SIMULATION TO PROOFS IN C2E2 Parasara Sridhar Duggirala 1
A simple (often the only) strategy π β’ Given start and target S β’ Compute finite cover of initial set β’ Simulate from the center π¦ 0 of each cover β’ Bloat simulation so that bloated tube contains all trajectories from the cover π¦ 0 β’ Union = over-approximation of reach set β’ Check intersection/containment with π β’ Refine [Girard et al 2006], [Donze et al 2008],... β’ How much to bloat? (obviously incomplete) β’ How to handle mode switches? 2
Discrepancy (Spirit of Loop Invariants) . Definition. πΎ: β 2π Γ β β₯0 β β β₯0 defines a discrepancy of the system if for any two states π¦ 1 and π¦ 2 β π , For any t, 1. |π π¦ 1 , π’ β π π¦ 2 , π’ | β€ πΎ π¦ 1 , π¦ 2 , π’ and 2. πΎ β 0 as π¦ 1 β π¦ 2 π¦ β 0 invariant π¦ β€ 10 until π¦ β₯ 10 do βπ π¦ 1 , π’ π¦ β π¦ + 1 βπ π π¦ 1 , π’ , π π¦ 2 , π’ βπΎ π¦ 1 , π¦ 2 , π’ od 3
Computing Discrepancy . If L is a Lipschitz constant for f(x,t) then |π π¦ 1 , π’ β π π¦ 2 , π’ | β€ π ππ’ π¦ 1 β π¦ 2 . π¦ = π΅π¦ Lyapunov function π¦ π ππ¦ that proves exponenial If stability, then |π π¦ 1 , π’ β π π¦ 2 , π’ | β€ πΏπ πΏπ’ π¦ 1 β π¦ 2 where πΏ = πΊπ£ππ(π) Similar observation by [Deng et al 2013] What about Nonlinear Systems? 4
Computing Discrepancy . If M is a contraction metric, that is, a positive definite matrix such that βπ π > 0 : πΎ π π + π πΎ + π π π βΌ 0, where J is the 2 β€ Jacobian for f, then βπ, π > 0 such that π π¦, π’ β π π¦ β² , π’ π π¦ β π¦ β² 2 π βππ’ [Lohmiller & Slotine β98]. New algorithm: computes local discrepancy by estimating maximum eigenvalue of the Jacobian matrix over a neighborhood [Fan & Mitra 2014]. Inferring Contraction Metric from simulations [Balkan et al 2014] What next? 5
Simulations+Annotation ο Reachtubes ππππππππππ(π π , π, π, πΌ) of gives a sequence S 0 , β¦ , π π : πππ π π β€ π & at any time π’ β [πβ, π + 1 β] , solution π π¦ 0 , π’ β π π . ππππ πππππ π», π, πΌ of π¦ = π π¦ is a sequence π 0 , β¦ , π π such that πππ(π π ) β€ π and from any π¦ 0 β π, for each time π’ β [πβ, (π + 1)β] , π π¦ 0 , π’ β π π . π 0 , β¦ , π π , π 1 β π€πππππ(π¦ 0 , π, π) For each π β [π] π 2 β sup πΎ π¦ 1 , π¦ 2 , π’ π’βπ π ,π¦,π¦ β² βπΆ π (π¦ 0 ) π π β πΆ π 2 π π 6
How to get completeness for hybrid systems? Track & propagate πππ§ and ππ£π‘π’ fragments of reachtube ππ£π‘π’ π β π ππππΊπππππ πΊ, πΈ = πππ§ π β© π β β πππ’ π β© π = β ππππππππππΈπππππ(π, π») = β©π 0 , π’ππ 0 , β¦ , π π , π’ππ π βͺ , such that either β² π‘ before it are must π’ππ π = ππ£π‘π’ if all the π π β² π‘ before it are at least may π’ππ π = πππ§ if all the π π and at least one of them is not must 7
Hybrid Reachtubes: Guards & Resets πππππΊππππππ(π) returns a set of tagged regions N. β π iff β π, π π such that π β² = πππ‘ππ’ π π π and: πβ², π’ππβ² π π β π»π£ππ π π , π’ππ π = π’ππ β² = ππ£π‘π’ π π β© π»π£ππ π π β β , π π β π»π£ππ π π , π’ππ π = ππ£π‘π’, π’ππ β² = πππ§ π π β© π»π£ππ π π β β , π’ππ π = π’ππβ² = πππ§ 8
Sound & Relatively Complete. Theorem. (Soundness). If Algorithm returns safe or unsafe, then π΅ is safe or unsafe. Definition Given HA π΅ = β©π, πππ, π΅, πΈ, π βͺ , an π -perturbation of A is a new HA π΅β² that is identical except, Ξ β² = πΆ π (Ξ) , β β β πππ, π½ππ€ β² = πΆ π (π½ππ€) (b) a β A, π»π£ππ π π = πΆ π (π»π£ππ π π ) . A is robustly safe iff βπ > 0 , such that Aβ is safe for π π upto time bound T, and transition bound N. Robustly unsafe iff β π < 0 such that π΅β² is safe for π π . Theorem. (Relative Completeness) Algorithm always terminates whenever the A is either robustly safe or robustly unsafe. 9
C2E2 10
Part II TWO APPLICATIONS Duggirala β Wang β Mitra β Munoz β Viswanathan (FM 2014) Huang β Fan β Meracre β Mitra β Kiwatkowska (CAV 2014) 11
SAPA-ALAS Parallel Landing Protocol Ownship and Intruder approaching parallel runways with small separation ALAS (at ownship) protocol is supposed to raise an ππΌ alarm if within T time units the Intruder can violate π¦π‘ππ safe separation based on 3 different projections ππΊ Verify Alert βΌ π Unsafe for different runway and aircraft ππΆ π§π‘ππ scenarios Scenario 1. With xsep [.11,.12] Nm ysep [.1,.21] Nm, π = 30 π π πππ¦ = 45 o vy o = 136 Nmph, vy i = 155 Nmph Duggirala, Wang, Mitra, Munoz, Viswanathan FM 2014 12
SAPA-ALAS Parallel Landing Protocol π΅πππ π’ π = π¦ β π’ β 0, π , ππ ππ π π¦, π’ β πππ‘πππ} , where ππ ππ π defined as solution of ODE π¦ = π π (π¦, π’) ππΌ π¦π‘ππ Use simulations and annotations of π π to compute ππΊ ππ£π‘π’ intervals when π¦ β π΅πππ π’ π ππΆ π§π‘ππ π΅πππ π’ βΊ π π 2 is satisfied by Reachtube π if β π½ 2 β ππ£π‘π’ π 2 βͺ πππ§ π 2 there exists π½ 1 β ππ£π‘π’ π΅πππ π’ such that π½ 1 < π½ 2 β π π΅πππ π’ βΊ π π 2 is violated by Reachtube π if β π½ 2 β ππ£π‘π’ π 2 for all π½ 1 β ππ£π‘π’ π΅πππ π’ βͺ πππ§ π΅πππ π’ such that π½ 1 > π½ 2 β π Duggirala, Wang, Mitra, Munoz, Viswanathan FM 2013 13
Real-time Alerting Protocol . Alert βΌ 4 Alert βΌ ? Running time Scenario Unsafe (mins:sec) Unsafe 6 False 3:27 2.16 7 True 1:13 β 8 True 2:21 β 6.1 False 7:18 1.54 7.1 True 2:34 β 8.1 True 4:55 β Sound & robustly completeness 9 False 2:18 1.8 10 False 3:04 2.4 C2E2 verifies interesting scenarios in 9.1 False 4:30 1.8 reasonable time; shows that false 10.1 False 6:11 2.4 alarms are possible; found scenarios where alarm may be missed 14
Exploiting Modularity Module 1 Module 1 ? Module 2 Module 3 Module 2 Module 3 Module 5 Module 4 π π π¦ 1 = π π (π¦ 1 , π¦ 2 , π¦ 3 ) Γ π π π¦ 2 = π π (π¦ 2 , π¦ 1 , π¦ 3 ) π¦ 3 = π π (π¦ 3 , π¦ 1 , π¦ 2 ) π π π π 15
Input-to-State (IS) Discrepancy π£(π’) π(π¦, π£, π’) π£ π¦ π¦ = π(π¦, π£) π£β²(π’) π(π¦ β² , π£ β² , π’) π¦β² time time π’ Definition. IS discrepancy is defined by πΎ and πΏ such that for any initial states π¦, π¦ β² and any inputs π£, π£ β² , π’ πΏ |π£ π‘ β π£ β² π‘ | ππ‘ |π(π¦, π£, π’) β π π¦ β² , π£ β² , π’ | β€ πΎ(π¦, π¦ β² , π’) + 0 πΎ β 0 as π¦ β π¦β² , and πΏ β 0 as π£ β π£ β² 16
Reduced System π(π 1 , π 2 , π 1 , π 2 ) . π¦ = π π π¦ π¦ = β©π 1 , π 2 , πππβͺ π 1 πΎ 1 π 1 , πππ + πΏ 1 π 2 π 2 = π π π¦ = πΎ 2 π 2 , πππ + πΏ 2 π 1 πππ 1 17
Bloating with Reduced Model π 1 = πΎ 1 π, π’ π¦ 1 = π 1 (π¦ 1 , π£ 1 ) +πΏ 1 (π 2 , π 3 ) π 2 = πΎ 2 π, π’ π 3 = πΎ 3 π, π’ +πΏ 2 (π 1 , π 3 ) π¦ 3 = π 3 (π¦ 3 , π£ 3 ) +πΏ 3 (π 1 , π 2 ) π¦ 2 = π 2 (π¦ 2 , π£ 2 ) π(π’) π(π’) π π¦ π(π’) time time The bloated tube contains all trajectories start from the π -ball of π¦ . The over-approximation can be computed arbitrarily precise. 18
Reduced π gives effective Discrepancy of π΅ . Theorem. For any π = β©π 1 , π 2 βͺ , π = β©π 1 , π 2 βͺ and π π ππππβ π΅ πΆ π π¦ , π β π’β€π πΆ π π’ (π π¦, π’ ) Theorem. For any Ο΅ > 0 there exists Ξ΄ = β©Ξ΄ 1 , Ξ΄ 2 βͺ such that π π’β€π πΆ π π’ (π π¦, π’ ) β πΆ π (ππππβ π΅ (πΆ π π¦ , π) Here π π’ is the solution of π(π 1 , π 2 , π 1 , π 2 ) . Huang & Mitra, HSCC 2013 19
Pacemaker + Cardiac Network . Action potential remains in specific range No alternation of action potentials Nodes Thresh Sims Run time (s) Property 3 2 16 104.8 TRUE 3 1.65 16 103.8 TRUE 5 2 3 208 TRUE 5 1.65 5 281.6 TRUE 5 1.5 NA 63.4 FALSE 8 2 3 240.1 TRUE 20 8 1.65 73 2376.5 TRUE
Recommend
More recommend