Optimizing the path of autonomous driving cars using PID controller and Deep reinforcement learning

Details

Supervisors

Macq, Benoît

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

Autonomous vehicles are gaining increasing prominence in enhancing the transportation sector. This thesis aims to analyze, evaluate, and understand the mechanisms behind control methods as well as reinforcement learning-based approaches to bring value to the transportation industry. Through a comprehensive review of existing literature and practical implementation of these methods, the research seeks to identify the strengths and limitations of current technologies. The study will focus particularly on shaping the reward signal as well as a comparison of algorithms to enable the agent to learn optimal road behavior. Experiments will be conducted using the Donkey Car simulator, set in an environment without other vehicles, with an emphasis on optimizing the vehicle's navigation to its destination. While reinforcement learning has shown promising results in various fields, traditional control methods offer a more predictable and reliable approach to autonomous driving. However, reinforcement learning offers greater adaptability to complex and dynamic environments. This research aims to provide a comprehensive understanding of the potential of reinforcement learning in autonomous driving and to identify the strengths of such approaches compared to traditional control methods.