PerceMon: Online Monitoring for Perception Systems

Perception algorithms in autonomous vehicles are vital for the vehicle to understand semantics of its surroundings, including detection and tracking objects environment. The outputs these turn used decision-making safety-critical scenarios like collision avoidance, automated emergency braking. Thus, it is crucial monitor such perception systems at runtime. However, due high-level, complex representations systems, a challenge test verify especially In this paper, we present runtime monitoring tool, PerceMon that can arbitrary specifications Timed Quality Temporal Logic (TQTL) extensions with spatial operators. We integrate tool CARLA simulation environment ROS middleware platform while properties on state-of-the-art object algorithms.