Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
1
AVERT Final Report Section 4.1 Combined
Issue 2 - 28 July 2015
Executive Summary
Terrorism threatens horrific loss of life, extensive disruption to city transport and damage to
commercial real estate. Vehicles provide an ideal delivery mechanism for IEDs (Improvised Explosive
Devices) because they can be meticulously prepared and deployed into the area of operations.
Blocking vehicles can prevent IED Disposal (IEDD) robots from accessing threat vehicles. Current
manual methods of extracting blocking vehicles from restricted access height areas are inadequate
and expose operators to unnecessary hazards.
The Autonomous Vehicle Emergency Recovery Tool (AVERT) project has, over 3 years, researched
and developed a tool for Police, Security Forces and Emergency Services to deploy a unique
capability to remotely remove blocking vehicles from access-denying positions in confined spaces,
tunnels, low bridges and underground car parks. The opportunity is taken to automate the task,
removing operators from the danger area and speeding up the times currently taken to give access.
The AVERT programme has addressed the development of a fully autonomous system which
operates alongside existing and envisaged IEDD/EOD robot technologies to move blocking vehicles
and provide safe, quick access for the disposal robot.
The project team worked throughout with reference to user views from different national IEDD
traditions and approaches. A representative user panel was established to guide the project on how
this new capability should be specified in order to replace manual methods effectively. Interviews
and workshops with the users shaped the Key User Requirements across the domains of Planning,
Situational Awareness, Monitoring, Interoperation, Communications and Information System,
Management, Security, Usability, Deployability and Sustainability. Operational mission tasks were
explored and representative use cases synthesised to provide practical test and verification
scenarios. The tactics, techniques and procedures (TTPs), to successfully deploy and undertake the
mission tasks, were explored using robotic simulations (Webots) to aid a common understanding
during concept development.
The AVERT approach is built on the concept of a swarm of four individual lifting devices being tasked
with moving to a user-identified blocking vehicle, each moving under the car to dock on one of the
four wheels. Then, in unison, they lift the vehicle and transport it to a user designated location which
will then leave an access space to allow the IEDD robot to be brought to the threat vehicle and
undergo its normal disarming or neutralising activities.
The four lifting devices have to run under the car and move sideways to dock from the inside so as to
allow the vehicle to be to be moved without any additional length or width constraints. The solution
2
was to use specially designed low profile mecanum-wheeled bogies which allow omnidirectional
movement even under severe loading. To demonstrate the concept the bogies were mounted on a
specially designed deployment unit (DU) which could be towed to the operating area and, through
mast mounted sensors autonomously prepare a 3D survey to present the scene to the operator who
could designate the vehicle to be moved and its destination. This information allowed the insertion
path for the unloaded bogies to the selected car and the extraction path for the loaded bogie-car
combination to be automatically generated. Once the paths are approved by the operator the bogies
are deployed from the unit and the vehicle is docked onto, lifted and extracted autonomously.
Usability has also been at the forefront of AVERT, providing console tools for environment
visualisation, sensor readings and controls to command fully autonomous or manual controlled
mission tasks.
The AVERT Website (www.avertproject.eu) was established for all interested parties to contact
AVERT partners. The project has had to manage security risks and constraints throughout and so the
website does not detail full operational details.
The Final Demonstration of AVERT followed two Use Cases including a Timed device and Victim
Operated IED. All objectives have been achieved and a successful Final Demonstration was given to
the Users, Exploiters and EU representatives in March 2015 at the Force Ware GmbH premises near
Stuttgart, Germany.
AVERT has delivered an IEDD tool which can be operated either autonomously or under remote
control, removing the need for users to work within the threat area. AVERT has exceeded
expectations on power management and can achieve the validation Use Cases within 30% to 50% of
the time taken by manual methods. The consortium is exceptionally grateful for the enthusiasm and
input from the User community. AVERT has received high praise from them at the User workshops
3
Summary description of context and objectives
Research and Development Objectives
The AVERT project was aimed at addressing a class of problems faced by IEDD teams when tackling
threats where operating access is denied to IEDD robots due to blocking vehicles. These problems
occur when trying to move the blocking vehicle in confined urban spaces such as underground or
multi-story car parks or other urban locations with restricted access which means conventional
remote controlled lifting and handling equipment cannot easily operate.
Currently the main techniques used are based on manual methods such as hook and line where
operators have to enter the danger zone for extended periods to set up the extraction path using
pulleys and strops and physically make attachment to the blocking car. This is time consuming and
unnecessarily risky.
The main objective of AVERT is to provide access for the IEDD robot through remote removal of a
blocking vehicle using an autonomous tool.
The research approach is based on the development of a lifting bogie system which is capable of
travelling under the blocking car either from the end or the side and splitting and docking onto each
of the road wheels so that the can be lifted and the vehicle moved along a safe path (not hitting the
threat vehicle) far enough away to allow the access required, and in a location which does not block
further operations.
The programme was executed in two phases. Period 1 of 18 months covered the research into the
requirement and development of an A-model for proof-of-concept trials at the interim review point.
Period 2, originally of 18 months but extended to 21 months covered the re-design of the A-model
and the development of a demonstrator system capable of being taken forward to exploitation.
It was important to interact with potential users from the very start of the programme in order to
capture the key requirements from different countries and traditions. The objective here was to
ensure that the research delivered a tool that enabled the wide range of IEDD approaches currently
used by different nations in Europe to be accommodated.
The user workshops helped develop and prioritise the user requirements within the broad bands of
Key User Requirements (KUR) priority 1 requirements and priority 2 requirements. These were
developed and verified against a series of illustrative use cases developed by subject matter experts
to be representative.
It was felt in the initial proposal that AVERT as a system should be highly autonomous in each of its
deployed actions. This was in order to reduce manual intervention and also to allow for the user to
conduct parallel activities which may be needed for speeding up the overall time taken to deal with
the threat.
During the research the user panel also placed great emphasis on providing remotely controlled as
well as autonomous capability in order to have an ability for direct intervention as a reversionary
capability.
4
The autonomous strategy for AVERT design was developed between the AVERT consortium partners
to provide an integrated system solution comprising :
1) Command Sub-system (CSS) responsible for operator interface, display and processing of
autonomous survey and mapping and also path planning and overall AVERT management;
2) Deployment Sub-system(DSS) responsible for the autonomous management of the functions
aboard the towed Deployment Unit (DU) , a supporting platform which is used to carry the bogies
from the command point and actively deploy them in the operational area. The DU also provides the
mast mountings for the survey and tracking sensors and the command communications, together
with the housing and power for the processors used by the other subsystems.
3) Bogie Sub-system(BSS) responsible for managing the four split-bogie units both on the path from
the DU to the target vehicle and once the , deployed as two locked pairs from the DU to the form-up
point prior to travelling under the vehicle to split and dock.
The main objective of the CSS is to provide the autonomous control of AVERT under operator
command, requiring development of autonomous sensing and processing of survey and mapping
information, including information from other sub-systems, together with an effective monitoring
and intervention workstation/GUI for effective User MMI.
The approach is based on overall ‘Global’ sensing of the operating environment to the command
sub-system, mapping and path planning. with updates of ‘Local’ obstacles, as discovered by the
bogie sub-system. The bogies operate under a closed loop system, measured and corrected within
the Global coordinate system, to safely rotate and move the vehicle for extraction in any direction,
with synchronised movement of the bogie swarm.
The DSS is used to provide the automatic control of the bogie deployment mechanism on the DU
and provide status and monitoring information for the power system and movement sensors
supporting the processors from the other sub-systems that are mounted on the DU
The four specially designed and mechanically robust lifting bogies are controlled via the BSS with a
coordinating BSS processor managing the overall tracking and swarming and embedded processing
on each bogie to manage the low level control of the drive, energy storage mechanisms and on-
board sensors. It is this configuration which enabled the AVERT to operate within the severe volume
and height constraints imposed by the vehicles they are intended to lift and move
The bogie elements represent the most complex section of the AVERT tool and design and
development was tackled in two Proof of concept was achieved using an full scale A-model early in
the programme. The results from this led to a significant design review of the bogie structure at mid-
project point. After analysis and further modelling appropriate modifications in geometry and
algorithims were applied to the build of the final demonstration bogies.
The system is structured so that a hybrid reversionary mode requested by the user panel can be
used. In this mode the autonomous mapping and planning is by-passed so the bogie sets can be
guided to the car by remote control. Once at the form up point the autonomous wheel detection
and docking/lifting can be activated, and then with the individual bogies still acting as a swarm, the
whole assembly can be moved as a single entity under remote control to the desired location.
5
The specialist mechanical engineering design, load testing and material selection for the mecanum
wheels, resulted in a very high load bearing capacity, given the low profile dimensions. This was a
highly significant achievement on this project and could, in its own right, be exploited.
The rapid survey and mapping capability was also a significant development in its own right. Specific
developments in fusing algorithms and path planning resulted in a very rapid location and mapping
capability which has led to a number of conference and research papers and is also capable of
further exploitation beyond AVERT itself.
The development objectives of AVERT were to deliver an IEDD tool which is effective in providing
access in urban situations, capable of operating with the expected range of vehicles and remains
safe and recoverable when communications are interrupted. Such a system, which provides faster
operation and removes Users further out of harm’s way than with current techniques has resulted
from this programme. AVERT has exceeded expectations on power management and can achieve
the validation Use Cases within 30% to 50% of the time taken by manual methods.
Dissemination and Exploitation Objectives
Dissemination and Exploitation objectives have been addressed throughout the programme. The
team has been active in raising user awareness by presenting AVERT at targeted conferences and
exhibitions throughout. In addition a number of academic papers on the underlying technology and
algorithm development have been prepared and published on peer reviewed open access sites.
The AVERT Website (www.avertproject.eu) was established for all interested parties to contact
AVERT partners. The project has had to manage security risks and constraints throughout and so the
website does not detail full operational details.
In addition the objective of wider dissemination has also been tackled through media outlets such as
Youtube video (300,000+) hits, INTERSEC magazine article and featuring on mainstream technology
shows such as BBC Click.
The dissemination also supports the exploitation objective, which has developed over the life of the
project. Initially potential exploitation partners were embedded in the consortium itself. The
business goals of these partners changed during the life of the project so they were not available to
directly market and exploit. Consequently a new exploitation plan with a revised set of objectives
has been developed.
The consortium has charged the coordinating partner, IDUS with heading up the search for, and
negotiation with, alternative exploiters capable of taking on board the current AVERT demonstrator
and developing it into an exploitable production unit within the next year.
Talks are currently in progress with suitable exploiters and identification of appropriate funding
sources has commenced with good prospects of meeting this objective.
6
Description of main S&T results/foregrounds
Summary
The AVERT project has identified and characterised the need for, and benefit from, developing a tool
to assist in IEDD operations by lifting and moving blocking vehicles so that IEDD robots can gain
access to deal with a discovered IED threat.
The main science and technology achievements commenced with the use of soft sciences, such as
Operational Analysis and Human Factors Research, to generate and consolidate the user and system
requirements across a wide multinational-user base with different cultures and national structures,
including organisation and training expectations.
The project then addressed the hard science areas of physics and maths for modelling and algorithm
assessment, whilst concurrently operating in the technological areas of the engineering sciences;
mechanical, electrical, electronic, computing etc to achieve a practical, demonstrable and trialable
system capable of being taken forward to exploitation.
Results
The results are measured in terms of the development of a system which meets the clear majority of
the KURs and Priority 1 Requirements and is technologically ready for development through to
exploitation (TRL 6 or above).
The verification and validation carried out against the requirements and detailed in the Period 2
periodic report has shown that 98.8% of the KURs were achieved and a total of 80.2% of all
requirements were reached at the demonstration stage.
The technology readiness levels achieved have also been impressive:
Bogie Mechanical Hardware - TRL 9
Electronics &Sensors TRL 6-7
Control & Swarming Software TRL 6-7
Survey & Mapping Software TRL 7-9
System Definition - WP2 Lead IDUS
The primary objective of this WP, planned entirely during Period 1, was to complete the overall
System Definition for the AVERT Tool. The sub-tasks address the overall system concept, the
preparation of Strawman user and system requirements, and the refinement of these with specialist
support to result in agreed user and system requirements. The final sub-task was the preparation of
the overall system design for the demonstration system, based on the requirements.
The WP was led by the IDUS team utilising skills in Operational Research, Systems Analysis,
Mechanical and Manufacturing Engineering, Electrical and Electronic Engineering as well as
Technical Management experience. The programme was arranged so that the requirements work
could be undertaken concurrently to programme research in order overcome project time
constraints.
7
User and system requirements were developed through: User questionnaires and Stakeholder
interviews; derivation of requirements by Systems Analysts; drafting of Strawman documents; and
development of Use Cases for analysis at the first User Workshop to define goals for project
acceptance. Finalisation of key deliverables included the User Requirements Document (URD),
Systems Requirements Document (SRD) and, derived from these, the detailed AVERT System Design
Definition document (SDD).
Five use cases were initially identified for detailed discussion and review at the User Working Group.
From these two (a) Victim Operated IED and (c) Timed Response were eventually confirmed as the
basis for the main operating requirements. In addition a further use case (e) Dirty Bomb in Extremis
was used for the development of stretch requirements. The impact and importance of autonomous
control and removing the user out of harm’s way was fully realised.
System Design Definition (SDD)
The SDD development provided the document describing the system design vision and identifying
the system and sub-system architecture and design concepts. The SDD was used as the design basis
for the A-Model proof of concept bogies (just one pair of experimental split bogies ) and also the
survey and mapping approaches (using fixed tripod mounted laser and camera). Normally the SDD
would have been updated following proof of concept review but, due to portal upload rules, a
separate Demonstrator Design Definition (DDD) had to be produced instead.
For the SDD the system was coordinated by IDUS using inputs from all the consortium partners. In
particular ZHAW, in consultation with BBI and IDUS, undertook mission task and sensor analysis,
which required assessment of local obstacle recognition (for avoidance and navigation), path-
planning and bogie position measurement. DUTH addressed the sensor strategy for mapping and
global path-planning, taking account of possible emission restrictions emerging from the
requirements exercise. The AVERT strategy was developed as follows:
a) The bogies were to be deployed from the DU and instructed to move to a Form Up Point (FUP) in
front or to the side of the vehicle to be moved.
b) The bogies would then go underneath the vehicle, locate the axles, separate and dock with the
vehicle wheels.
c) The bogies, acting as a swarm, would lift and extract the vehicle along a pre-planned path.
Each step of the operation required sensors to detect sub-surface, surface and overhead obstacles,
both locally and globally, for safe navigation when operating as joined bogie pairs, as individual split
bogies and also when working autonomously as a bogie swarm. The obstacles under consideration
were those typical of the scenario and were considered to be static during the mission. The
algorithm was to be capable of being extended to dynamic obstacles if required during exploitation.
The appropriate choice of sensors was essential to meet the User Requirements such that obstacles
could be identified and avoided, the IED threat would not be activated, and also so that AVERT could
be manufactured at a competitive price, within the budget of EOD organisations. For classified
operational reasons the users placed constraints on the use of certain active sensor options.
Consequently the system design was developed in a way that different sensor types could be used
8
on a modular basis. This will assist in tailoring AVERT during exploitation to overcome particular
threats and also allow more cost-effective sensors to be used in certain scenarios.
The extent to which users also wanted full observation, intervention and manual override at all
times was greater than originally expected. It is believed this was due to user's current reliance on
hands-on remote control with existing equipment and has indicated the need to build trust in
autonomous control under user command supervision.
The AVERT research has been predicated on the provision of autonomous capability and this
continued as the prime objective. The architecture was developed so that manual reversion
capability could be incorporated at the exploitation stage but the formal demonstration of the
AVERT research achievements was confined to the autonomous operation itself.
The SDD also explored two design approaches for the deployment unit: towed or powered, self-
propelled remote-controlled. These were reviewed at concept level at the interim review but
selection and build was not required at this stage to support the proof of concept activities.
Both options contained the autonomous bogie deployment system and the review recognised that
the powered, self-propelled remote-controlled DU approach would not demonstrate further
autonomy. Consequently the cheaper and lower risk towed unit was selected for development at
the demonstration. This would not affect the option to provide a powered DU as an alternative for
exploitation.
Demonstrator Design Definition (DDD)
The DDD was refined from the SDD following proof of concept review by EU, the project team and
the users. It included the changes decided after a series of options studies arising from the review,
which were developed in a small, 3 month, extension to the programme timeframe. This resulted in
refinement of the underlying architecture, revision of the bogie design and greater definition of the
DU and the AVERT command console/GUI.
The refined architecture was still based on the three sub-systems (CSS, DSS and BSS) and their
intended interfacing is described in the DDD. Physically the system comprised the following key
hardware components: the detachable AVERT Command Console (ACC); the towed DU; and the 4
split-bogie Units (BU), deployed as two joined bogie pairs.
The requirements, with associated User objectives and acceptance criteria were re-visited, and the
primacy of two of the Use Cases (UC): UC1 “Victim Operated IED (VOIED)” and UC3 “Timed
Response”, with stretch requirements identified by UC5 “Dirty Bomb in Extremis” was confirmed,
with additional operational advice.
Command & Control System WP3 Lead IDUS
Overview
The Command and Control system was designed to service the architectural concept based on the
three functional subsystems, CSS, DSS and BSS and the work package was specifically tasked with the
development of the CSS and DSS by DUTH and IDUS and their linkage to the BSS developed by
ZHAW. Key system architecture components included the communications systems, video relay to
9
the User, onboard processing for each functional subsystem, and command and control of mounted
sensors as well as mobility and actuators.
During development there was continued concern regarding the integration of the sub-systems and
a number of activities were instigated to reduce the inherent risks in this area.
To assist risk mitigation a ‘Command System Distributed Integration Rig’, template was developed
and utilised using laptop emulation at IDUS and DUTH for use as a system integration support tool
and later acted as part of the validation environment for system upgrades. It was originally planned
to network the two elements together over the internet but corporate security policy constraints at
DUTH prevented this, instead files were exchanged and emulators built to represent the missing
elements.
CSS - DUTH/IDUS
The CSS supported mission planning and command decisions by the User, presented status and
monitoring information to the User as well as provision of requested supervision camera and sensor
views. The design philosophy was to enable the user to intervene and facilitate reversionary control
mode and emergency stop commands at any time.
The CSS design was guided by a detailed Concept of Use (CONUSE) presented and developed at the
Second User Workshop in October 2013. This showed how the User would interact with AVERT and
EOD equipment in a range of operations or scenarios In accordance with the CONUSE, the two-
position survey method, developed by DUTH which required the User to pause at an intermediate
point before towing the DU to its final overwatch position was agreed and incorporated.
Robot simulations of Use Cases 1 and 3, developed by DUTH during Period 1 in the Webots tool,
were employed during the Period 2 User Workshops 3 and 4 as well as the Final Demonstration to
help visualise and evaluate how AVERT would be deployed and understand the constraints of the
environment for which it has been designed.
The command and control system itself was researched during period 1in terms of communications
capabilities from a mix of off the shelf and bespoke solutions. This included research into IP
compatible trunk and local communications and antenna solutions specific to under-car operations.
It included the development of a novel narrowband 800MHz safety and reversionary control link
tailored to the swarm.
The communications solution was selected built and integrated during Period 2, initially using a
simulation facility and subsequently being integrated into the full system for final testing. At
exploitation most users would wish to use their own trunk communications system for compatibility
and security purposes so it was ensured that the design used in the demonstration was open
architecture enabling a user's alternative system, e.g. COFDM to be easily substituted.
The CSS was developed to meet the following criteria:
a) responsibility for survey and mapping including control of relevant sensors,
b) responsibility to support mission planning and command decisions by the user,
10
c) obligation to receive obstacle update data for path re-planning from BSS,
c) responsibility to provide path movement and re-planned path movement data to the BSS both for
bogie move to the FUP and swarm extraction of the vehicle to the selected location,
d) responsibility to present status and monitoring information to the user, including data from BSS
and DSS,
e) responsibility to manage the DU mounted survey sensor
f) responsibility to provide requested supervision camera and sensor views to the user
g) responsibility to enable a user to intervene and facilitate reversionary control mode and
emergency stop commands at any time.
CSS Hardware
At the mid-project stage the AVERT Command Console was represented by partially-integrated
console components using discrete screens and laptops. The keypad and joystick were commercial-
off-the-shelf with USB connectivity.
A design scheme was developed for a portable integrated command console in terms of processors,
screens and controls as well as the implementation of the open-source ROS (Robot Operating
System) framework for integration. However it was found in Phase 2 that several of the potential
exploiters involved in preliminary discussions already had suitable console approaches implemented.
It was eventually considered that there was no additional research benefit in developing a more
sophisticated stand-alone console at this stage and so the non-integrated arrangement was carried
forward to the demonstration.
CSS Software
The command system was developed to command autonomous AVERT operations under user
supervision. Separate mechanisms provided video surveillance and emergency stop so that the
command system software did not need to be developed under safety-critical criteria. The CSS had
two processing nodes, one represented by a laptop on the DU managing the sensor processing and
the other, represented by a laptop at the command console managing the GUI.
A command system integration rig, with representative processors, displays and communications
links, was used to provide the basis for trialling elements of the AVERT hardware and software
before they were integrated. It featured the ability to emulate primary operator command and
control inputs and information exchange between CSS/DSS and other system elements.
The command console provided the tools for environment visualisation, sensor readings and User
control input. Multiple parallel processes were required for communication, graphic visualisation of
scanned point cloud data, commands and sensor data processing, as well as User selected solution
processing employing obstacle avoidance algorithms.
The Graphical User Interface (GUI) was developed within ROS utilising the Qt framework, which
enables the User to interact with and examine the ROS environment in a visual manner. The
11
workshops proved helpful in allowing the DUTH team to customise the GUI and widgets contained
within the toolkit to meet the Users’ requests and feedback. From the assessment of the User
Requirements detailed in WP9, the GUI and User Controls exceeded User expectations and have
been a significant achievement under this project.
The GUI activity continued in period 2 in order to simplify the AVERT User experience. The Second
User Workshop provided the forum for discussing User interaction, and provided options for trials.
The User required a relatively simple and graphical diagram interface for selection of bogies, camera
view as well as bogie and DU status information. The bogie release mechanism on the DU itself was
also part of the GUI.
CSS-BSS Integration
CSS-BSS integration was a difficult task throughout the programme as it required assured interfacing
between two software systems each of which was continuously evolving. Initial agreements were
made at integration meetings in June 2013. These were reviewed and challenged during the
subsequent periods in a series of meetings, workshops and teleconferences held between DUTH and
ZHAW and chaired by IDUS. The workshops addressed integration clarification between the three
sub-systems and sought mechanisms to identify a practical way forward. The issue was considered
high risk throughout the programme and continued despite using various risk reduction measures.
Information exchanges considered for communications, sensors and actuators as well as planning
and situational awareness data continued to raise difficulties through the remainder of period 1 and
during period 2 despite several attempts to address it by the universities. One of the problems was
the late availability of BSS software which was a knock-on consequence of the delayed demonstrator
bogie design and subsequent trials. This affected the initial mitigation plan to overcome the
difficulties experienced by DUTH and ZHAW in using the original interface specification given in the
DDD to develop and prove the CSS software (built on a ROS framework) interfacing with the BSS
software (developed from a previous non-ROS application) using the rig.
As a neutral compromise IDUS proposed taking the interfaces as far as they had been developed and
preparing a separate ROS/non-ROS (RnR) transfer application to sit between them and provide two-
way communication between the sub-systems. The RnR application itself was built and pre-tested in
UK on the rig. Then the IDUS team brought it to the demonstrator system integration preparation
and worked directly with the software researchers from DUTH and ZHAW to integrate and test the
application with the latest builds of the CSS and BSS software in the run-up to the demonstration.
This was finally achieved during the last phase of integration and was successfully used at the
demonstration in March 2015 to show the operation of a fully autonomous sequence from initial
survey through to vehicle extraction and placement.
CSS Trunk Communications
The original communications approach was superseded by improvements in off-the-shelf technology
and, following communications trials in period 1 and 2, a modern Multiple-Input:Multiple-Output
(MIMO) based Wi-Fi system was found to provide range and resilience of equal or superior
performance to the original system for trunk linkage at the demonstration stage, and at much lower
12
cost. The COTS Wi-Fi system would be expected to be replaced by a modern COFDM system
compatible with the particular user's other robot controllers in most exploitation circumstances.
The Final Demonstrator also showed integration potential for a dedicated backup communications
system for safety and operator reversionary control, as well as sufficient bandwidth to support laser-
based survey and monitoring functions.
The reversionary communication system was designed to support a simple fail-safe and easily-
evaluated emergency stop system over a separate communication link in case of failure of any of the
primary control elements.
DSS
The DSS was originally considered as the centre of autonomous control, as the major processing
burden was expected to be located on the DU. It would have supported real-world mapping and
mission monitoring as well as processing situational awareness of the AVERT Area of Operations
(AOA) where the bogies are operating.
Following review, these activities were transferred to the CSS and BSS as appropriate, with separate
BSS and CSS processors (located on the DU), each managing the associated mast mounted BSS and
CSS laser sensors respectively.
The DSS was re-purposed to cover the specific DU control & autonomous actions. As a consequence,
a network-connected microcontroller was found to be the most appropriate solution for the DSS
processor. This bespoke unit was designed to manage the CSS commanded autonomous pneumatic
deployment sequencing system, the off-tow parking brake and the on-tow laser sensor guard
(deployed to protect the lower sensor during movement). It was also programmed to monitor and
transmit the DU battery status and odometry sensor readings to the CSS for use across the system.
BSS - ZHAW
The BSS controlled the bogies in joint bogie pair, split-bogie pair and swarm modes in pre-
programmed sequences commanded from the CSS.
a) Joined - For moving to the vehicle in joined mode, the CSS managed the relevant Sick laser based
tracking sensor. When joined the BSS received the planned or re-planned path from the CSS.
b) Split - When the joined bogie split for docking, the bogie on-board sensors operated an
autonomous sequence which detected the wheels to be lifted and split and guided each bogie to the
correct wheel where other onboard sensors managed the lifting.
c) Swarm - Once the vehicle was lifted, the bogie swarm could move a vehicle in any direction. The
actual processing was coordinated at high level by the main CSS processor located on the DU with
embedded pic processors on each bogie, which managed the low-level motor and actuator actions.
The BSS was required to sense local obstacles, in the vicinity of the bogies, such as concrete pillars
and kerb stones, whilst updating these detected obstacles to the CSS so that the path could be re-
planned and re-issued if necessary.
The BSS development is described in the report for WP8.
13
Bogie and Deployment Platform Design WP4 - Lead BBI
Overview
The mechanical design of the A-Model and Demonstrator bogie sets (and also the DU), together with
support for design changes during manufacturing and assembly of all the units, was undertaken by
BBI in this work package.
The bogies, being the key research component, were subjected to an iterative design process with
an initial A-Model design for one pair of bogies as proof of concept being built for mid-project trials.
The results of these were subjected to extended review and a number of alternative design concepts
were developed and debated as a result. Whilst this had significant impact potential on the project
timeline, it did mean that the eventual design for the demonstrator bogies, and especially the high-
load mecanum wheels, was well considered. This opportunity to incorporate the key lessons learned
ensured that the fundamental mechanical design operated effectively during the rest of the project
at a high TRL level, ready for exploitation.
The mechanical load testing, in support of the bogie mecanum wheel redesign, and selection of a
harder rubber material for the ‘tyre’ rollers, was undertaken by FW (this testing had to be
undertaken with Period 2). Design of a bogie system that can lift a large vehicle, such as a 4x4,
including potential overload, using mecanum wheels of such small diameter that could be inserted
underneath any vehicle, was a highly significant achievement on this project and could (in its own
right) be exploited.
A-Model Bogies
The A-Model bogies were designed using 3D CAD and the engineering calculations were carried out
using the CAD package itself, supplemented by load, drive and braking calculations executed in an
Excel-based model. A commercial off the shelf FE package was used for key stress calculations.
Figure 1 - A-model Bogie Design
14
Even at A-Model, the design output was in the form of detailed manufacturing drawings for all the
specialist parts including casings, gears etc. This enabled the purchasing, manufacturing and
assembly procedures to be "dry-run" at the mid-point of the project, in order to reduce logistics risks
for the manufacture and assembly of the demonstrator bogies later on. After the design phase,
design support was provided for impending mechanical design modifications arising from integration
and trials of the A-Model.
From the outset, the design concept for the split bogie incorporated the idea that the half bogies
would rotate and join to form the full bogie. This clever approach means that a single design of half
bogie could be used for all four units in a set, which gave several cost and time advantages during
production. This will be a significant consideration when moving to exploitation.
Demonstrator Bogie Design
The Demonstrator Bogie design commenced with a major consortium-level review. The A-Model
bogies had been successful in proof of concept, but the mechanical layout was questioned in terms
of geometry and steering control in split mode. A comparison and assessment of different Bogie
mechanical design alternatives against System and User Requirements was conducted to optimise
the mechanical system design approach.
User requirements from the first User Workshop for a ‘free-wheel-functionality’ placed an additional
challenge to the bogie design. A redesign was carried out so that, in the event of a severe system
failure, this new capability would allow users to manually push or tow the bogies, with a vehicle
loaded on top, towards a safe area. Furthermore, a more detailed investigation of tyre sizes and
vehicle dimensions showed that the anticipated bogie height had to be reduced to 100mm to cover
the full range of vehicles which may need to be moved if they are built to European ground-
clearance guidelines.
As a result of earlier load testing, a revised mecanum wheel with harder rubber plastic and
redesigned spindles was developed and successfully tested using the A-Model structure but, due to
long lead time for the rollers, this could not occur until the second period. This design, used in the
demonstrator, enabled AVERT to be capable of lifting a large vehicle (such as a 4x4) including
potential overload, using mecanum wheels of such small diameter that they could be inserted
underneath any vehicle.
The final demonstrator design provided for braked-in-line omniwheels to assist the steering in split
mode and utilised the upgraded mecanum wheel components. It also provided a lower-profile cable
management arrangement and revised electronics housing, based on trials of the docking and
forward sensing laser positions using the A-Model. The bogie ground clearance itself was raised from
10mm to 15mm, to overcome potential grounding on some floor types commonly used in parking
areas. This forced the bogie’s overall height also to rise from 100mm to 110mm. This has a knock-on
consequence that some extremely low cars may not have enough clearance to be underridden.
During development for exploitation it is proposed that a trade-off study will be made into the
optimum bogie height for clearance and load-bearing across the population of vehicles likely to be
encountered.
15
Figure 2 - Demonstrator Bogie Design
DU Design
The DU design was not considered as critical as the bogies from a research viewpoint. To reduce
project cost and risk, a simple towed platform was prepared and operated as a proof of concept
model. From an exploitation viewpoint the DU design would be developed to suit the deployment
strategy of the relevant force, whether passive-towed or powered remote-controlled.
Figure 3 - Towed DU Design
16
The basic structure was designed to illustrate deployment within a typical operating constraint,
passing through a narrow doorway. The resulting design was provided with an autonomous
pneumatic lifting and lowering mechanism controlled by the DSS processor. This demonstrated the
proof of principle sufficiently, but would have needed a re-design for greater lifting power during
exploitation, due to the weight increase of the bogie units themselves following modification. The
design also incorporated a dual-mast structure, separating the sensor head from the
communications mast in order to reduce system interaction risks.
As with the bogies the DU framework was developed using Computer Aided Design (CAD) and also,
in line with the bogies, key drawings of the DU were reviewed by the AVERT Consortium Partners to
form the demonstrator manufacturing data-pack. Some detailed elements were omitted, including
some cable runs and type and location of some electronics enclosures. This was due to the
concurrent development approach used for the DU, where enclosures and wiring were being
resolved quite late due to knock-on impact of the delays in the overall programme. Late design
changes were captured in integration documentation and photographs.
Deployment Platform Development & Fabrication WP5 - Lead FW
Overview
The primary objective of this WP was to develop, manufacture, assemble, integrate and test a
representative deployment platform. Originally this would have taken place at the premises of an
experienced platform integrator (MSDG), which was unfortunately not available following its
withdrawal for business reasons earlier in the programme. Consequently the manufacture was
hosted by FW in the same workshop as that where the bogies were fabricated and the integration
was supported with platform and sub-system expertise from IDUS, BBI and DUTH.
Within WP5 the DU was built by FW incorporating sub-system communications and sensor-mast
units developed by IDUS and DUTH respectively. The frame manufacture at FW followed on from the
demonstrator bogie build, and hence occurred slightly later than planned in the original DOW.
The DU was designed with a forward sensor mast and a separate rear communications mast. The
design used readily-available COTS components to minimise procurement and assembly issues, and
the separate mast approach meant that the units could be trialled as sub-assemblies prior to DU
integration in Eningen (Germany).
17
Figure 4 - DU with Forward Sensor Mast Design
In practice the late assembly of the DU due to other programme issues meant that DU integration
and overall demonstrator timeframes overlapped and the pragmatic decision was made to co-locate
the IDUS and DUTH teams with FW for final integration at both the DU and overall system level. The
result was a more tailored DU integration than would have been possible with multi-site working.
However, as there was only one DU there was a need to forgo certain platform details in order to
concentrate on the overall system needs. None of these areas impacted directly on the delivery of a
successful demonstration, but a number of issues were noted which would require further re-design
when going forward to exploitation.
Bogie Module Development & Fabrication WP6 - Lead FW
Overview
Work package 6 covered all manufacturing and development tasks for both the A-Model bogie and
the demonstrator bogie units. Both tasks were undertaken by the same participant in order to
include lessons learnt experience from A-Model manufacturing into manufacturing of the
demonstrator bogie units. In addition, a mechanical performance test to verify the A-Model
capabilities for lifting and displacing a vehicle was performed within WP6, coordinated and
conducted by FW.
18
Figure 5 - Bogie Mechanical Load Test
It was originally planned to build two complete sets of bogie units at FW, one for use by FW and the
other for MSDG during subsequent exploitation activities. With the withdrawal of MSDG, it was
agreed with the EU that just one full set of demonstration bogies and a spare sub-set would be
needed. This resulted in six, rather than eight, split bogies being built. This approach allowed the
diversion of planned resource to support production cost-benefit analysis for exploitation planning.
The manufacturing practices and procedures for AVERT bogies were developed from the normal FW
procedures for equipment and detectors. They had to accommodate a much greater mechanical
complexity and wider range of specialist subcontractors than normal, in order to procure parts for
and built a robotic system of this nature. The approach was trialled with input and re-design
flexibility from BBI in support and the A-Model was successfully constructed.
Following mechanical redesign for the demonstration bogies, these were assembled in the same
manner and with the same suppliers as the A-Model. It was possible to make preliminary estimates
of assembly time and material cost from this exercise, but this has to be heavily caveated when
extrapolating for exploitation, as there were a number of time and cost factors imposed by the
nature of the programme which could be significantly reduced once the design is productionised and
batch sizes are optimised during exploitation.
Two complete sets of Demonstration Bogies were delivered by FW, on 14th November 2014, to
ZHAW for integration and trials. The third set was assembled by FW in time for the Genoa CP Expo
exhibition in December 2014.
19
Figure 6 Demonstration Bogie Set
Simulation & Driving Trajectory Calculation, Synchronised Platform WP7 - Lead DUTH
Overview
The primary objective of this WP was to research and develop strategies and path-planning
algorithms for the safe and fast movement of bogies: a) from the DU to the Form-up Point (FUP: a
User-defined end-point near the blocking vehicle); and b) carrying the lifted blocking vehicle from its
original position along the extraction path to the user-designated final position. These strategies
were simulated in a practical context using the level of point cloud data typically obtained from an
AVERT DU deployment as the basis for navigation.
WP7 was also originally planned to deliver the global navigation for the trajectory, with local
navigation being developed in WP8. Following integration discussions with IDUS and ZHAW towards
the end of Period 1, the local navigation to support obstacle avoidance during transit was not
considered by ZHAW to be covered by WP8. Consequently it was proposed by the consortium, and
accepted by the EU, that there would be an additional sub-task in WP7 during Period 2 covering the
integration of Global and Local trajectory planning to support obstacle avoidance.
The global survey was undertaken using the output from the sensor head on the DU. This head
comprised a vertically scanning SICK LMS500 PRO 2D laser scanner co-located with a Prosilica
GC780C camera mounted on a Schunk PW-70 PTU. To avoid blind spots, at least 2 scans were taken
from different locations and then their point clouds merged using algorithms developed by DUTH
and modified for the AVERT problem by using inferences to rapidly speed up registration time by an
order of magnitude.
The resulting 3D representation of the area was then transformed to a 2 ½D map (2D map with
knowledge of height constraints) for planning the bogie approach and the vehicle extraction. The
20
path-planning algorithm itself was based on improvements made to the D*Lite fast path-
planning/re-planning algorithm for efficiency and speed.
These fast computations provided significant advantage in the time needed to survey and plan
AVERT deployment, resulting in fast-into-action times even in previously un-surveyed locations.
Additionally the underlying speed, and the dynamic re-planning ability, made it possible to bring in
additional obstacle information during transit and re-calculate and re-issue an updated path from
the CSS to the BSS during transit without affecting extraction times.
Planning and re-planning development
DUTH commenced with design and development of the path-planning algorithm itself using discrete
and then Webots simulations prior to the sensor head trials. The path-planning algorithm itself was
selected for development after a literature survey. D*lite was chosen as a fast path-planning/re-
planning algorithm suited to partial or unknown terrain. It was then implemented in a programming
language and parameters modified experimentally to enhance speed of execution. The next step
was to integrate it into the simulation environment.
Figure 7 - Discrete Path Planner Development
Global mapping
The second simulation stage created the global mapping function and used it with the path-planning
in a simulation environment. The approach replicated typical laser scanning from a sensor mast
mounted on a DU equivalent platform. Then the point-cloud-based mapping from this was
converted to a two-dimensional (2D) map, working with the path-planning algorithm in a simulation
of the AVERT situational assessment and planning mode. The point-cloud mapping was undertaken
from two separate static locations to build up the 3D dataset.
21
Figure 8 - AVERT Simulated in Webots
After the simulation, the path-planner was integrated into the overall AVERT system. Mainly this
included the appropriate compilation and definition of the dependencies in the ROS framework
running on the CSS processors located on the DU and with the GUI. The system itself was already
being integrated with the survey functionality from the sensor head.
Figure 9 - Example Sensor Head 3D Capture
Managing local obstacle avoidance
In the additional sub-task for integrating Global and Local trajectory planning to support obstacle
avoidance, advantage was taken of the CSS fast path-planning capability to overcome the issue that
ZHAW considered local obstacle avoidance during transit outside WP8 scope. It was agreed that the
BSS, which would have knowledge of its bogies' own positions in relation to the global map, would
report the position of the obstacles it had seen as it scanned ahead during transit to the CSS. The
CSS, which holds the global map, would then add the obstacle to the map and, if necessary, re-plan
the swarms' transit path and send an update to the BSS. The BSS would receive the update and alter
the trajectory accordingly. This would mean that the bogie swarm elements were still required to
22
operate under closed-loop control in the BSS such that their relative positioning was measured and
corrected within the global coordinate system to follow the overall transit re-plan around the local
obstacle.
A summary of the key achievements of this research includes:
a) The development of the SICK laser as well as the SCHUNK ‘pan and tilt’ software drivers;
b) Interactive software drivers to acquire a 360˚ scan from the SICK and SCHUNK units (completed
within 2 minutes, as demonstrated);
c) Software developed to accurately register successive scans for mapping the environment;
d) Development of the path-planning algorithm (detailed above);
e) Software to merge the mapping and acquired planning data;
f) Path re-planning in transit for local obstacle avoidance.
Bogie Local Control and Obstacle Avoidance Sensor System and Software WP8 - Lead ZHAW
Overview
WP8 covered the design, implementation and commissioning of the electronic hardware as well as
the bogie software controller tasks. This task was led by ZHAW and started in parallel to the detailed
AVERT Concept Design work. The autonomous requirement was for the joined bogies to travel to a
FUP (selected to the front or side of the (blocking) vehicle) using a path provided from the CSS,
detect the exact position of wheels, then insert underneath the vehicle and locate the wheel centres
prior to splitting and docking, i.e. position each split bogie's "jaws" either side of the wheel and
clamp together. Once each bogie was docked, they were each to be activated together to lift the
vehicle and, in a common ‘swarm’ mode, the bogies were required to move it along the planned
extraction path.
Software architecture/framework
The software architecture and framework for the BSS was developed by ZHAW from a previous
internal project using a modular but proprietary approach. This allowed the basic software structure
to be prepared relatively quickly and the specific algorithms for bogie dynamics to be researched
and integrated. The framework also facilitated the internal communication between the bogie
elements and provided a platform for different sensor option experiments. However this decision to
use a closed rather than open system did lead to integration problems with the open-source ROS-
based CSS later in the programme. These were exacerbated by the late development of the BSS
functions due to knock-on delays from bogie design and re-design.
Framework implementation and hardware abstraction
From an early stage the framework was implemented in a Linux laptop computer to simulate the BSS
processor which would be situated on the DU, with the tracking laser, and manage the bogie swarm.
The communications with the bogies were over a dedicated hi-speed Wi-Fi network using the same
type of MIMO communications devices as for the trunk system. On board the bogies themselves, a
23
dedicated PIC-based electronics board with Ethernet and other interfaces was developed to support
on-bogie sensing and control of the joining and lifting systems. The laser sensors and mecanum drive
motors/controllers on board each bogie were also Ethernet connected to the on-board switched
network and through that to the DU over the Wi-Fi.
Early mock-up and A-model
The concept was being developed from scratch so, ahead of the A-model bogies being available, a
heavy-duty bogie mock-up was built to provide test facilities for the servo motors and the bogie
kinematics using a real vehicle. An Opel Vectra station wagon was used as the test vehicle, both for
bogie algorithm development and as a basis for developing the wheel-detection algorithms.
The motors used in the mock-up were relocated to the A-model bogie set when it became available
and the A-model, which also included the lifting system, was then used to demonstrate the proof of
concept for joined bogie movement to the FUP, wheel detection from the FUP, moving under the
car, splitting, docking and lifting. As this was a single bogie set the other wheels were supported by
passive castored dollies (Go-Jacks) during trials.
The findings of the A-model trials resulted in the discussions regarding geometry and algorithms for
reliable movement of the split bogies from their joined position to their docking on individual
wheels. This resulted in the addition of an in-line braked omniwheel being added to each mecanum
wheel pair in the demonstration bogies. The knock-on effect was to delay the availability of the
demonstrator bogies for swarm tests and also to require late modifications to the already developed
bogie kinematics once the bogies were available, which delayed their availability for integration.
Swarming and sensor experiments
Once it was realised that there would be a substantial delay in the availability of the demonstration
bogies, it was decided that a second pair of the heavy-duty mock-up bogies would be constructed by
FW and ZHAW so that the delayed swarming trials could be de-risked. These bogies again used the
motors that would eventually be used in the demonstration bogies and were run using umbilical
Ethernet cables ahead of the wireless network being available. After this was built, it was possible to
prove the swarm concept and tune the software and algorithms using the trial vehicle in the
underground car park at ZHAW.
Figure 10 - Skeleton Bogies Used For Swarm Experiments
24
In addition to swarming, a number of sensor experiments were undertaken using the A-model
bogies to improve tracking, docking and obstacle avoidance for the demonstration bogies in time for
the findings to be fed back to the demonstrator bogie design.
The “Swarm simulation/emulation demonstration” delivered the functionality of bogie subsystem
for local navigation and swarm control. The tests successfully drove a vehicle on a swarm of four
independent bogies, following a trajectory, and demonstrated the complex constraints including the
physical connection of the bogies through the vehicle suspension. These tests were then repeated
with the demonstration bogies themselves, once they became available.
Bogie Tracking
The original approach by ZHAW was based on using a single tracking laser (SICK LMS511) mounted
on the DU to keep track of the bogies and the car wheels. This was amended, following the A-model
experiments, to allow for local tracking and obstacle sensing with a forward-facing upgraded
(version UST) Hokuyo laser sensor. This was mounted on a servo tilt mount to allow horizontal, fixed
angle and variable tilt scanning. The arrangement allowed a scan of the car underbody from the FUP
to detect the risk of impacting low hanging (below 110mm) obstacles before moving off. It also
allowed fixed angle fan scanning to allow detection of obstacles formed of adverse dips or hollows.
Figure 11 - Underbody Scan
For the demonstration, the DU-mounted SICK sensor maintained the tracking of the bogies relative
to the global co-ordinates up to the FUP. For the bogies moving from the FUP under the vehicle to
dock, the forward Hokuyo laser was used for wheel-centre detection and tracking.
A separate laser sensor was fitted to view the wheel area for docking and also to view and track the
partnering bogie in support of autonomous rejoin at the end of the mission.
The upgraded Hokuyo sensors became available during period 2 and gave significantly better
performance in terms of range, resolution and ability to work with the absorbent behaviour of car
tyres than the ones originally used for docking experiments at the project midpoint. This meant that,
25
as well as being used in the FUP to docking segment as described above, the forward sensor could
also be properly tasked with obstacle detection and reporting during the two transit segments.
In later trials it became clear that the rearward-facing laser, with the 16m range of the UST model,
would have the capability to scan for and recognise the DU itself, meaning that the deployment
could be envisaged operating with a simple reflective identifying panel at the DU in place of the SICK
laser. Whilst it was too late in the programme to investigate this in detail, it is considered to be one
of the initial areas to be looked at when preparing for exploitation, due to the lower cost and higher
resilience of such a solution.
BSS Software Integration
The final stage for WP8 was to integrate the demonstrator hardware with the AVERT control
software, ready for commissioning. This encompassed all software changes made, including
algorithm improvements, in preparation for the successful Final Demonstration.
The bogie-tracking algorithm was modified to detect only the bogie edges (as opposed to the full
bogie shape), making it more robust against external movement. The Swarm-tracking algorithm was
also integrated to increase the position accuracy of the swarm. These final improvements were
worked on by the ZHAW team prior to system integration activities in Stuttgart
Integration of Systems Components into Demonstrator WP9 - Lead IDUS
Overview
In WP9 all the technologies developed in the previous work packages were brought together to
integrate and test the AVERT system as a whole. IDUS was responsible for coordinating this task
which was initially rig-based but in its final stages was undertaken with the actual demonstrator sub-
system components as they became available. The primary integration site was at the FW facilities at
Eningen near Stuttgart.
Integration
The demonstration system was assembled and integrated in accordance with the Demonstrator
Design Definition Document developed in WP3. Technical specialists from each of the beneficiaries
prepared each sub-system and then worked together in an integrated team to ensure functional
operation of mechanical, electrical and communications and data interfacing across the system.
Once the functionality was established, including the interfacing with a typical EOD robot kindly lent
by the German user organisation, the team moved to the designated test site provided by FW for
operational integration and performance testing. The site was in a disused workshop with
characteristics similar to those found in a typical parking garage in terms of floor and supporting
pillars. In this area it was possible to set up the vehicle dispositions representative of the two use
cases and to commence operational integration in conjunction with members of the user panel who
provided input and also commanded the legacy EOD robot during the work-up.
26
Verification and Validation
Verification is the (internal) process of evaluating system requirements against the design
specification. Validation is the (external) process of evaluating User Requirements against the
customer or identified stakeholder operational criteria.
The use of the use of the “V diagram” systems engineering process for these tasks was identified by
IDUS at the start of the project and adopted at the first PMB meeting in July 2012.
Figure 12 - V Diagram
(Systems Engineering, coping with complexity”, Stevens, Brook, Jackson and Arnold, ISBN 0-13-095085-8)
The left side of the V defines what must be built in descending order of detail from the User
Requirements to the system requirements to the sub-system requirements. The right hand side of
the V represents the corresponding sub-system, system and user test levels
The electrical, mechanical and software components (built by each consortium partner) were
verified as they were integrated and tested at the component level, sub-systems level and systems
level, and then validated against the Use Cases as described by the Users.
Verification of each sub-system was undertaken within its related WP and only the demonstrator
systems testing and integration work was undertaken in WP9 prior to User validation. A specific
integrated test plan was prepared to cover these WP9 activities.
The level of autonomy was derived from the AVERT URD, to which validation information was
supplemented through feedback from the EOD Users. As part of the test plan document, two Use
Cases were validated by the Users supporting the AVERT Project. The Use Cases were initially
simulated within Webots to present to the Users. This enabled informed discussion of the key issues
regarding how AVERT could be used, with specific reference to the level of autonomy and the needs
and expectations of the Users.
To prepare the test framework the AVERT CONUSE sub-tasks, detailing how AVERT conducts EOD
missions were developed with the users and captured as UML Sequence Diagrams. This formed the
basis, against which a detailed analysis was undertaken for each AVERT sub-system, to map, using a
spreadsheet model, all (unique and individual) User and System Requirements needed to achieve
each and every action, executed through all the sub-tasks. The requirements themselves being
graded into KURs, Priority 1 and Priority 2. This was ratified at the Third User Workshop against the
KUR areas of Planning, SA, Monitoring, Interoperation, CIS Management, Security, Usability,
27
Deployability, and Sustainability. It was also confirmed that only KUR and Priority 1 requirements
would be featured in the demonstration itself.
A demonstration assessment capture document was populated during trials and testing prior to the
Final Demonstration, based on each of the User Requirements grouped against the sequence
diagrams. The outcome of each requirement test was assessed against specific numerical criteria
but, due to the security classification of certain performance data, the published results are rated
only in categories. the four assessment categories used were:
a)significant achievement,
b) passed,
c) failed
d) mission critical failure
Scores and comments were then recorded based on the assessment category and summarised in a
presentation table.
Table 1 - AVERT Verification and Validation Summary
The outcome of the assessment is that 98.8% of the KURs were achieved and a total of 80.2%
(11.1% significant achievement and 69.1% pass) of all requirements were reached at the
demonstration stage.
Additionally of the remaining 19.8% which failed only 1.2 % were KUR , 2.5% were Priority 1 and 16%
were Priority 2 ( desirable).
The 1.2% User Requirements that failed as mission success critical related to Bogie tracking
(demonstrated from the rear of the bogies only) and wireless messaging problems that meant that
the bogie Hokuyo local sensors could not be integrated in time for Final Demonstration. However,
the work to correct these had been delivered by ZHAW under deliverable D8.5 “Bogie Local
Navigation Concept” and can be incorporated during development for exploitation.
Demonstration Assessment of 81 Requirements
Failed and
mission critical
(-5 points)
Failed and
compensated
for or
overcome (0
points)
Pass (1 point) Significant
Achievement (2
points)
Total
Significant
Achievements
that could be
achieved
Key User Requirements (KURs) AVERT Total 1.2% 18.5% 69.1% 11.1% 40.7%
Planning 0.0% 3.7% 8.6% 0.0% 4.9%
Situational Awareness (SA) 1.2% 1.2% 17.3% 0.0% 9.9%
Monitoring 0.0% 4.9% 1.2% 3.7% 4.9%
Interoperation 0.0% 0.0% 3.7% 0.0% 0.0%
Communication and Information System (CIS) Management 0.0% 1.2% 3.7% 0.0% 2.5%
Security 0.0% 1.2% 0.0% 0.0% 1.2%
Usability 0.0% 3.7% 16.0% 6.2% 14.8%
Deployability 0.0% 1.2% 12.3% 1.2% 2.5%
Sustainability 0.0% 1.2% 6.2% 0.0% 0.0%
Demo Priority 1 and KUR (48 requirements) 1.2% 2.5% 45.7% 9.9% 34.6%
Demo Priority 2 (33 requirements) 0.0% 16.0% 23.5% 1.2% 6.2%
28
Key assessment findings
a) AVERT has exceeded the expectations in the KUR areas of Monitoring and Usability:
Embedded watchdog monitoring automatically placed AVERT in a known safe state when normal
communications are interrupted and was backed up by a separate dedicated emergency stop and
reversionary command communications channel, allowing direct user override of the autonomous
system.
The AVERT GUI and User Controls provided mapping, bogie trace and systems monitoring. The
system autonomy allowed the entire survey and deployment sequences to be commanded without
any user intervention, except for the designation of the vehicle to be moved and the place where it
was to be deposited.
b) The goal of fully autonomous operation has been reached:
The AVERT system has demonstrated proven swarm technology to rotate and translate bogies in any
direction, meaning that a vehicle may be extracted from complex and constrained environments.
This has been achieved through a fully-autonomous system, from the point of deployment by the
User to the FUP.
The link between the CSS ( extraction path-planning/ obstacle avoidance re-planning functions) and
the BSS (autonomous bogie docking and bogie swarm control functions) has been established via the
RnR application to achieve integration as an autonomous system.
c) The system is potentially practical in terms of safety, speed, capability and endurance:
AVERT has successfully demonstrated that it can take EOD Users out of harm’s way, whether bogies
are deployed on the DU or deployed by an EOD Operator within near range of the blocking vehicle.
The AVERT capability has been successfully validated against a series of Use Cases and has shown
that they can be achieved faster, with extraction times between 30% to 50% of the time taken by
manual methods.
Further improvements on the demonstrated timeline have been shown to be possible through
further optimisation of the survey and deployment sequence and concurrent rather than
consecutive operation of the bogies during deployment. These can be addressed in the development
for exploitation.
AVERT has exceeded User expectations on power management based on total number of vehicle
extractions per battery charge over short distances. In addition the use of quick change batteries
with spares carried on the DU means that extended operation with multiple extractions is possible
without having to withdraw and recharge the bogie units themselves.
d) The mechanical and survey elements of the system are close to exploitation whilst the electronics
and control software will require further development.
This is illustrated in the TRL assessment table.
29
Table 2 - AVERT Demonstrator TRL Assessment
Dissemination and Exploitation - WP10 - Lead IDUS
Dissemination & Exploitation was undertaken through preparation of academic papers and articles
for trade press (Intersec Magazine) attending international conferences and also a number of
exhibitions, at which printed leaflets were distributed, banners were displayed and short videos
were presented. Towards the end of the project the AVERT website was refurbished and a
professional video made of the final demonstration system. The publicity from the conferences and
the video (which was posted on YouTube in March with 340,850 hits by 22nd July 2015) has shown
considerable interest from the general public. In addition discussions have already been held with a
number of companies interested in exploiting AVERT as a product; however none of these has yet
agreed to go to the next stage.
Figure 13 - AVERT Stand at CPExpo Genoa December 2014
AVERT Element TRL
Bogie Mechanical Hardware 9
Electronics & Sensors 6-7
Control & Swarming Software 6-7
Survey & Mapping Software 7-8
30
Demonstration WP11 – Lead IDUS
Following the integration and validation activities a dedicated international workshop was held to
demonstrate the achievements of the AVERT project to a range of users, potential exploiters and the
EU programme manager and EU technical assessor.
At the workshop a presentation of the programme goals and achievements, including a summary of
the validation assessment was given to provide a context to the demonstration itself. The
demonstration encompassed the whole system but some elements ( in particular the DU
deployment mechanism and the reversionary video mechanism had had integration problems and
were not fully functional. These aspects were accommodated in the demonstration planning and did
not impact on the fundamental demonstration of a fully autonomous system. An additional
presentation summarised future and primary markets for exploitation, AVERT team partnering,
exploitation challenges and presented the assessed Technology Readiness Levels (TRL) of each of the
technologies used within AVERT.
The demonstration that was then held showed, with accompanying commentary from the relevant
technical leads, the full autonomous sequence from the towing in of the loaded DU to the docking,
lifting and extraction of a designated blocking car.
The Final Demonstration of AVERT followed Use Case “Timed” (UC3) and was in a representative
environment. The entrance came through a narrow doorway and then the DU was towed by the
EOD Robot about a concrete column in a turning circle much less than the constraints of a standard
car park of 5m to 6m ‘road’ width.
Figure 14 - Robot Towing Loaded DU
The AVERT CONUSE sub-tasks were demonstrated including (i) Dispatch, initialise and deploy AVERT
(EOD Operator); (ii) Bogie deployment and docking (fully autonomous); (iii) Vehicle lift and
extraction (fully autonomous); and (iv) Lower vehicle; recall bogies and DU (the re-docking of the
split-bogie pair and extraction of the joined bogies was demonstrated).
31
The bogies located the blocking vehicle axles, the bogies split and docked with the tyres in
autonomous mode. The vehicle was lifted and immediately extracted.
Figure 15 - Vehicle Extracted During Demonstration
Following the demonstration an extended question and answer session was held with the users and
exploiters around the equipment to discuss issues to be addressed during development for
exploitation and the exploitation prospects themselves.
The key message that came from the user community in this session was that the autonomy of the
overall system is well in advance of the capability that they would need immediately. For
exploitation most of the users identified that they would be keen to see a lower cost system which
would allow them to remote control the bogies to the FUP and the autonomy would then be
concentrated in the docking and lifting and the station keeping of the swarm as the assembly is
extracted under remote control.
In response to this, at the request of the Exploiters attending, a "bogie only" demonstration was
quickly arranged to prove the principle of the reduced scope approach.
The second demonstration showed that the joined bogie pair could be deployed as an independent
tool which could be either carried by the EOD Robot or deployed on a trolley by an EOD Operator.
The bogies were manually located in front of the vehicle and then autonomously docked. The
extraction and rotation was then conducted by the EOD Operator using a joystick, with ‘semi-
autonomous’ control of the bogie swarm, i.e. the User was controlling the ‘vehicle’ whereas the BSS
controller was processing the direction and speed of the mecanum wheels to resolve the User-
controlled vector direction. This was approved by the User community present as a tactically
acceptable option.
32
Specialist Support to Security and Safety Assessment (WP12) and Consortium Advisory Board
(CAB) & Workshops (WP13) – Lead IDUS
Overview
These two workpackages covered the engagement of specialists and arrangement of user workshops
throughout the programme to ensure that AVERT was being developed as a safe system capable of
meeting user needs and that, during development and especially public dissemination of the results,
security guidelines regarding IEDD operations were adhered to
The CAB was established at the start of the programme with appropriate specialist expertise from a
range of European countries. The initial challenge was to agree the key user requirements and
synthesise the operational use cases in such a way that they were representative of a composite
approach compatible with the very different approaches used across the different countries.
Managing the security level of AVERT has been challenging but has been helped by the active
involvement of the CAB members in reviewing internal requirements documents and external
dissemination documents before publication.
Great care was taken in the compilation of the user requirements from user interviews and
workshops to remove sensitive operational information from the internal documents with redacted
versions being specially prepared before posting on the EU portal. This is also true of the Verification
and Validation results where the actual performance figures have been translated into categories in
the posted documentation.
With the changes in the programme demonstration goals following the exit of MSDG the revised trial
scope did not necessitate the involvement of a separate safety specialist prior to development for
exploitation, sufficient expertise for trials level was available within the development team and the
existing CAB members.
There were four user workshops which were held regularly through the programme with a core
team of user representatives from UK, Germany and Switzerland reviewing the progress made and
giving advice on the interpretation of the requirements and the development of the use cases. The
same personnel sat on the separate security committee with suitable personnel from within the
consortium. The security committee sat formally when each workshop was held and ratified ex-
committee approvals of documents for publication. At the fourth workshop the UK representative
had retired and additional user representatives from Germany were included.
At the final demonstration workshop 23 users from ten separate countries were invited and of
these 12 , representing four countries (Germany, Greece, Austria and Switzerland) were able to
attend. In addition five potential exploiters were invited of which three were able to attend. The EU
Project Officer and Technical Assessor for AVERT were also invited and attended the Final
Demonstration.
33
Figure 16 - Team, Users & Assessors at Final Demonstration
Conclusions of S&T achievements across all Work Packages (WPs)
The AVERT programme has taken a concept and designed and demonstrated a practical autonomous
system capable of being exploited as an IEDD tool:
Within this the key overall conclusions are:
The objective of taking the operator out of harm's way has been achieved;
The goal of fully autonomous operation has been reached;
AVERT has exceeded the expectations in the KUR areas of Monitoring and Usability;
The system has been assessed to be potentially practical in terms of safety, speed, capability
and endurance;
The system is capable of exploitation either as a fully autonomous survey and extraction tool
or as a remotely controlled tool with autonomous docking and swarming to aid vehicle
extraction.
The specific conclusions which have been made across the workpackages are:
WP1 (Project Management) The effort required to monitor, report on and co-ordinate the project,
whilst kept to a minimum, was significantly greater than assumed by EU guidelines. Much of this
stemmed from the large volume of deliverables requiring managing and processing, significantly
greater than comparable projects of this size in other spheres.
WP2 (Overall System Definition) The early use of the CAB panel to develop the system requirements
and their continued involvement during the development and interpretation of the system
requirements was essential to the achievement of a practical system design definition.
The system design encompassed the principle of autonomous operation, providing the operator with
a tool which does not require detailed control but which can be paused or overridden by the user at
34
any point . This approach was designed to give the user confidence in using AVERT in IEDD
operations by reducing, not increasing operator workload.
WP3 (Command and Control System) This workpackage showed that sophisticated and effective
command and control systems can be built using modern off-the-shelf networking and processing
with very little adaptation. In particular the adoption of commercially sourced MIMO wireless
networking provided very robust communications in the less than ideal operating conditions
amongst and under vehicles and in steel framed buildings. Many of the techniques used are similar
to those in the professional COFDM systems used in modern EOD robot control so it is expected that
the transition to exploitation will be straightforward.
The GUI (developed in ROS) played a critical role and was scrutinised by the Users during the
Workshops, providing valuable feedback to meet the Users’ needs. The GUI and User Controls have
exceeded User expectations and have been a significant achievement under this project.
WP4 (AVERT Bogie and Deployment Platform Design) The overall design effort to produce a TRL8
level system with only one iteration was phenomenal in the timeframe, especially the reaction to the
results of the A-model trials and the restructuring of the profile and geometry in under 3 months
extension.
One major design success was the totally new mecanum wheel design and construction capable of
unprecedented load bearing within such a low profile. This is a highly significant achievement on this
project and could (in its own right) be exploited.
Another key design success was the imaginative use of 3D printing to provide rapid solutions to
complex housings and components necessitated by the design constraints, especially the low profile.
The rapid turnround, ease of modification and relatively low batch costs of 3D printing make it an
essential tool in research, development and rapid prototyping.
WP5 (AVERT Deployment Platform Development and Fabrication) The main achievement for WP5
was the ability of the team to rally round and produce a relatively complex platform rapidly, even
with the loss of the experienced platform integration partner. The fabrication task was necessarily
disjointed as the development of the automation sequence required the unit to be temporarily
shipped to the UK whereas ideally all tasks would have been completed in one location.
Nevertheless the prototype platform shown at the demonstration, albeit under-powered
pneumatically, provided all the key functions and showed a potential route to exploitation which
could be refined considerably in the future development phases.
WP6 (AVERT Bogie Module Deployment and Fabrication) The choice to fabricate the A-model to a
high standard was instrumental in understanding and refining the processes for procurement and
assembly ready for application to the Demonstrator bogies in Period 2.
This meant that during the development of the Demonstrator bogies a better understanding of the
critical parts, as well as access to specialist suppliers, helped prioritise purchasing, fabrication and
assembly of the demonstrator bogies and provided valuable information to support future
exploitation.
35
WP7 (Simulation & Driving Trajectory Calculation - Synchronised Platform) The application of
simulation early in the programme not only enabled rapid development of the path planner prior to
equipment availability but also enable the rest of the team to visualise and understand the approach
being taken. This became important during the discussions regarding local and global path planning
held between the leads of WP7 and WP8 which resulted in the local navigation path
planning/replanning and obstacle avoidance for the approaching bogies and the extracted vehicle
being subsumed into the global path planning through an additional task in WP7.
WP8 (AVERT Bogie Local Control & Obstacle Avoidance Sensor) For software development the
operational states of the bogies were determined for each stage in the operation in the sequence: (i)
locating the bogies from the DU to in front of the SICK laser scanners; (ii) controlling the joined bogie
pair along a trajectory to the FUP; (iii) wheel tracking to align the bogies to the axle of the blocking
vehicle; (iv) splitting the bogies and docking against the vehicle tyres; (v) synchronised lifting of the
vehicle; (vi) providing coordinated bogie swarm motion to follow a vehicle trajectory (autonomous
or manual joystick vector); (vii) lowering of the vehicle; and (viii) rejoining the bogie pair.
The development of these states could not be fully simulated in the timescale and hence the actual
bogie hardware was needed to develop and tune the software. This was difficult within the overall
programme due to design and fabrication delays, especially for the full set of demonstration bogies.
The preparation of the skeleton bogie set for swarm software development, and the extended use of
the A-model for docking sensor positioning, were essential to the completion of WP8 in time for the
demonstration. Even so several features which were allowed for in the software framework were
not able to be completed (e.g. side entry ). These are documented in the software code and can be
developed on the road to exploitation.
WP9 (Integration of the System Components into Demonstrator) focussed on the development of
AVERT Sub-Systems integration, at the Systems Requirement verification and User Requirements
acceptance validation levels, moving forward to systems integration as each sub-system became
available. The main conclusion from the verification was that the system is already well integrated at
the demonstration stage and scores highly against the requirements. The remaining capablities
needed for exploitation are readily identifiable.
WP10 (Dissemination & Exploitation) The workpackage has seen significant achievement in
dissemination, with several papers accepted into on-line journals and presented to conferences
during the programme itself, together with the featuring of AVERT at international exhibitions where
printed leaflets were distributed, banners were displayed and short videos were presented. Towards
the end of the project the AVERT website was refurbished and a professional video made of the final
demonstration system. The publicity from the conferences and the video (which was posted on
YouTube in March 340,850 hits by 22nd July 2015) has shown considerable interest from the general
public. In addition discussions have already been held with a number of companies interested in
exploiting AVERT as a product; however none of these has yet agreed to go to the next stage.
WP11 (Demonstration) The high attendance and enthusiasm of the users at the demonstration
reinforced the belief that an AVERT system would make a positive contribution as an IEDD tool. The
option to exploit using partial autonomy was specifically requested and demonstrated as a practical
stepping stone to exploitation of a fully autonomous system.
36
WP12 (Specialist Support to Security & Safety Assessment), WP13 (Specialist Support to CAB &
Workshops) The recruitment of specialist, subject matter expert advice from users has been vital to
the understanding of the AVERT role within the IEDD toolset and the use of the CAB committee
structure has been essential to the management of sensitive issues and the internal and external
documentation release.
The consortium is exceptionally grateful for the enthusiasm and input from the User community and
the high praise that AVERT has received at the workshops and demonstrations.
37
Potential Impact
AVERT provides a tool for Police and Emergency Services; with a unique capability to rapidly deploy,
extract and remove blocking vehicles (without disturbance) from vulnerable positions such as
confined spaces, tunnels, low bridges, as well as under-building and underground car parks. AVERT
operates autonomously alongside existing technologies, enhancing bomb disposal response speed
and safety.
There is a general consensus, amongst the Users approached during the development of AVERT, that
current methods of extracting blocking vehicles from restricted access height areas are inadequate.
This is currently a slow and cumbersome operation that exposes IEDD personnel to unnecessary
hazards. AVERT will improve IEDD operators ability to work out of harm’s way, and has shown the
potential to significantly improve IEDD response and capability to disrupt devices. AVERT is a tool
which extends the accessibility of legacy IEDD robotics to hitherto difficult-to-get-at vehicles and will
be welcomed by the IEDD community. AVERT will operate alongside legacy and future IEDD
equipment, utilising existing incident command vehicles and structures, to introduce the benefits of
autonomous vehicle extraction quickly, thereby providing additional capability to existing Police,
Security Forces and Fire and Rescue Services responsible for IEDD operations in Europe.
Competitive advantage is brought by direct application of AVERT to a known European infrastructure
security problem, within a framework that enables systems scalability to other autonomous control
and surveillance systems.
38
Table of Abbreviations
2D Two-dimensional
3D Three-dimensional
ACC AVERT Command Console
AVERT Autonomous Vehicle Emergency Recovery Tool
BBI Consortium member: BB-Ingenieure Ingenieurbüro, Germany
BSS Bogie Subsystems
BU Bogie Unit
CAB Consortium Advisory Board
CAD Computer Aided Design
CIS Communications and Information System
CONEMP Concept of Employment
CONOPS Concept of Operations
CONUSE Concept of Use
COTS Commercial Off The Shelf
CSS Command Subsystems
DDD Demonstrator Design Definition
DOW Description of Work
DSS Deployment Subsystems
DU Deployment Unit
DUTH Consortium member: Democritus University of Thrace, Greece
EOD Explosive Ordnance Disposal
EU European Union
FE Finite Element
FUP Form-Up Point
FW Consortium member: ForceWare
GUI Graphical User Interface
IED Improvised Explosive Device
IEDD IED Disposal
IEEE Institute of Electrical and Electronics Engineers
KUR Key User Requirements
LER Legacy EOD Robot
MIMO Multiple Input, Multiple Output
MS Microsoft
MSDG Consortium member: Marshall System Design Group Ltd, UK
PMB Project Management Board
RCV Remote Controlled Vehicle
ROS Robot Operating System
SA Situational Awareness
SDD System Design Document
SEIT Systems Engineering Integration Team
SMEs Subject Matter Experts
SMS Safety Management System
SOPs Standard Operational Procedures
SRD Systems Requirements Document
TTPs Tactics, Techniques and Procedures
TRL Technology Readiness Levels
UC Use Case
URD User Requirements Document
UML User Mark-up Language
VBIED Vehicle Borne IED
VOIED Victim Operated IED
Wi-Fi Wireless fidelity. Group of wireless data standards defined by the IEEE 802.11 standard series
WP Work Package
ZHAW Consortium member: Zurcher Hochschule Fur Angewandte Wissenschaften, Switzerland