Autonomous Berthing and Unberthing of Unmanned Surface Vessels via Harmonised Bézier Path Planning and Proximal Policy Optimisation

ABSTRACT

Berthing and unberthing are among the most demanding ship manoeuvring tasks because they are performed at low speed in restricted waters and are highly susceptible to environmental disturbances. Existing methods generally rely on relatively accurate mathematical models and may exhibit degraded performance when model uncertainty and external disturbances become significant. To address these limitations, this study proposes an autonomous berthing and unberthing framework, termed harmonised Bézier trajectory planning with proximal policy optimisation (HB‐PPO). In this paper, a harmonised segmented Bézier curve is developed to generate smooth berthing trajectories with continuous curvature, thereby alleviating the practical problems of abrupt steering demand, control oscillation and degraded manoeuvring stability caused by the curvature discontinuities commonly observed at the junctions of conventional piecewise Bézier curves. Furthermore, a proximal policy optimisation–based continuous control policy is established to directly output rudder angle and propulsive thrust, enabling coordinated regulation of heading, path‐following and speed. Then, multiple berthing and unberthing simulation scenarios are constructed to evaluate the HB‐PPO under different port layouts and ship scales against representative control approaches. The results show that the HB‐PPO achieves superior tracking accuracy while markedly reducing rudder oscillation, resulting in smoother and more realistic control actions. In addition, the HB‐PPO demonstrates good adaptability across different berthing and unberthing scenarios and ship scales, while maintaining satisfactory robustness under environmental disturbances. These findings indicate that the HB‐PPO provides a promising solution for autonomous berthing and unberthing of unmanned surface vessels in port waters.