Assume Guarantee

Assume-guarantee reasoning is a modular approach to verifying complex systems by decomposing them into smaller, independently verifiable components. Current research focuses on applying this paradigm to reinforcement learning, formal verification of strategic abilities (often using temporal logics), and controller synthesis for multi-agent systems, leveraging techniques like funnel-based control and contract-based specification refinement. This methodology offers significant advantages in tackling the scalability challenges inherent in verifying large-scale systems, with applications ranging from automated driving to robotic mission planning and improving the reliability of AI-controlled systems.