Obstacle Avoidance System on Autonomous Car Using D3QN

An autonomous vehicle’s triumphant and safe navigation, which can circumvent obstacles, necessitates a skill set encompassing steering wheel control, sometimes called obstacle avoidance. One potential approach to address this issue is using a simulation framework wherein an automobile is subjected to various barriers. During this simulation, the sensory input of the car, as well as its corresponding actions, are recorded and analyzed. An alternative approach involves allowing the vehicle to autonomously acquire knowledge to optimize its performance towards the desired objective. The Dueling Deep Double QNetworks (D3QN) approach is a strategy that enables the model to autonomously learn and optimize its performance to attain the most favorable conclusion. The D3QN architecture is a computational framework incorporating Dueling and double-Q processes. The implementation of D3QN is anticipated to result in a reduction in the training time required for an autonomous vehicle. This study is expected to substitute for training an autonomous vehicle.