Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control最新文献

{"title":"k-Inductive Barrier Certificates for Stochastic Systems","authors":"Mahathi Anand, Vishnu Murali, Ashutosh Trivedi, Majid Zamani","doi":"10.1145/3501710.3519532","DOIUrl":"https://doi.org/10.1145/3501710.3519532","url":null,"abstract":"Barrier certificates are inductive invariants that provide guarantees on the safety and reachability behaviors of continuous dynamical systems. For stochastic dynamical systems, barrier certificates take the form of inductive “expectation” invariants. In this context, a barrier certificate is a non-negative real-valued function over the state space of the system satisfying a strong supermartingale condition: it decreases in expectation as the system evolves The existence of barrier certificates, then, provides lower bounds on the probability of satisfaction of safety or reachability specifications over unbounded-time horizons. Unfortunately, establishing supermartingale conditions on barrier certificates can often be restrictive. In practice, we strive to overcome this challenge by utilizing a weaker condition called c-martingale that permits a bounded increment in expectation at every time step; unfortunately this only guarantees the property of interest for a bounded time horizon. The idea of k-inductive invariants, often utilized in software verification, relaxes the need for the invariant to be inductive with every transition of the system to requiring that the invariant holds in the next step if it holds for the last k steps. This paper synthesizes the idea of k-inductive invariants with barrier certificates. These refinements that we dub as k-inductive barrier certificates relax the supermartingale requirements at each time step to supermartingale requirements in k-steps with potential c-martingale requirements at each step, while still providing unbounded-time horizon probabilistic guarantees. We characterize a notion of k-inductive barrier certificates for safety and two distinct notions of k-inductive barrier certificates for reachability. Correspondingly, utilizing such k-inductive barrier certificates, we obtain probabilistic lower bounds on the satisfaction of safety and reachability specifications, respectively. We present a computational method based on sum-of-squares (SOS) programming to synthesize suitable k-inductive barrier certificates and, demonstrate the effectiveness of the proposed methods via some case studies.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116440267","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Verifying Neural Network Controlled Systems Using Neural Networks","authors":"Qingye Zhao, Xin Chen, Zhuoyu Zhao, Yifan Zhang, Enyi Tang, Xuandong Li","doi":"10.1145/3501710.3519511","DOIUrl":"https://doi.org/10.1145/3501710.3519511","url":null,"abstract":"Safety verification is an essential requirement of neural network controlled systems when they are adopted in safety-critical fields. This paper proposes a novel approach to synthesizing neural networks as barrier certificates, which can provide safety guarantees for neural network controlled systems. We first propose the construction conditions of neural network barrier certificates, followed by an iterative framework to synthesize them. Each iteration trains a neural network as the candidate barrier certificate using the training datasets sampled from the neural network controlled system. After training, identifying whether the candidate barrier certificate is a real one for the neural network controlled system is transformed into a group of mixed-integer programming problems, which the numerical optimization solver solves with guaranteed results. We implement the tool NetBC and evaluate its performance over 6 practical benchmark examples. The experimental results show that NetBC is more effective and scalable than the existing polynomial barrier certificate-based method.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"123 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127041670","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Correct-By-Construction Exploration and Exploitation for Unknown Linear Systems Using Bilinear Optimization","authors":"Kwesi J. Rutledge, N. Ozay","doi":"10.1145/3501710.3519536","DOIUrl":"https://doi.org/10.1145/3501710.3519536","url":null,"abstract":"This paper addresses the problem of controlling an unknown dynamical system to safely reach a target set. We assume we have a priori access to a finite set of uncertain linear systems, to which the unknown system belongs to. This set can contain models for different failure or operational modes or potential environmental conditions. Given a desired exploration-exploitation profile, we provide a bilinear optimization based solution to this control synthesis problem. Our approach provides a family of controllers that enable adaptation based on data observed at run-time to automatically trade off model detection and reachability objectives while maintaining safety. We demonstrate the approach with several examples.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133380698","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimality and Asymptotic Stability in Two-Player Zero-Sum Hybrid Games","authors":"S. J. Leudo, R. Sanfelice","doi":"10.1145/3501710.3526948","DOIUrl":"https://doi.org/10.1145/3501710.3526948","url":null,"abstract":"In this work, we formulate a two-player zero-sum game under dynamic constraints given in terms of hybrid dynamical systems. Find the full version in [8], including the main results and outlines of the corresponding proofs. We propose sufficient conditions to guarantee attaining a solution to the game. When the players select the optimal strategy, the value function can be evaluated without the need of computing solutions. Under additional conditions, the optimal feedback laws render a set of interest asymptotically stable. Using this framework, we address an optimal control problem under the presence of an adversarial action in which the decision-making agents have dynamics that might exhibit both continuous and discrete behavior.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"113 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122552328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Requirement Testing Using Focused Falsification","authors":"Johan Lidén Eddeland, Alexandre Donzé, K. Åkesson","doi":"10.1145/3501710.3519521","DOIUrl":"https://doi.org/10.1145/3501710.3519521","url":null,"abstract":"Testing of Cyber-Physical Systems (CPS) deals with the problem of finding input traces to the systems such that given requirements do not hold. Requirements can be formalized in many different ways; in this work requirements are modeled using Signal Temporal Logic (STL) for which a quantitative measure, or robustness value, can be computed given a requirement together with input and output traces. This value is a measure of how far away the requirement is from not holding and is used to guide falsification procedures for deciding on new input traces to simulate one after the other. When the system under test has multiple requirements, standard approaches are to falsify them one-by-one, or as a conjunction of all requirements, but these approaches do not scale well for industrial-sized problems. In this work we consider testing of systems with multiple requirements by proposing focused multi-requirement falsification. This is a multi-stage approach where the solver tries to sequentially falsify the requirements one-by-one, but for every simulation also evaluate the robustness value for all requirements. After one requirement has been focused long enough, the next requirement to focus is selected by considering the robustness values and trajectory history calculated thus far. Each falsification attempt makes use of a prior sensitivity analysis, which for each requirement estimates the parameters that are unlikely to affect the robustness value, in order to reduce the number of parameters that are used by the optimization solver. The proposed approach is evaluated on a public benchmark example containing a large number of requirements, and includes a comparison of the proposed algorithm against a new suggested baseline method.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"46 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120899343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ABS: A formally correct software tool for space-efficient symbolic synthesis","authors":"Alexander Weber, Elisei Macoveiciuc, G. Reissig","doi":"10.1145/3501710.3519519","DOIUrl":"https://doi.org/10.1145/3501710.3519519","url":null,"abstract":"We present ABS, a software for Abstraction-Based Synthesis of controllers for continuous-state control systems. The tool distinguishes itself from previously known such software by being formally correct, i.e., any controller synthesized by ABS is mathematically guaranteed to solve the control problem provided as input. ABS achieves this quality by providing an input language with mathematically defined semantics and a respective compiler, and by carefully taking into account all numerical and rounding errors that might be incurred at either compile- or run-time. To mitigate computational overhead caused by the aforementioned approach, ABS implements, e.g. on-the-fly synthesis algorithms with greatly reduced memory requirement. The tool is currently applicable to invariance and reachability problems and requires state measurement. We discuss structure, algorithmic details and basic usage of ABS, and we demonstrate on two examples that its performance compares favorably with that of competing, not formally correct synthesis software. The source code of ABS is publicly available. See http://www.reiszig.de/gunther/pubs/ABS.html","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128463582","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Poster Abstract: Learning from Demonstrations with Temporal Logics","authors":"Aniruddh Gopinath Puranic, Jyotirmoy V. Deshmukh, S. Nikolaidis","doi":"10.1145/3501710.3524914","DOIUrl":"https://doi.org/10.1145/3501710.3524914","url":null,"abstract":"Learning-from-demonstrations (LfD) is a popular paradigm to obtain effective robot control policies for complex tasks via reinforcement learning without the need to explicitly design reward functions. However, it is susceptible to imperfections in demonstrations and also raises concerns of safety and interpretability in the learned control policies. To address these issues, we propose to use Signal Temporal Logic (STL) to express high-level robotic tasks and use its quantitative semantics to evaluate and rank the quality of demonstrations. Temporal logic-based specifications allow us to create non-Markovian rewards, and are also capable of defining interesting causal dependencies between tasks such as sequential task specifications. We present our completed work that proposed LfD-STL framework that learns from even suboptimal/imperfect demonstrations and STL specifications to infer rewards for reinforcement learning tasks. We have validated our approach through various experimental setups to show how our method outperforms prior LfD methods. We then discuss future directions for tackling the problem of explainability and interpretability in such learning-based systems.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121370387","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sufficient Conditions for Optimality and Asymptotic Stability in Two-Player Zero-Sum Hybrid Games","authors":"S. J. Leudo, R. Sanfelice","doi":"10.1145/3501710.3519514","DOIUrl":"https://doi.org/10.1145/3501710.3519514","url":null,"abstract":"In this paper, we formulate a two-player zero-sum game under dynamic constraints given in terms of hybrid dynamical systems. We present sufficient conditions with Hamilton-Jacobi-Isaacs-like equations to guarantee attaining a solution to the game. It is shown that when the players select the optimal strategy, the value function can be evaluated without the need of computing solutions. Under additional conditions, we show that the optimal feedback laws render a set of interest asymptotically stable. Using this framework, we address an optimal control problem under the presence of an adversarial action in which the decision-making agents have dynamics that might exhibit both continuous and discrete behavior. Applications of this problem, as presented here, include disturbance rejection and security scenarios, for which the effect of the worst-case adversarial action is minimized.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126306991","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Linear Time Monitoring for One Variable TPTL","authors":"Bassem Ghorbel, Vinayak S. Prabhu","doi":"10.1145/3501710.3519537","DOIUrl":"https://doi.org/10.1145/3501710.3519537","url":null,"abstract":"The temporal logic Timed Propositional Temporal Logic () extends with freeze quantifiers in order to express timing constraints, and is strictly more expressive than Metric Temporal Logic () over future modalities. The monitoring problem is to check whether a particular timed trace satisfies a given temporal logic specification, and monitoring procedures form core subroutines of testing and falsification approaches for Cyber-Physical Systems. In this work, we develop an efficient linear time monitoring algorithm, linear in the length of the trace (for traces that have at most a constant number of sample points in any unit interval), for one variable in the pointwise semantics. This one variable fragment is known to be already more expressive than and thus allows specifications of richer timed properties. Our algorithm carefully combines a divide and conquer approach with dynamic programming in order to achieve a linear time algorithm. As a plus, our algorithm uses only a simple two-dimensional table, and a syntax tree of the formula, as the data structures, and hence can be easily implemented on various platforms. We demonstrate the tractability of our approach with our prototype tool implementation on Matlab; our experiments show the tool scales easily to long trace lengths.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130461776","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mortality and Edge-to-Edge Reachability are Decidable on Surfaces","authors":"Mateus de Oliveira Oliveira, O. Tveretina","doi":"10.1145/3501710.3519529","DOIUrl":"https://doi.org/10.1145/3501710.3519529","url":null,"abstract":"The mortality problem for a given dynamical system S consists of determining whether every trajectory of S eventually halts. In this work, we show that this problem is decidable for the class of piecewise constant derivative systems on two-dimensional manifolds, also called surfaces (). Two closely related open problems are point-to-point and edge-to-edge reachability for . Building on our technique to establish decidability of mortality for , we show that the edge-to-edge reachability problem for these systems is also decidable. In this way we solve the edge-to-edge reachability case of an open problem due to Asarin, Mysore, Pnueli and Schneider [4]. This implies that the interval-to-interval version of the classical open problem of reachability for regular piecewise affine maps (PAMs) is also decidable. In other words, point-to-point reachability for regular PAMs can be effectively approximated with arbitrarily precision.","PeriodicalId":194680,"journal":{"name":"Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114644199","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}