The article presents a research on the use of deep reinforcement learning (DRL) for the control and positioning of scaled robotic vehicles. The authors propose a federated extended Kalman filter (FEKF) and a DRL path tracking controller trained via an expert demonstrator to increase robustness and speed up the learning phase. The effectiveness of the proposed model and control strategies is validated through experimental results. The research shows that the proposed DRL path tracking strategy outperforms traditional model-based control strategies.
Publication date: 2024-01-11
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.05194
