Shield Synthesis

Shield synthesis focuses on developing methods to ensure the safe operation of autonomous systems, particularly in reinforcement learning and control systems, by incorporating safety constraints into their behavior. Current research emphasizes efficient algorithms for generating "shields"—external components that monitor and correct unsafe actions—often employing techniques like state-space transformations, reason-based approaches, and hybrid models combining Kriging and random forests to improve computational efficiency and accuracy. This work is crucial for deploying autonomous systems in safety-critical applications, ranging from robotics and autonomous driving to infrastructure projects like tunneling, by providing verifiable safety guarantees without significantly compromising performance.