Robust Decision Making for Autonomous Driving using Deep Reinforcement Learning with Bayesian Particle Filter-based World Model

Making trustworthy decisions in noisy environments is a critical task for autonomous vehicles. In this paper, we propose a robust decision-making scheme for autonomous vehicles using Deep Reinforcement Learning (DRL) with a Bayesian Particle Filter-based world model (BPFW). The proposed method consists of two main components: the DRL-based decision-making module and the Bayesian Particle Filter-based world model. The DRL-based decision-making module is designed to learn the optimal policy for autonomous vehicles driving in complex environments. The BPFW is used to estimate the state of the environment and provide observations for the DRL-based decision-making module. We evaluated the proposed method in a simulated environment under various noise levels. The results show that our method can make trustworthy decisions for autonomous vehicles in complex environments, even when the observations are noisy. This method can be applied in real-world scenarios to improve the safety and reliability of vehicles.