Rare Collision Risk Estimation of Autonomous Vehicles with Multi-Agent Situation Awareness

This paper offers a formal framework for the rare collision risk estimation of autonomous vehicles (AVs) with multi-agent situation awareness, affected by different sources of noise in a complex dynamic environment. The estimation framework consists of two complementary parts: modeling formalism and a rare event estimation method using sequential Monte Carlo (MC) simulation instead of importance sampling. By defining incremental levels of severity that must be passed before a collision, a sequence of MC simulations can be applied from one level to the next. This particular sequential MC method consists of the simulation of an Interacting Particle System (IPS) in combination with Fixed Assignment Splitting (FAS) of particles that reach the next level. We model AVs equipped with the situation awareness as general stochastic hybrid systems (GSHS), including the IPS-FAS relevant severity levels, and assess the probability of collision in a lane-change scenario where two self-driving vehicles simultaneously intend to switch lanes into a shared one while utilizing the time-to-collision measure for decision-making as required. The IPS-FAS method is subsequently used to estimate collision risk for this GSHS model of the lane-changing scenario. The results show that in contrast to straightforward MC simulation, IPS-FAS is able to quantify the very low collision risk for the scenario of interest.