30 July 2023
Proposes distributional RL method to handle uncertainty in marine navigation
Learns policies that adaptively balance speed and risk avoidance
Outperforms classical algorithms in challenging simulated environments
Converges to smooth, efficient trajectories that maintain safe distances
Open-sources code and simulation environments for research
Demystifying autonomous marine navigation with reinforcement learning
This paper proposes a reinforcement learning approach to enable autonomous navigation for unmanned surface vehicles in challenging marine environments with unknown currents and obstacles. The method learns to plan safe, efficient trajectories by interacting with simulated environments. Compared to classical algorithms, the learned policies are more robust to disturbances and avoid collisions more successfully.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper