15-424/15-624: Foundations of Cyber-Physical Systems
Transcription
15-424/15-624: Foundations of Cyber-Physical Systems
15-424/15-624: Foundations of Cyber-Physical Systems 01: Overview André Platzer [email protected] Carnegie Mellon University, Pittsburgh, PA http://symbolaris.com/course/fcps14.html http://www.cs.cmu.edu/~aplatzer/course/fcps14.html 0.8 1.0 0.2 0.4 0.6 0.5 0.4 0.3 0.2 0.1 André Platzer (CMU) FCPS/01: Overview FCPS 1 / 24 Outline 1 CPS: Introduction Hybrid Systems & Cyber-Physical Systems Applications Robot Labs 2 15-424: Foundations of Cyber-Physical Systems Objectives Outline Assessment Labs Resources 3 Approach CPS Contracts CPS Logic Differential Dynamic Logic Family 4 Summary André Platzer (CMU) FCPS/01: Overview FCPS 1 / 24 Can you trust a computer to control physics? André Platzer (CMU) FCPS/01: Overview FCPS 2 / 24 Can you trust a computer to control physics? Rationale 1 Safety guarantees require analytic foundations 2 Foundations revolutionized digital computer science & society 3 Need even stronger foundations when software reaches out into our physical world Cyber-physical Systems CPS combine cyber capabilities with physical capabilities to solve problems that neither part could solve alone. How can we provide people with cyber-physical systems they can bet their lives on? – Jeannette Wing André Platzer (CMU) FCPS/01: Overview FCPS 2 / 24 Can you trust a computer to control physics? Rationale 1 Safety guarantees require analytic foundations 2 Foundations revolutionized digital computer science & society 3 Need even stronger foundations when software reaches out into our physical world Cyber-physical Systems CPS combine cyber capabilities with physical capabilities to solve problems that neither part could solve alone. 1 CPS Foundations: intellectual grand challenge 2 Research & Industry applications abound André Platzer (CMU) FCPS/01: Overview FCPS 2 / 24 Outline 1 CPS: Introduction Hybrid Systems & Cyber-Physical Systems Applications Robot Labs 2 15-424: Foundations of Cyber-Physical Systems Objectives Outline Assessment Labs Resources 3 Approach CPS Contracts CPS Logic Differential Dynamic Logic Family 4 Summary André Platzer (CMU) FCPS/01: Overview FCPS 2 / 24 CPS Analysis: Car Control Challenge (Hybrid Systems) Fixed rule describing state evolution with both Discrete dynamics (control decisions) Continuous dynamics (differential equations) x v a 7 3.0 2 6 2.5 m 5 1 2.0 4 0 1 2 3 4 5 t 1.5 3 1.0 -1 0.5 -2 0.0 0 André Platzer (CMU) 2 1 1 2 3 4 FCPS/01: Overview 0 5t 0 1 2 3 4 FCPS 5 t 3 / 24 Hybrid Systems & Cyber-Physical Systems Mathematical model for complex physical systems: Definition (Hybrid Systems) systems with interacting discrete and continuous dynamics Technical characteristics: Definition (Cyber-Physical Systems) (Distributed network of) computerized control for physical system Computation, communication and control for physics André Platzer (CMU) FCPS/01: Overview FCPS 4 / 24 What CPS are around us? What CPS will be around us in the future? Which CPS do we trust with our lives? André Platzer (CMU) FCPS/01: Overview FCPS 5 / 24 Successful CPS Proofs (rx , ry ) fy (vx , vy ) ey (lx , ly ) xb ex fx FM’11,LMCS’12,ICCPS’12,ITSC’11,ITSC’13,IJCAR’12 André Platzer (CMU) FCPS/01: Overview FCPS 6 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accelerate / brake (discrete dynamics) 2.0 1.5 1.0 1D motion (continuous dynamics) 0.5 0.0 0 1 v a 2 3 4 6 5 p 0.2 5 0.8 0.1 2 4 6 8 10 t 4 0.6 3 0.4 -0.1 px 2 0.2 -0.2 1 -0.3 2 André Platzer (CMU) 4 6 8 FCPS/01: Overview 10 t py 2 4 6 8 FCPS 10 t 7 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accelerate / brake (discrete dynamics) 2.0 1.5 1.0 1D motion (continuous dynamics) 0.5 0.0 0 a 1 2 3 4 6 5 d Ω 1.0 0.2 dx 0.00008 0.1 0.8 0.00006 2 4 6 8 10 t 0.6 0.00004 0.4 -0.1 0.00002 0.2 -0.2 2 -0.3 4 6 8 10 t 2 André Platzer (CMU) FCPS/01: Overview 4 dy 6 8 FCPS 10 t 7 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accelerate / brake / stop (discrete dynamics) 2.0 1.5 1.0 1D motion (continuous dynamics) 0.5 0.0 0 a 2 4 6 8 10 1 2 3 4 6 5 p v t1.2 px 3.5 1.0 3.0 0.8 2.5 0.6 2.0 -0.05 -0.10 1.5 0.4 1.0 -0.15 0.2 -0.20 2 André Platzer (CMU) py 0.5 4 6 8 FCPS/01: Overview 10 t 2 4 6 8 FCPS 10 t 8 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accelerate / brake / stop (discrete dynamics) 2.0 1.5 1.0 1D motion (continuous dynamics) 0.5 0.0 0 a 2 4 6 8 10 1 2 3 4 d Ω t0.00012 1.0 dx 0.00010 -0.05 6 5 0.5 0.00008 0.00006 -0.10 2 4 6 dy 8 10 t 0.00004 -0.15 -0.5 0.00002 2 -0.20 André Platzer (CMU) 4 6 8 FCPS/01: Overview 10 t-1.0 FCPS 8 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accelerate / brake (discrete dynamics) 2.0 1.5 1.0 1D motion (continuous dynamics) 0.5 0.0 0 1 2 4 6 5 p v a 3 0.2 2.5 0.8 px 0.1 2.0 2 4 6 8 10 t0.6 1.5 0.4 -0.1 1.0 0.2 -0.2 -0.3 2 André Platzer (CMU) py 0.5 4 6 8 FCPS/01: Overview 10 t 2 4 6 8 FCPS 10 t 9 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accelerate / brake (discrete dynamics) 2.0 1.5 1.0 1D motion (continuous dynamics) 0.5 0.0 0 a 1 2 3 4 6 5 d Ω 1.0 0.2 dx 0.00008 0.1 0.8 2 4 6 8 10 t 0.00006 0.6 0.00004 0.4 -0.1 0.00002 0.2 -0.2 2 -0.3 4 6 8 10 t 2 André Platzer (CMU) FCPS/01: Overview 4 dy 6 8 FCPS 10 t 9 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accel / brake / steer (discrete dynamics) 2.0 1.5 1.0 2D motion (continuous dynamics) 0.5 0.0 0 1 v a 2 3 4 6 5 p 1.0 0.2 8 2 4 6 8 10 t 0.8 6 0.6 -0.2 px 4 0.4 -0.4 0.2 -0.6 2 -0.8 2 André Platzer (CMU) 4 6 8 FCPS/01: Overview 10 t py 2 4 6 8 FCPS 10 t 10 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Accel / brake / steer (discrete dynamics) 2.0 1.5 1.0 2D motion (continuous dynamics) 0.5 0.0 0 a 1 2 1.0 0.5 2 4 6 8 4 6 5 d Ω 0.2 3 10 t 2 4 6 8 10 t dx 0.5 -0.2 -0.5 -0.4 2 4 dy 6 8 10 t -1.0 -0.6 -0.5 -0.8 André Platzer (CMU) FCPS/01: Overview FCPS 10 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Dynamic obstacles (other agents) 2.0 1.5 1.0 Avoid collisions (define safety) 0.5 0.0 0 a 2 4 6 8 10 1 v t1.0 2 3 4 6 5 p px 4 0.8 -1 3 0.6 -2 2 0.4 -3 1 0.2 -4 2 André Platzer (CMU) 4 6 8 FCPS/01: Overview 10 t py 2 4 6 8 FCPS 10 t 11 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Dynamic obstacles (other agents) 2.0 1.5 1.0 Avoid collisions (define safety) 0.5 0.0 0 a 1 2 4 6 8 10 4 6 5 d Ω 2 3 t 1.0 dx 0.5 -1 0.5 2 4 6 8 10 t -2 2 -0.5 4 6 8 10 t dy -3 -1.0 -0.5 -4 André Platzer (CMU) FCPS/01: Overview FCPS 11 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Control robot (respect delays) 2.0 1.5 1.0 Environment interaction (obstacles, agents, uncertainty) 0.5 0.0 0 1 2 4 6 5 p v a 3 1.2 7 1.0 6 0.4 0.2 2 4 6 8 10 t 5 0.8 4 0.6 px 3 -0.2 0.4 2 -0.4 0.2 1 -0.6 2 André Platzer (CMU) 4 6 8 FCPS/01: Overview 10 py t 2 4 6 8 FCPS 10 t 12 / 24 CPS Analysis & Design: Robot Lab Challenge (Hybrid Systems) Design & verify controller for a robot avoiding obstacles 3.5 3.0 2.5 Control robot (respect delays) 2.0 1.5 1.0 Environment interaction (obstacles, agents, uncertainty) a 0.5 0.0 0 1 2 3 4 6 5 d Ω 1.0 0.4 0.5 dx 0.2 2 4 6 8 10 t 2 4 6 8 10 t 0.5 dy -0.5 -0.2 2 4 6 8 10 t -0.4 -1.0 -0.6 -0.5 André Platzer (CMU) FCPS/01: Overview FCPS 12 / 24 Proof Theory Modal Logic Computer Algebra Algebraic Geometry Theorem Proving Model Checking Differential Algebra Logic Algebra Lie Algebra Differential Equations Closure Ordinals Logical Foundations of Cyber-Physical Systems Fixpoints & Lattices Algorithms Proof Search Dynamical Systems Analysis Differentiation Decision Procedures Limit Processes Weierstraß Approximation Stochastics Numerics Hermite Interpolation Numerical Quadrature André Platzer (CMU) Error Analysis Doob’s Supermartingales Stochastic Differential Equations Differential Generators Dynkin’s Infinitesimal Generators FCPS/01: Overview FCPS 13 / 24 Outline 1 CPS: Introduction Hybrid Systems & Cyber-Physical Systems Applications Robot Labs 2 15-424: Foundations of Cyber-Physical Systems Objectives Outline Assessment Labs Resources 3 Approach CPS Contracts CPS Logic Differential Dynamic Logic Family 4 Summary André Platzer (CMU) FCPS/01: Overview FCPS 13 / 24 About 15-424/14-624 Foundations of CPS Foundations! Modeling & Control 1 2 3 Understand the core principles behind CPSs. Develop models and controls. Identify the relevant dynamical aspects. Computational Thinking 1 2 3 4 5 Identify safety specifications and critical properties of CPSs. Understand abstraction and system architectures. Learn how to design by invariant. Reason rigorously about CPS models. Verify CPS models of appropriate scale. CPS Skills 1 2 3 Understand the semantics of a CPS model. Develop an intuition for operational effects. Use higher-level model-predictive control. Byproducts 1 2 Exposure to numerous math areas in action. ... André Platzer (CMU) FCPS/01: Overview FCPS 14 / 24 Course Outline 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Cyber-physical systems: introduction Differential equations & domains Choice & control Safety & contracts Dynamical systems & dynamic axioms Truth & proof Control loops & invariants Events & delays Differential equations & differential invariants Differential equations & proofs Dynamic logic & dynamical systems Dynamical systems: discrete & continuous & hybrid Robots / railway / air traffic / car CPS & applications Virtual substitution & real arithmetic Hybrid systems & games André Platzer (CMU) FCPS/01: Overview FCPS 15 / 24 Assessment Read Collaboration and Academic Integrity Policy ≈22% Theory homework Policy Due at beginning of lecture ≈51% Labs, including ≈22% final project Due at 22:00 Whitepaper For final project Proposal For final project Term paper Due with final project ≈11% Midterm ≈11% Final ≈5% Participation in class and in online comments André Platzer (CMU) FCPS/01: Overview FCPS 16 / 24 Robot Labs 1 Robot on Rails a b 2 Robot on Highways a b 3 with event-driven control with time-triggered control Robot on Racetracks a b 4 Autobots, Roll Out Charging Station stay on the circular racetrack slow down to avoid collisions Robot in a Plane a b with obstacle avoidance Robot vs. Roguebot: avoid collisions with moving obstacles 5 Robot in Star-lab: self-defined final project 6 Final project presented at CPS V&V Grand Prix André Platzer (CMU) FCPS/01: Overview CPS v&V Grand Prix FCPS 17 / 24 Resources Prerequisites 15-122 Principles of Imperative Computation 21-122 Integration, Differential Equations, and Approximation (21-241 Matrix algebra or 18-202 Mathematical Foundations of Electrical Engineering or 15-251 Great Theoretical Ideas in Computer Science) You will be expected to follow extra background reading material as needed. Further reading and background material on the course web page Check course web page periodically http://symbolaris.com/course/fcps14.html KeYmaera Piazza Autolab Ask! André Platzer (CMU) FCPS/01: Overview FCPS 18 / 24 Textbook André Platzer. Logical Analysis of Hybrid Systems. Springer, 426p., 2010. DOI 10.1007/978-3-642-14509-4 http://symbolaris.com/lahs/ CMU library e-book André Platzer. Foundations of Cyber-Physical Systems. Lecture notes. Computer Science Department Carnegie Mellon University. http://symbolaris.com/course/ fcps14-schedule.html André Platzer (CMU) FCPS/01: Overview FCPS 19 / 24 Outline 1 CPS: Introduction Hybrid Systems & Cyber-Physical Systems Applications Robot Labs 2 15-424: Foundations of Cyber-Physical Systems Objectives Outline Assessment Labs Resources 3 Approach CPS Contracts CPS Logic Differential Dynamic Logic Family 4 Summary André Platzer (CMU) FCPS/01: Overview FCPS 19 / 24 CPS Design & CPS Contracts in Programs Reveal in layers HP Contracts Reason about CPS @ r e q u i r e s (0<=x & x=H & v =0) @ r e q u i r e s ( g>0 & 1>=c>=0) @ensures(0<=x & x<=H) { {x ’=v , v’=−g , x >=0}; i f ( x = 0) { v := −c ∗ v ; } }∗ @ i n v a r i a n t ( 2 ∗ g ∗ x <= 2∗ g ∗H − v ˆ2 & x>=0) CPS Simulate for intuition André Platzer (CMU) CT FCPS/01: Overview Design-by-invariant FCPS 20 / 24 CPS Contracts & CPS Logic Logic for CPS dL Contracts Reason in Logic 0<=x & x=H & v=0 & g>0 & 1>=c>=0 −> [{ {x ’=v , v’=−g , x >=0}; i f ( x = 0) { v := −c ∗ v ; } }∗ @ i n v a r i a n t ( 2 ∗ g ∗ x <= 2∗ g ∗H − v ˆ2 & x>=0) ] (0<=x & x<=H) CPS Analyze for precision André Platzer (CMU) CT FCPS/01: Overview Proof-by-invariant FCPS 21 / 24 Family of Differential Dynamic Logics c o n ti n rsar i nondet tic adv e has stoc al uous d i s cr e te JAR’08,CADE’11,LMCS’12,LICS’12,LICS’12 André Platzer (CMU) FCPS/01: Overview FCPS 22 / 24 Family of Differential Dynamic Logics differential dynamic logic [α]φ dL = DL + HP φ α c o n ti n uous d i s cr e te dGL = GL + HG SdL = DL + SHP nondet adv e φ tic hαiφ has rsar i al stochastic differential DL stoc differential game logic hαiφ φ quantified differential DL QdL = FOL + DL + QHP JAR’08,CADE’11,LMCS’12,LICS’12,LICS’12 André Platzer (CMU) FCPS/01: Overview FCPS 22 / 24 Proof Theory Modal Logic Computer Algebra Algebraic Geometry Theorem Proving Model Checking Differential Algebra Logic Algebra Lie Algebra Differential Equations Closure Ordinals Logical Foundations of Cyber-Physical Systems Fixpoints & Lattices Algorithms Proof Search Dynamical Systems Analysis Differentiation Decision Procedures Limit Processes Weierstraß Approximation Stochastics Numerics Hermite Interpolation Numerical Quadrature André Platzer (CMU) Error Analysis Doob’s Supermartingales Stochastic Differential Equations Differential Generators Dynkin’s Infinitesimal Generators FCPS/01: Overview FCPS 23 / 24 Outline 1 CPS: Introduction Hybrid Systems & Cyber-Physical Systems Applications Robot Labs 2 15-424: Foundations of Cyber-Physical Systems Objectives Outline Assessment Labs Resources 3 Approach CPS Contracts CPS Logic Differential Dynamic Logic Family 4 Summary André Platzer (CMU) FCPS/01: Overview FCPS 23 / 24 Can you trust a computer to control physics? André Platzer (CMU) FCPS/01: Overview FCPS 24 / 24 Can you trust a computer to control physics? Rationale 1 Safety guarantees require analytic foundations 2 Foundations revolutionized digital computer science & society 3 Need even stronger foundations when software reaches out into our physical world Cyber-physical Systems CPS combine cyber capabilities with physical capabilities to solve problems that neither part could solve alone. How can we provide people with cyber-physical systems they can bet their lives on? – Jeannette Wing André Platzer (CMU) FCPS/01: Overview FCPS 24 / 24 Can you trust a computer to control physics? Rationale 1 Safety guarantees require analytic foundations 2 Foundations revolutionized digital computer science & society 3 Need even stronger foundations when software reaches out into our physical world Cyber-physical Systems CPS combine cyber capabilities with physical capabilities to solve problems that neither part could solve alone. 1 CPS Foundations: intellectual grand challenge 2 Research & Industry applications abound André Platzer (CMU) FCPS/01: Overview FCPS 24 / 24