Abdollah (Ebbie) Homaifar Duke Energy Eminent Professor and Director of Autonomous Control and...
37
Abdollah (Ebbie) Homaifar Duke Energy Eminent Professor and Director of Autonomous Control and Information Technology (ACIT) Institute Testing, Evaluation, and Control of Heterogeneous Large- Scale Systems of Autonomous Vehicles (TECHLAV) Department of Electrical and Computer Engineering North Carolina A&T State University 1601 E. Market Street/517 McNair Hall Greensboro, NC 27411 Email: [email protected]Office: 336-285-3709 Lab: 336-285-3271 Fax: 336-334-7716 http://acitcenter.ncat.edu/index.html Winning a DoD Center of Excellence: Process and Protocol Summary of Funded Research in Progress DoD 2015 “Taking the Pentagon to the People” HBCU/MI Technical Assistance Training Greensboro, NC 8 June 2015
Abdollah (Ebbie) Homaifar Duke Energy Eminent Professor and Director of Autonomous Control and Information Technology (ACIT) Institute Testing, Evaluation,
Abdollah (Ebbie) Homaifar Duke Energy Eminent Professor and
Director of Autonomous Control and Information Technology (ACIT)
Institute Testing, Evaluation, and Control of Heterogeneous
Large-Scale Systems of Autonomous Vehicles (TECHLAV) Department of
Electrical and Computer Engineering North Carolina A&T State
University 1601 E. Market Street/ 517 McNair Hall Greensboro, NC
27411 Email: [email protected]@ncat.edu Office:
336-285-3709 Lab: 336-285-3271 Fax: 336-334-7716
http://acitcenter.ncat.edu/index.html Winning a DoD Center of
Excellence: Process and Protocol Summary of Funded Research in
Progress DoD 2015 Taking the Pentagon to the People HBCU/MI
Technical Assistance Training Greensboro, NC 8 June 2015
Slide 2
Outline Tips for successfully winning a proposal Summary of
Funded research in progress Testing, Evaluation and Control of
Heterogeneous Large-scale systems of Autonomous Vehicles (TECHLAV)
Mining Climate and Ecosystem Data The Center for the Study of
Evolution in Action (BEACON) The Crash Imminent Safety (Cris)
University Transportation Center (UTC) Perception Inference Engine
(PIE) for Test & Evaluation of autonomous systems
Acknowledgment Questions 2
Slide 3
Tips for successfully winning a Proposal Start early and plan
ahead Build life-long relationships with stakeholders Understand
the requirements and needs of the stakeholders External Read and
execute the RFP Have a team of internal and external reviewers
Collaborating team Have diverse individual(s) in each subject
matter Understand the expectations, strengths, and needs of each
individual 3
Slide 4
Tips continued.. Be organized and delegate responsibilities
Expect the unexpected (have a backup plan) Dont take things
personal and do not internalize issues Be a good listener and treat
everyone with respect Good ethics History Track record Be
passionate about what you do and have fun Remember: Winning and
sustaining is about team effort (Credit goes to all) 4
Slide 5
Testing, Evaluation and Control of Heterogeneous Large-scale
systems of Autonomous Vehicles (TECHLAV) 5
Slide 6
6 DoD Roadmap
Slide 7
About TECHLAV 7 A. Homaifar Testing, Evaluation and Control of
Heterogeneous Large-scale Systems of Autonomous Vehicles (TECHLAV)
Testing, Evaluation, and Control of Heterogeneous Large Scale
Systems of Autonomous Vehicles (TECHLAV) is a multidisciplinary
research Center on the cutting edge of Control, Communication,
Computation and Human Cognition to address two main problems:
1.Teaming and Cooperative Control of Large Scale Autonomous Systems
of Vehicles (LSASVs) integrated with human operators. 2.Testing,
Evaluation, Validation, and Verification of LSASVs.
Slide 8
Systems of Vehicles Systems of Vehicles: a (large) team of
heterogeneous (aerial, ground, sea surface, and underwater)
vehicles that together create a new, more complex networked system
that is more capable and robust. 8 A. Homaifar Testing, Evaluation
and Control of Heterogeneous Large-scale systems of Autonomous
Vehicles (TECHLAV)
Slide 9
Challenges A. Homaifar Testing, Evaluation and Control of
Heterogeneous Large-scale systems of Autonomous Vehicles (TECHLAV)
9 TEACHLAV How to mathematically describe the collective behaviors
of systems of vehicles? How to reliably design decentralized
control and communication for systems of vehicles to achieve
desired collective performance? How to test and evaluate the
control and communication between systems of vehicles? TEACHLAV
creates a comprehensive umbrella covering different aspects of
systems of vehicles ranging from Modeling and Analysis to Control
and Communication and Testing and Evaluation. Modeling and Analysis
Reliable Control and Communication Test and Evaluation
Slide 10
About TECHLAV 10 A. Homaifar Testing, Evaluation and Control of
Heterogeneous Large-scale Systems of Autonomous Vehicles (TECHLAV)
Department of Defense (DoD) Office of the Secretary of Defense
(OSD) Air Force Research Laboratory (AFRL) Remark: The views and
conclusions being discussed here are those of the presenters and
should not be interpreted as necessarily representing the official
policies or endorsements, either expressed or implied, of Air Force
Research Laboratory, OSD, or the U.S. Government. TECHLAV Center
has been funded at $5 M by the Department of Defense (DoD) as a
Center of Excellence in Autonomy in 2015. The grant is operated by
the Air Force Research Laboratory (AFRL) and the Office of the
Secretary of Defense (OSD).
Slide 11
TECHLAV Mission and Vision 11 A. Homaifar Testing, Evaluation
and Control of Heterogeneous Large-scale systems of Autonomous
Vehicles (TECHLAV) 1.Address fundamental problems in modeling,
analysis, control, coordination, test and evaluation of autonomous
systems 2.Serve as a national resource in education and research in
Autonomy 3.Provide outreach services in autonomy related area
4.Foster linkages among national institutions of higher education,
government agencies and industries 5. Commercialize TECHLAV
technologies for the benefit of the national economy. The TECHLAV
as a Center of Excellence in Autonomy will
Slide 12
Collaborators 12 A. Homaifar Testing, Evaluation and Control of
Heterogeneous Large-scale systems of Autonomous Vehicles (TECHLAV)
Southwestern Indian Polytechnic InstituteUniversity of Texas at San
Antonio North Carolina A&T State University TECHLAV is led by
North Carolina Agricultural and Technical State University (N.C.
A&T) in collaborate with the University of Texas at San Antonio
to conduct strong integrated multi-disciplinary research and
education activities on Large Scale Autonomous Systems of Vehicles
(LSASV). The Center also partners with Southwestern Indian
Polytechnic Institute (SIPI) to provide and promote education and
outreach activities and curriculum development to the larger Native
American community.
Slide 13
TECHLAV Core Research 13 A. Homaifar Testing, Evaluation and
Control of Heterogeneous Large-scale systems of Autonomous Vehicles
(TECHLAV)
Slide 14
Research Thrust 1 14 A. Homaifar Testing, Evaluation and
Control of Heterogeneous Large-scale systems of Autonomous Vehicles
(TECHLAV) Thrust 1 (Modeling, Analysis and Control of Large-scale
Autonomous Vehicles (MACLAV)) will develop scalable methodologies
to improve modeling, analysis, localization, navigation, and
control of LSASV. Managing modern large scale systems requires new
scalable techniques to integrate communication, control and
computation that can interact with humans through many new and
emerging modalities. Lead: Dr. M. Jamshidi, UTSA. Members: Dr. J.J.
Prevost and Mr. P. Benavidez, UTSA; Dr. A. Karimoddini, N.C.
A&T
Slide 15
Research Thrust 2 15 A. Homaifar Testing, Evaluation and
Control of Heterogeneous Large-scale systems of Autonomous Vehicles
(TECHLAV) Thrust 2 (Resilient Control and Communication of
Large-scale Autonomous Vehicles (RC2LAV)) will develop systematic
methods to enhance the reliability and efficacy of the control
structure and the communication backbone for LSASV integrated with
human operators in dynamic and uncertain environments such as a
battlefield. Lead: Dr. A. Karimoddini, N.C. A&T Members: Drs.
F. Afghah and S. Yi, N.C. A&T; Drs. M. Jamshidi and B. Kelley,
UTSA
Slide 16
Research Thrust 3 16 A. Homaifar Testing, Evaluation and
Control of Heterogeneous Large-scale systems of Autonomous Vehicles
(TECHLAV) Thrust 3 (Testing, Evaluation and Verification of
Large-scale Autonomous Vehicles (TEVLAV)) will develop and provide
technologies and tools for testing, evaluation, validation, and
verification of heterogeneous LSASV. This will be done through a
runtime formal verification approach which uses a
divide-and-conquer technique to modularly check whether LSASV can
satisfy the desired (complex) high level specifications. Lead: Dr.
Y. Seong, N.C. A&T Members: Drs. A. Homaifar, A. Karimoddini
and J. Stephens, N.C. A&T
Slide 17
TECHLAV Education and Outreach 17 A. Homaifar Testing,
Evaluation and Control of Heterogeneous Large-scale systems of
Autonomous Vehicles (TECHLAV) Collaborative Curriculum Development:
Developing or reorganizing autonomy related courses at NC A&T,
UTSA, SIPI which will be shared with the collaborating institutions
and 11 other Tribal Colleges and Universities Training Pathway to
Next Generation of STEM Leaders TECHLAV effectively trains and
educates students with backgrounds in robotics, and control systems
Student mobility-doctoral students from N.C. A&T and UTSA will
visit other institutions to share research results and exchange
research ideas The graduate students from N.C. A&T and UTSA
will visit SIPI to help students with their engineering education
and to assist in the preparation of educational laboratories
Outcome Graduating approximately 45 PhD and MS from N.C. A&T
and UTSA Supporting 105 Undergraduate students from all three
institutions Influence more than 300 Native American students in
Albuquerque, NM, and 150 high school teachers and students in
Greensboro, NC, via year-round courses offered through outreach
activities
Slide 18
TECHLAV People 18 A. Homaifar Testing, Evaluation and Control
of Heterogeneous Large-scale systems of Autonomous Vehicles
(TECHLAV) Dr. A. Homaifar, Director NC A&T Dr. M. Jamshidi Lead
for Thrust 1 UTSA Dr. A. Karimoddini Deputy Director and Lead for
Thrust 2 NC A&T Dr. Y. Seong, Lead for Thrust 3 NC A&T Dr.
F. Afghah Lead for DII NC A&T Dr. J. Kelly Lead for Education
NC A&T Dr. N. Vadiee Lead for Outreach SIPI Dr. B. Kelley
Member, UTSA Dr. S. Yi Member, NC A&T Dr. J. Stephens Member,
NC A&T Mrs. Shar Seyedin Program manager NC A&T
Slide 19
TECHLAV Advisory Board 19 A. Homaifar Testing, Evaluation and
Control of Heterogeneous Large-scale Systems of Autonomous Vehicles
(TECHLAV) Dr. Barry L. Burks, chair, Strategic Planning and Policy
Board (SPPB) Edward Tunstel, Jr., Senior Robotics at John Hopkins
University Applied Physics Laboratory, Chair of Scientific and
Industrial Advisory Board (SIAB) Garry Roedler, Engineering
Outreach Program Manager at Lockheed Martin Corporate Engineering
Richard A. Carreras, Senior Researcher at Laser Division at Air
Force Research Lab/RDLE Paul C. Hershey, Senior Engineering Fellow
at Raytheon Intelligence Information and Services Daniel
DeLarentis, Professor, Department of Aerospace Engineering, Purdue
Russell L. (Rusty) Roberts, Director of Aerospace Transportation
& Advanced Systems Lab, Georgia Tech Research Institute Roberto
Horowitz, Professor, James Fife Endowed Chair, Department of
Mechanical Engineering at UC, Berkeley and Director of Partners for
Advanced Transportation Technology (PATH) Dr. Julius Yellowhair,
(Navajo), Senior Member of the Technical Staff at Sandia National
Laboratories Ms. Sandra Begay-Campbell (Navajo)-[manages Sandia
National Laboratories tribal energy technical assistance program]
Wallace Alton S., Senior Research Consultant Dr. Mark Noakes,
Senior Engineer, Robotics and Remote Systems, at ORNL
Slide 20
Mining Climate and Ecosystem Data 20
Slide 21
Analysis of time series with time-varying parameters Many real
systems are time-varying, i.e., the parameters of mathematical
models of the system are changing through time. The time series
generated by a time-varying system has some regimes, while the
parameters of the model remains constant in each regime. Given the
time series, the aim is to find the number and time of changes
between the regimes and to detect the parameters in each regime.
Solving this problem is generally difficult, since there are more
unknown parameters than known parameters. It has many applications
in climatology, economy, engineering, and etc. Dr. Abdollah
HomaifarMohammad GorjiDr. Stefan LiessJaquana McNeill NSF-funded
project on understanding climate changes from data, in
collaboration with North Carolina State Univ., Univ. of Minnesota,
and Northeastern Univ. Michael Lowe
Slide 22
Example: How US temperature trends have changed We found that
US temperature time series can be modeled by a piecewise linear
trend, with a change point in 1958. There is a warming trend in
many areas of the US after 1958. Also, there is a change point in
the climate forcings time series (including greenhouse gases, solar
activity, etc) in 1960. Other studies suggest that the post- WWII
economic expansion is the reason for increases in anthropogenic
forcings. This fact led to the occurrence of the common change
point in forcings and temperature series, marking the beginning of
sustained global warming. Forcing (W/m2)
Slide 23
Example: Modeling of switched dynamical systems Switched
dynamical systems are mixtures of continuous (transfer function)
and discrete events (switching mechanism). Given the input/outputs
of the system, find the transfer functions and switching times.
Switched dynamical systems are mixtures of continuous (transfer
function) and discrete events (switching mechanism). Given the
input/outputs of the system, find the transfer functions and
switching times. The position and angular velocity of the car on
the road are controlled by the gear and throttle. The system has
four different dynamical systems with unknown switching times.
Given the position and velocity of the car, we can find when the
gears have been changed. The position and angular velocity of the
car on the road are controlled by the gear and throttle. The system
has four different dynamical systems with unknown switching times.
Given the position and velocity of the car, we can find when the
gears have been changed. Modeling switched systems has many
applications for understanding the behavior of complex and
nonlinear systems.
Slide 24
Chaunt Lacewell 1, Dr. Abdollah Homaifar 1, Dr. Yuh-Lang Lin 2,
and Dr. Kenneth Knapp 3 Autonomous Control and Information
Technology (ACIT) Center Department of Electrical and Computer
Engineering 1 Department of Energy and Environmental Systems 2
North Carolina A&T State University NOAA - National Climatic
Data Center 3 Title III PhD. Fellow Graduate Research Assistant
Duke Energy Eminent Professor This work is partially supported by
the Expeditions in Computing by the National Science Foundation
under Award Number: CCF-1029731 Professor Senior Scientist Products
Branch Chief
Slide 25
Problem Definition Investigates predictive features, such as
geometric center and eccentricity, that contribute to the
development of cloud clusters (CC) into a tropical cyclone (TC) in
the North Atlantic Ocean Forecasters need better data-mining and
data-driven techniques to provide advanced warning of rare events
such as TCs Only global gridded satellite data that are readily
available are used Without numerical weather prediction models
Challenges Imbalanced data # non-developing CCs >>> #
developing CCs Our proposed Selective Clustering-based Oversampling
Technique (SCOT) Outperformed state-of-the-art methods
Classification of multivariate time series with varying lengths
Evolution of a CC is complex CCs can change shape and size rapidly
No ground truth data of identified and tracked CCs 25
Slide 26
Neural Network Results Comparison of the geometric means and
the performance of developing (recall) and non-developing
(specificity) cloud clusters using (a) the imbalanced dataset and
(b) the balanced dataset for each forecast hour for the neural
network simulation. (a) (b)
Slide 27
The Center for the Study of Evolution in Action (BEACON)
27
Slide 28
Mohammad Gorji Dr. Abdollah Homaifar Real-life t ime series
prediction with soft computing approaches Mina Moradi Joy Edward
Larvie Jaquana McNeill This work is partially supported by the
BEACON by the National Science Foundation under Cooperative
Agreement No. DBI-0939454 : the Study of Evolution in Action BEACON
is an NSF Science and Technology Center, headquartered at Michigan
State University, with partners at North Carolina A&T State
University, University of Idaho, University of Texas at Austin, and
University of Washington. Dr. Joseph L. Graves, Jr. Dr. Scott
Harrison
Slide 29
29 Sunspot time series is a chaotic and nonlinear system in the
fields of geophysics and climatology, that has attracted the
interest of many researchers to model the behaviors of the system
by a variety of methods. It contains the yearly number of dark
spots on the sun from 1700 to 1987, consisting of 288 observations
Real-life t ime series prediction with soft computing approaches A
time series is a sequence of data points that are measured over a
time interval. Time series forecasting is the use of a model to
find the internal dynamics of a system and predict future values
based on previously observed values. Artificial neural network
(ANN) is a data-driven approach capable of transforming the inputs
into meaningful outputs when dynamics of the system is unknown The
performance of ANN improves if an efficient set of inputs and lag
observations are chosen by the designer We proposed partially
connected artificial neural network with evolvable topology
(PANNET) for predicting time series with complex dynamics PANNET is
able to predict observations that have never been observed in a
training dataset
Slide 30
30 Real-Life Time Series Prediction: Reconstruction of Gene
Regulatory Network Reconstruction of Gene Regulatory Network (GRN)
Constructed network for target gene lexA which identifies the
previously reported regulations of gene lexA by uvrD, umuD, lexA
Temporal gene expression of the E. coli SOS DNA Repair System This
GRN is known for repairing the DNA after some damage GRN is an
abstract mapping of gene regulations in living cells. DNA
microarray technology simultaneously measures the expression levels
of thousands of genes inside cells in response to specific
environmental conditions. Identification of GRNs and prediction of
gene expression can open up a window on the disease progression and
lead to accurate diagnosis, drug development, and treatment. The
proposed PANNET is able to use temporal genes expression values and
model the GRN by considering different time delays in gene
interactions.
Slide 31
The Crash Imminent Safety (Cris) University Transportation
Center (UTC) 31
Slide 32
32 Dr. Abdollah Homaifar Duke Energy Eminent Professor Graduate
Research Assistants: Undergraduate Research Assistants: Saina
Ramyar PhD Student Allan Anzagira PhD Student Seifemicheal Amsalu
PhD Student Developing an Autonomous Driving Assistance System
Senior Design Team: Thy Le, Keyur Vaidya, Josh White, Clarence
Houser Xuyang Yan This work is partially supported by US Department
of Transportation under Award Number: 60040605 Collaborating
institutions:- Ohio State University (OSU), North Carolina A&T
State University (NCAT), University of Wisconsin (UW), University
of Massachusetts (Umass) Amherst Autonomous Control and Information
Technology (ACIT) Center Department of Electrical and Computer
Engineering 1 North Carolina A&T State University
Slide 33
Problem Definition Development of Autonomous Driver Assistance
System (ADAS) Modeling and predicting the behavior of the driver in
pre-crash scenarios Detection, modeling, and mitigation of driver
distraction Challenges The reaction time before a crash happens is
very small, so the ADAS must predict the crash possibility and
react quickly 33 Driver distraction detection There are no direct
indicators of driver cognitive distraction The distraction
detection equipment (e.g., eye sensor) may disturb and distract the
driver Our Work: The Real-time Technologies Inc. Driving Simulator
is used to mimic real life driving scenarios alongside the eye
tracking system to detect driver distraction behavior. There are no
direct indicators of driver cognitive distraction The distraction
detection equipment (e.g., eye sensor) may disturb and distract the
driver Our Work: The Real-time Technologies Inc. Driving Simulator
is used to mimic real life driving scenarios alongside the eye
tracking system to detect driver distraction behavior.
Slide 34
34 Estimation of state at each time step for left and right
maneuvers with SVM(red) and HMM (blue) Driver behavior modeling The
driver behavior model has temporal as well as spatial dynamics The
model takes dynamic inputs about the changing environment,
therefore finding the switching time is important The model must
simulate perception, attention, cognition, and control behavior of
the driver Our Work: Driver behavior models are being developed
using machine learning techniques: Hidden Markov Model, Support
Vector Machines and Fuzzy Logic The driver behavior model has
temporal as well as spatial dynamics The model takes dynamic inputs
about the changing environment, therefore finding the switching
time is important The model must simulate perception, attention,
cognition, and control behavior of the driver Our Work: Driver
behavior models are being developed using machine learning
techniques: Hidden Markov Model, Support Vector Machines and Fuzzy
Logic
Slide 35
My students at the ACIT center particularly Chaunt Lacewell,
Mina Moradi Kordmahalleh, Mohammad Gorji Sefidmazgi, Joy Edward
Larvie, Norbert Agana, Saina Ramyar, Seifemichael Amsalu, and Allan
Anzagira My colleagues, particularly Drs. Goodman and Adami
(Michigan State University), Vipin Kumar (UM), Umit Ozguner (Ohio
State), Guiseppi-Elie (Clemson University), Dozier, Karimoddini,
Graves and Lin (NC A&T University), and Dr. Kenneth Knapp at
NOAA - National Climatic Data Center Department of Defense (DoD) -
the Office of the Secretary of Defense (OSD) NSF Science and
Technology Center NSF Expedition in Computing U.S. DOT University
Transportation Center Test Resources Management Center (TRMC)
Scientific Research Corporation (SRC) Acknowledgment 28/29 35
Slide 36
Developing a Data-Driven Perception Inference Engine (PIE) for
Test & Evaluation of autonomous systems 36
Slide 37
Work is still in progress......feedback welcome! 37