Efficient Path Planning with G2-Smoothing for USV in the Presence of Known Obstacles

Unmanned surface vehicles (USVs), have enabled many practical applications due to their maneuverability. Yet achieving efficient and reliable performance for USVs is a challenging problem. It is known that path planning plays a crucial role in solving the problem. By limiting the curvature and improving the degree of smoothness of the path, USVs can reduce tracking error and maintain distance from obstacles. Dubins paths are commonly used in USV path planning as they can be used to construct shortest curvature-bounded paths. However, their lack of smoothness in curvature leads to the physical impossibility for USV to follow them accurately, even in ideal conditions. This paper proposes a method of constructing a path through a set of waypoints with G2-smoothness in an environment with known maps and obstacles. The approach involves searching for an efficient curvature-bounded path via the genetic algorithm, then checking for obstacles within the safe distance limit and updating the path to avoid collision before smoothing it using Fermat spirals. A simulation was built to validate and compare the method’s performance with the Dubins interpolation’s performance. The results showed that the proposed method has reduced both RMS and maximal cross-track error, along with travel length.