In this paper, we exploit the generalization capabilities of Deep Reinforcement Learning (DRL) to enhance adaptive steering control of autonomous vehicles. With this aim, we investigate three different approaches to tackle the problem of lateral control of a vehicle: i) a pure DRL algorithm such as Twin Delayed Deep Deterministic Policy Gradient (TD3) or Soft Actor-Critic (SAC) as the sole controller, ii) a Proportional-Integral-Derivative (PID) controller combined with RL for error compensation, where the PID regularizes the RL search space to gain stability and iii) a PID controller utilizing RL as an adaptive and online selection mechanism for the PID gains. All three approaches must comply with hard runtime and memory consumption constraints as we deploy the RL actor on the Parallel Processing Unit (PPU) of Infineon's AURIX™ TC4x microcontroller. We also follow a model-based development approach, which eases the whole process. We show that combining DRL with a classical controller adds adaptivity to new complex roads and different vehicle dynamics for high speeds of more than 120 km/h, while using less resources on the PPU as detailed in the paper.
Hardware-Compatible Deep Reinforcement Learning-Based Lateral Trajectory Controller
24.09.2024
436204 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Crossroad lighthouse trajectory planning method based on deep reinforcement learning
Europäisches Patentamt | 2024
|Deep Reinforcement Learning-Based 3D Trajectory Planning for Cellular Connected UAV
DOAJ | 2024
|