Bootstrapping Reinforcement Learning with Sub-optimal Policies for Autonomous Driving
This work addresses training challenges for RL agents in autonomous driving, but it is incremental as it builds on existing methods with demonstration guidance.
The paper tackles the problem of sample inefficiency and poor exploration in reinforcement learning for autonomous driving by integrating a sub-optimal rule-based lane change controller with the Soft Actor Critic algorithm, resulting in improved driving performance.
Automated vehicle control using reinforcement learning (RL) has attracted significant attention due to its potential to learn driving policies through environment interaction. However, RL agents often face training challenges in sample efficiency and effective exploration, making it difficult to discover an optimal driving strategy. To address these issues, we propose guiding the RL driving agent with a demonstration policy that need not be a highly optimized or expert-level controller. Specifically, we integrate a rule-based lane change controller with the Soft Actor Critic (SAC) algorithm to enhance exploration and learning efficiency. Our approach demonstrates improved driving performance and can be extended to other driving scenarios that can similarly benefit from demonstration-based guidance.