The current market cars are getting additional features like driver assistance systems which add comfort and safety to the commuters. This driver assistance system contains features like Cruise Control and Automatic braking system, which can be added for automated driving vehicles on highways in a limited scenario. In this work Deep Reinforcement Learning technique is proposed to develop an automated driving vehicle assistance system for highway driving. This helps vehicle to understand the surrounding environment and make decisions according to its observations on the highway. The Automated driving assistance system consists of supervisor agent (Machine Learning agent), which assists the system through Deep Reinforcement Learning algorithms. The supervisor agent is trained through images captured (camera installed in front of the automated driving vehicle) and Light Detecting and Ranging sensor (sensor installed on automated driving vehicle), which directly maps both camera image and Light Detecting and Ranging data to make decisions (acceleration, braking and lane change). Collision avoidance is included to prevent collisions on highways. The proposed method is verified by automated driving on highway simulation, Unity ML agents execute this simulation environment.
Distributional Reinforcement Learning For Automated Driving Vehicle
2022-10-16
1181265 byte
Conference paper
Electronic Resource
English
DataCite | 2021
|British Library Conference Proceedings | 2019
|