Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
ARTIST2ARTIST2 Embedded Systems Design
Cooperating objects in Wireless Sensor networks
Dr. Paul J.M. HavingaUniversity of Twentethe Netherlands
ARTIST2ARTIST2 Embedded Systems Design
Agenda
Wireless sensor networksSystem Software architectureBeyond WSN
ARTIST2ARTIST2 Embedded Systems Design
Wireless sensor network
A network that is formed when a set of small sensor devices that are deployed in an ad hoc fashion cooperate for sensing a physical phenomenon
What’s the difference with conventional sensor?Not just sensors, but sensors with tiny brains!
+ + =
ARTIST2ARTIST2 Embedded Systems Design
Sensors with brains?
What’s the advantage of having brains??!!Improves the quality of readings obtained
E.g. false readings due to malfunctioning sensors are discarded
Sensors can calibrate themselvesShorter response time
Allows nodes to function autonomouslyE.g. how to route data when a certain node fails or
moves out of range? (Improves robustness!)Makes the network more efficient
Increased network lifetime – works longer without any need for user intervention
New functionalitiesE.g. localizationLocal execution of business rules
ARTIST2ARTIST2 Embedded Systems Design
Networked Embedded Systems
“Everything” is networkedEven very small things like sensors and actuatorsExplosion in the number of connected end devices
In-network processingProtocol stack plus some ability to process data in network end devicesDecentralized and localized controlServices executing inside the network
ARTIST2ARTIST2 Embedded Systems Design
Ideal Sensor Networks
“Smart Dust”Tens of thousands of sensor nodesNode lifetime longer than 3 yearsNode size smaller than 1 mm3
Node price less than 5 cents
ARTIST2ARTIST2 Embedded Systems Design
Real Sensor Nodes
Sensor nodesMatchbox size Limited resources
energy, memory, CPU powerLow data-rate radioExpensive ($20 – $500)Limited lifetime (1 day – 3 years)
ARTIST2ARTIST2 Embedded Systems Design
fine grained sensing systems
Nodes with very limited resources16 bit processor, 2KB RAM, 60KB FLASHLow data-rate radio (100 kbits/sec)Limited energy available (1-2 small batteries) for several years of operationSmall physical size
Networks characteristicsDistributed multi-hop mesh networkmobile unreliable nodesDeployed in a harsh environmentSelf-organizing and self-healing
SensorsSimple, low cost, low accuracyLight, temperature, pressure, movement, humidity,
ARTIST2ARTIST2 Embedded Systems Design
Application domains
Wireless Sensing and Acting Networks regarded as one key emerging technology platform for a wide range of applications
EnvironmentEnvironment
InfrastructureInfrastructure
AgricultureAgriculture
ARTIST2ARTIST2 Embedded Systems Design
Same concepts can be applied to Futuristic Applications!
ARTIST2ARTIST2 Embedded Systems Design
Beyond traditional sensor networks….
ARTIST2ARTIST2 Embedded Systems Design
Facing the real world…
Limited resourcesProcessor, memory, communication, and energy
The real-world is continually changingEnvironments, technologies, and resources are constantly shiftingNew technologies have to co-exist
The real-world is inherently unpredictable and uncertain, and incomplete
Inaccurate and incomplete informationImperfect communicationBatteries drain out unpredictableTechnologies move and change
ARTIST2ARTIST2 Embedded Systems Design
A dynamic world
The physical world is dynamic Dynamic operating conditionsDynamic availability of resources
Including energy!Dynamic tasksDynamic network conditions (mobility, errors)
Devices must adapt automatically to the environment
Too many devices for manual configurationEnvironmental conditions are unpredictable
Self-configuring, autonomous, dependable, robust, scalable
ARTIST2ARTIST2 Embedded Systems Design
Agenda
Wireless sensor networksSystem Software architectureBeyond WSN
ARTIST2ARTIST2 Embedded Systems Design
Abstract
Existing OS for WSN limitedEither merely runtime systems, no real-time, no dynamics, hard to develop withOr too big and complex
AmbientRT: a real-time data centric operating systems
for resource poor devicesefficient memory organizationdynamic (task level) reconfigurationevent (data) driven with real-time scheduling
Superior capabilities at comparable cost of alternative solutions
ARTIST2ARTIST2 Embedded Systems Design
Current WSN architectures
Rather staticfixed layered protocol stackfixed applicationsMerely static networks
Each protocol or algorithm covers its worst-case scenario Hard-coded interfaces between layersConfiguration only at compile timeNo central coordination of functionalitySoftware-homogenous nodes
ARTIST2ARTIST2 Embedded Systems Design
Example: TinyOS archictecture
RFM
Radio byte
i2c
Tempphoto
Messaging Layer
clocksbit
byte
packet Radio Packet
Routing Layer
sensing applicationapplication
HW
SW
ADC
messaging
routing
UART Packet
UART byte
ARTIST2ARTIST2 Embedded Systems Design
Dynamic WSN architectures
Wide variety of wireless sensor applicationsWith very different requirements and characteristicsNeeds adaptation geared towards needs and possibilities
DynamicsMobile nodesChanging services, protocols and resourcesAdaptable network protocols
Self configurationSelf adaptationSelf organizingSelf adaptation
ARTIST2ARTIST2 Embedded Systems Design
Example: EYES Architecture
Location Time Synchronisation Network Management
Security
Failure Detection
Service Discovery
User Application
Aggregation
Topology
Agreement
ID-centric routing
Link Layer Control
Geo-centric routing Data-centricrouting
Frame Tx/Rx
PHY
Mainly tasked with packet forwarding/exchange
Mainly tasked with inter-buildingblock information exchange
Message passingInformation exchange
Contention & Scheduling
ARTIST2ARTIST2 Embedded Systems Design
Typical Hardware Constraints
CPU (MSP430)Clock frequency: 4.6 MHz2048 bytes RAM60 Kilobyte flash (program, read-only data)
Radio115.2 Kbps
Power supply2 AA alkaline batteries
Major research questions:
Do we need, and can we afforta real time operating system on such a platform?
ARTIST2ARTIST2 Embedded Systems Design
Why do we need anoperating system?
For the same reason as we use them for bigger machines !
Ease of programmingHardware abstractionMulti tasking supportResource managementEfficient resource usage
Direct processing of data to prevent the need of buffers
Efficient use of memory
ARTIST2ARTIST2 Embedded Systems Design
Why should it be real time?
Timely behaviourComputation correctness based on both the logical result and the time of that result“Immediate” response to external events
Ease of programmingCode has to be written carefullyCorrect execution is not based on tediuos programming (cooperative scheduling), user needs to plan the scheduling Separation of concerns
RobustnessPredictable system Prevent time critical tasks (e.g. radio) to fail because of execution of other less important tasks
Saving energy by real time operationJust in time operation: real-time saves energyLess failure is less retrying: energy efficient
ARTIST2ARTIST2 Embedded Systems Design
Any more reasons?
Yes, to be able to cope with the dynamics!Self *
Self configurationSelf calibrationSelf organizingSelf adaptation
By reconfiguration and adaptation
ARTIST2ARTIST2 Embedded Systems Design
WSN operating systems
TinyOS“TinyOS is a component-based runtime environment designed to provide support for deeply embedded systems which require concurrency intensive operations while constrained by minimal hardware resources.”Developed mainly at University BerkeleyTargeted microcontrollers: Atmel, TI, etc.Freeware, sustained by a large active community
ARTIST2ARTIST2 Embedded Systems Design
TinyOS characteristics
Event based architecture conceptIncoming events are served by specific tasksNon-blocking execution is required (run to completion)Free CPU cycles are spent in the sleep stateFast context switches, small footprintNo real time scheduling capabilities
Each component designed to cover all the (worst) possible cases
Cooperative (manual) schedulingConfiguration and tuning available only at compile timeGood operating system for small tasks & task sets
Polling a sensor periodicallyBroadcasting data
ARTIST2ARTIST2 Embedded Systems Design
Our approach: Architectural Overview
Combine system architecture with underlying operating system
All protocols, algorithms, protocol stacks implemented as tasks or sets of tasks (also called modules)Operating system is a local centralized system:built around the task scheduler
Introduce a central componentEvery module communicates with other modules via a publish/subscribe mechanismCan select/enable/disable layersCan turn on/off certain functionalities of each layer
ARTIST2ARTIST2 Embedded Systems Design
EDFI Scheduling
EDFI (Earliest Deadline First with Inheritance)Dynamic priority preemptive scheduling methodOne stack used by all tasksAutomatic shared resource synchronization (mutual exclusion)Deadlock free
ARTIST2ARTIST2 Embedded Systems Design
AmbientRT implementation
We have implemented AmbientRT on an MSP430Very limited RAM usage
min. 32 bytesPer task 9 bytes extra
Small code (3800 bytes)Max. 12500 task switches per second
available on http://ambient-systems.net
ARTIST2ARTIST2 Embedded Systems Design
System software architecture - conclusions
Proposed architecture allowsDynamic configuration of the protocol stack usedEfficient cross-layer communicationSelf adaptation to external conditionCentralized control over the performance of the protocol stackDynamic reconfiguration of tasks
Even over the air!
ARTIST2ARTIST2 Embedded Systems Design
Agenda
Wireless sensor networksSystem Software architectureBeyond WSN
ARTIST2ARTIST2 Embedded Systems Design
Transport and Logistics
Currently RFID applications are emergingCheap tags are availableRequires infrastructure
ExpensivePlanningSoftware backendLimited reading distance
Novel services are possible using WSN technologiesLocalizationAutonomous controlReconfigurableLarge coverage areaScalable
ARTIST2ARTIST2 Embedded Systems Design
WSNs in Transport and Logistics
Identify opportunities for WSNs in Transport and Logistics Specific scenario: distribution of flowers from auction hall to shopUse embedded devices with wireless ad-hoc networking and sensingGoal: improve efficiency of process with active computing components
ARTIST2ARTIST2 Embedded Systems Design
Process diagram
Auction hall pipeline
Expedition floor Trailer
Loading Gate
ARTIST2ARTIST2 Embedded Systems Design
Major sources of errors
RTIs filled with wrong goodsRTIs placed in wrong cell on expedition floor, or lostthe expedition floor lacks space for static allocation of cells,resulting in ad-hoc, error-prone solutionsdelays gathering RTIs on expedition floorRTIs are placed in the wrong trailerRTIs are not returned from retail storeswrong environmental conditions during transport (temperature...)
Can these problems be solved using WSNs?
ARTIST2ARTIST2 Embedded Systems Design
Functional requirements
WSN node is active component: keep track of all activity in distribution processAlert when error detectedNodes keep track of own location on expedition floor alert if in wrong positionNodes know contents, destination and transporting trailer of RTIs alert when in wrong trailer, or delivered at wrong customerEnvironmental monitoring higher quality of transport process…
ARTIST2ARTIST2 Embedded Systems Design
Process diagram – revised
Expedition floor
Central System
Trailer
Loading Gate
Auction hall pipeline
ARTIST2ARTIST2 Embedded Systems Design
Requirements – Networking
Up to 5 nodes per RTITrailers contain <= 60 RTIsExpedition floor contains <= 10.000 RTIsRTIs are moved around
Very high density, high mobilityuse small communication distances
Ad-hoc multi-hop networkLow operational power, batteries not replaceable
Minimal 5 years on a single batteryEnergy efficient network protocols needed
ARTIST2ARTIST2 Embedded Systems Design
Requirements – Localization
Localization needed on expedition floorMaximal accuracy of ~2 mNodes must be robust and cheap
no special distance measurement sensors like ultrasound
Support of large number of wireless devicesSupport of multihop network possibleDistributed mechanism needed
ARTIST2ARTIST2 Embedded Systems Design
Requirements – Rule Engine
Specify rules of logistics processVerify correct execution of logistical process, alert when errors occurRules dynamically updated when neededHandle different data typesProvide communications abstractionEasy to program, used by non-experts
ARTIST2ARTIST2 Embedded Systems Design
FlowerPower
Our application scenario makes use of innovative solutions:
energy efficient networking (MAC, routing)Localization Sensing
New concepts are introduced: Distributed rule engine
Useful application scenario to evaluate additional value of WSN protocolsprovides an alternative, and solution beyond RFID
Beyond RFID and beyond sensor networks
ARTIST2ARTIST2 Embedded Systems Design
Conclusion
Future WSNs are:HeterogeneousDynamicLarge-scaleHave multiple different services and stakeholders
Collaboration is essentialCross-layer optimizationTight integration needed of networking protocols (MAC, transport, routing) and applicationsProper system software support
Face the real world, do experiments
Still numerous challenges, but…Even more opportunities!