Safe Reinforcement Learning for Autonomous Vehicles through Parallel Constrained Policy Optimization
This work addresses safety concerns for autonomous vehicles, but it appears incremental as it builds on existing actor-critic architectures with added constraints and parallelization.
The paper tackles the problem of applying reinforcement learning to autonomous vehicles by addressing safety and explainability issues, resulting in a new algorithm called Parallel Constrained Policy Optimization (PCPO) that ensures safety and achieves fast learning in simulations.
Reinforcement learning (RL) is attracting increasing interests in autonomous driving due to its potential to solve complex classification and control problems. However, existing RL algorithms are rarely applied to real vehicles for two predominant problems: behaviours are unexplainable, and they cannot guarantee safety under new scenarios. This paper presents a safe RL algorithm, called Parallel Constrained Policy Optimization (PCPO), for two autonomous driving tasks. PCPO extends today's common actor-critic architecture to a three-component learning framework, in which three neural networks are used to approximate the policy function, value function and a newly added risk function, respectively. Meanwhile, a trust region constraint is added to allow large update steps without breaking the monotonic improvement condition. To ensure the feasibility of safety constrained problems, synchronized parallel learners are employed to explore different state spaces, which accelerates learning and policy-update. The simulations of two scenarios for autonomous vehicles confirm we can ensure safety while achieving fast learning.