Abstract:This paper is concerned with a compositional approach for the construction of control barrier certificates for large-scale interconnected stochastic systems while synthesizing hybrid controllers against high-level logic properties. Our proposed methodology involves decomposition of interconnected systems into smaller subsystems and leverages the notion of control sub-barrier certificates of subsystems, enabling one to construct control barrier certificates of interconnected systems by employing some max-type small-gain conditions. The main goal is to synthesize hybrid controllers enforcing complex logic properties including the ones represented by the accepting language of deterministic finite automata, while providing probabilistic guarantees on the satisfaction of given specifications in bounded-time horizons. To do so, we propose a systematic approach to first decompose high-level specifications into simple reachability tasks by utilizing automata corresponding to the complement of specifications. We then construct control sub-barrier certificates and synthesize local controllers for those simpler tasks and combine them to obtain a hybrid controller that ensures satisfaction of the complex specification with some lower bound on the probability of satisfaction. To compute control sub-barrier certificates and corresponding local controllers, we provide two systematic approaches based on sum-of-squares (SOS) optimization program and counter-example guided inductive synthesis (CEGIS) framework. We finally apply our proposed techniques to two physical case studies.

Barrier Certificates for Assured Machine Teaching

Myopically Verifiable Probabilistic Certificates for Safe Control and Learning

Safe Control With Learned Certificates: A Survey of Neural Lyapunov, Barrier, and Contraction Methods for Robotics and Control

Data-Driven Permissible Safe Control with Barrier Certificates

An Optimal Control Approach to Sequential Machine Teaching

Machine Teaching of Active Sequential Learners

Understanding the Power and Limitations of Teaching with Imperfect Knowledge

Learning for Safety-Critical Control with Control Barrier Functions

From Small-Gain Theory to Compositional Construction of Barrier Certificates for Large-Scale Stochastic Systems

Verification of Uncertain POMDPs Using Barrier Certificates

Using Machine Teaching to Investigate Human Assumptions when Teaching Reinforcement Learners

Distribution Matching for Machine Teaching

Safe Barrier-Constrained Control of Uncertain Systems via Event-triggered Learning

Safety Barrier Certificates for Stochastic Hybrid Systems

Transfer of Safety Controllers Through Learning Deep Inverse Dynamics Model

Learning Control Barrier Functions with High Relative Degree for Safety-Critical Control

Safe Online Dynamics Learning with Initially Unknown Models and Infeasible Safety Certificates

Adaptive Teaching of Temporal Logic Formulas to Learners with Preferences

Synthesizing Barrier Certificates of Neural Network Controlled Continuous Systems Via Approximations.

On the Optimality, Stability, and Feasibility of Control Barrier Functions: An Adaptive Learning-Based Approach

Machine Teaching: A New Paradigm for Building Machine Learning Systems