dr.ir. tamás keviczky · 2018. 11. 23. · reconfigurable control design in aerospace applications...
TRANSCRIPT
Networked Cyber-Physical Systems
Dr.ir. Tamás Keviczky
Delft Center for Systems and ControlDelft University of TechnologyThe Netherlands
[email protected]://www.dcsc.tudelft.nl/~tkeviczky/
September 4, 2018 Systems & Control MSc Information Days
NCPS Team
Dr. Ir. Tamas Keviczky(Group head)
Dr. Ir. Manuel Mazo Jr.
Dr. Ir. Peyman Mohajerin Esfahani
Dr. Ir. Giulia Giordano
6 PhD students2 Post Doctoral researchers
Perspective on CPS
�= f(�, υ)
Cyber-Physical
Collision Avoidance
Combustion Control
ABS/ESP
CAN/FlexRay
dx/dt = f(x,u)
Networked CPS
September 4, 2018 5
Zoo of Systems
Interconnection
(Sub)systemdynamics
COMPLEXITY
CO
MP
LEX
ITY
The Challenge of Distributed Control
?
The Challenge of Distributed Control
The Challenge of Distributed Control
?????
Optimization and Team Decision Making
How can we distribute optimal team decision-making?
How can this work in a real-time control system?
What information should be exchanged and when?
What is the impact of the communication topology?
…
Pick an MSc project
in Networked Cyber Physical Systems
1. Increased autonomy via embedded, real-time optimization-based solutions for control and estimation
Research Agenda
Utilize problem structure, parallel computing,for real-time solutions in a closed feedback loop
Subsystem interactions(dynamics, objective, constraints)
Componentsubsystems
2. Distributed decision-making for interconnected systems in an optimization-based framework
Research Agenda
Distributed optimization as a paradigm fordistributed control and decision-making
Subsystem interactions(dynamics, objective, constraints)
Componentsubsystems
3. Decision-making under uncertainty
Research Agenda
Scalability/ComputationUncertainty/InformationDynamic nature
Control Theory
Applied Probability
Optimization
Interdisciplinary solutions
MSc projects
Security-aware Control Synthesis in Networks
• Your are a white-hat hacker!
• Game between the system operator and the attacker
• Mimic natural system behaviors while maximally damage
the system!
• How to mitigate such a damage
Sustainable Buildings: an Advanced Diagnosis Approach
• Modeling the building dynamics
• Design filter to identify certain abnormalities
• Validate the performance through the real measurements
Joint project with DWA (www.dwa.nl)
Dynamical System Filter
©DWA
4. Control under communication constrains
Research Agenda
Event-based control advances: Modelling of traffic, Security, Implementations, …
5. Correct-by-design synthesis of controllers
Research Agenda
Population
+
b
x
cb
a+a
Genetic operatiors
Fitness assessmentand parent selection
Controller
User-defined grammar
Formal verification
#1
#2
Temporal logicproperties
Formal methods in control: Abstraction/refinement, Symbolic regression, SMT…
Talk onThursday
6. Networked systems
Research Agenda
Network-decentralised control, coordination and estimation
Network-decentralized height/position estimation for cartography: infer a global map based on local information exchanges.
Network-decentralised coordination of robots with collision avoidance.
Interacting dynamic subsystems with local information and local action
Talk onMonday
6. Networked systems
Research Agenda
Structural analysis of biological systems
Natural and biological systems global behaviour (complex interplay of local interactions) astoundingly robust to environmental changes, fluctuations and perturbations.
• “Verification” of structural properties: given a graph structure, generate random functions that satisfy the assumptions and check that the property always holds.
• Simulation-based/analytical study of biological systems and chemical reaction networks.
Structural analysis: assess whether a class of systems always enjoys a given property, due to its structure (topology of the interaction graph) and not to specific parameter values.
Automotive and Mobile Robot Applications• Vehicle platooning
– Nonlinear trajectory planning (driver assist, traffic constraints)
– Optimize fuel economy, ride comfort– Optimize communications– Verifiable Safety
• Driving on the limits of handling
– Autonomous driving, complex vehicle dynamics
– Exploit problem structure to enable real-time computations
• Distributed Robotic Networks
– Connectivity maximization– Distributed estimation,
SLAM in active sensor networks– Emergent behaviours, swarms
Aeronautical and Space Applications
• Reconfigurable fault-tolerant predictive flight control
– Real-time, verifiable computation of constrained optimal control laws
– Online optimization-based robust fault estimation for constrained nonlinear systems
• Satellite proximity operations
• Distributed control of spacecraft formations and swarms
Large-Scale Infrastructure Applications
• CPS security and privacy
• Wireless efficient control
• Demand side management
• Distributed control of water/power/heat networks
• Distributed stochastic MPC of Aquifer Thermal Energy Storage smart grids
• Power-to-X
• Smart building control
• Cooperative control of offshore wind farm power networks
Canal networkHydro power network
September 4, 2018 22
Zoo of Systems
Interconnection
(Sub)systemdynamics
COMPLEXITY
CO
MP
LEX
ITY
Real-Time Nonlinear Trajectory PlanningFor Long Heavy Truck Combinations
• Exploit problem structure to enable real-time computations
• Serve as a driver assist system and in future autonomous driving solutions
MScProject
• Respect vehicle dynamics and constraints imposed by the truck and the environment
• Optimize fuel economy, ride comfort, etc.
September 4, 2018 24
Zoo of Systems
Interconnection
(Sub)systemdynamics
COMPLEXITY
CO
MP
LEX
ITY
September 4, 2018 25
Reconfigurable Control Design inAerospace Applications
• Integrated fault detection, isolation, and predictive control design
• Real-time and verifiable computation of constrained optimal control laws
• Improve penetration of state-of-the-art methods into industrial practice
MScProject
September 4, 2018 26
Zoo of Systems
Interconnection
(Sub)systemdynamics
COMPLEXITY
CO
MP
LEX
ITY
September 4, 2018 27
Multi-Agent Systems
Different agents (e.g. robots, spacecraft) have to collaborate
autonomously in order to reach a common goal
Organic Air Vehicle Formation Flight
September 4, 2018 30
Distributed Control and FDIR Design for Satellite Formations
MScProject
Delfi-C3
• Individual satellites are tightly coupled by performance requirement
• FDIR is critical for mission success
• Single central computing unit is undesirable
• Collaboration with ongoing space projects at Aerospace Faculty (e.g. Delfi nanosat)
September 4, 2018 31
Cooperative IndoorFormation Flight Control
• Cooperating quadrotor fleet
• Sensor network-based indoor localization scheme
• Coupled dynamics and control task of lifting and transporting a hung mass
• Optimal synchronization problem with constrained subsystem dynamics
• Implementation in Networked Embedded Robotics Lab
MScProject
In collaboration with LR, Micro Air Vehicle Group (Bart Remes, Christoph De Wagter)
Check out the recent IMAV competitions!!!
September 4, 2018 32
Distributed Control of Multi-Agent Systems on a Mobile Robot Testbed
MScProject
iRobot CreateCricket RF-USrange sensor
Networked Embedded Robotics Lab
September 4, 2018 34
Zoo of Systems
Interconnection
(Sub)systemdynamics
COMPLEXITY
CO
MP
LEX
ITY
September 4, 2018 35
Distributed Optimization and MPC for Large-Scale Infrastructures
• Decomposition methods in optimization and dynamic programming
• Application to Distributed MPC schemes
• Study of performance versus uncertainty in DMPC schemes
• Achievable performance bounds
MScProject
Canal network
Traffic network
Hydro power network
September 4, 2018 36
Distributed Optimization and MPC for Large-Scale Infrastructures
• Decomposition methods in optimization and dynamic programming
• Application to Distributed MPC schemes
• Study of performance versus uncertainty in DMPC schemes
• Achievable performance bounds
MScProject
Distributed Optimization and MPC forHigh-Performance Buildings
MScProject
September 4, 2018 38
Exploit Problem Structure toReduce Complexity of Control Design
• Enable control design for large-scale systems by exploiting symmetry in
– interconnection– subsystem dynamics
• Construct distributed (structured) controllers, estimators
• Respect global objectives and analyze performance of distributed solution
MScProject
September 4, 2018 39
Exploit Problem Structure toReduce Complexity of Control Design
• Enable control design for large-scale systems by exploiting symmetry in
– interconnection– subsystem dynamics
• Construct distributed (structured) controllers, estimators
• Respect global objectives and analyze performance of distributed solution
MScProject
September 4, 2018 40
Distributed Constraint Fulfillment
• Design methods for coupled constraints (e.g. collision avoidance)
• Guaranteed feasibility in distributed MPC schemes
• Approximation schemes, controlled invariant sets and reachability
• Robust constraint fulfillment with negotiation
• Reducing conservativeness
MScProject
September 4, 2018 41
Consensus inDistributed Predictive Control
• Interplay between consensus seeking and MPC
• Incremental subgradient methods
• Optimal synchronization problems with constrained subsystem dynamics
• Application to multi-vehicle coordination, oscillator networks, etc.
MScProject
September 4, 2018 42
Distributed Moving Horizon Estimation
• Linear and nonlinear distributed solutions for dynamic systems
• Approximation of arrival cost with local particle filters
• Respecting physical constraints
• Local unobservability, information exchange with neighboring sensors
MScProject
September 4, 2018 43
Some Recommended Courses
• SC42095 Digital Control
• SC42010 Robust and Multivariable Control Design
• SC42040 Adaptive and Predictive Control
• SC42055 Optimization in Systems and Control
• WI4218 Convex Optimization and Semi-Definite Programming
• SC42100 Networked and Distributed Control Systems
• SC42075 Modeling and Control of Hybrid Systems
• SC42060 Modeling and Nonlinear Systems Theory
• AE4-305 (Spacecraft Attitude Dynamics and Control)
September 4, 2018 44
Preparation for Control Theory Course
• Refresh linear algebra knowledge(see material also on Brightspace)
• Order textbook(B. Friedland, Control System Design – An Introduction to State-Space Methods)