Path Planning for Autonomous Inland Vessels in Complex Harbor Environments

In this paper, we present a flexible path generation framework for complex maneuvers of autonomous inland vessels in harbor environments. Our approach is based on the common motion primitives approach. By sampling our motion primitives from the motion model of an inland ship, we are able to generate realistic path solutions for autonomous vessels. The underlying graph search approach considers 5 dimensional state space, to account for the characteristics of the ship motion, with huge inertia and latency between the control command and response of the vessel. The framework can be easily modified to include other dynamic properties of the vessel or to perform path planning with other vehicles with different characteristics. We present experimental results of simulated maneuvers of an autonomous inland cargo vessel in a narrow harbor basin. The new planning framework provides real-time performance and ensures feasible paths with simplified control logic. In our evaluation, we examine the individual steps of the solution path and explain the quality of the result in comparison to an expertly executed solution.