Optimal path planning for autonomous berthing of unmanned ships in complex port environments.

This paper addresses the problem of automatic berthing of unmanned surface vehicles (USVs) in complex scenarios and proposes a multitask optimal control berthing method (MT-OCBM). The berthing process is divided into stages, each with distinct goal constraints and path restrictions. Through a combination of topology and computational geometry, a sparse Delaunay triangulation map segmentation (SDTMS) is introduced. SDTMS reduces map complexity while preserving accuracy by extracting map feature points and dividing the map into multiple triangular regions. The application of the traditional hp-Radau pseudospectral method (hp-RPM) to solve optimal control problems (OCPs) in complex marine environments is restricted by problems such as slow convergence and susceptibility that arise because of greatly different magnitudes of path constraints. To overcome these problems, we propose the hp-normalized Radau pseudospectral method (hp-NRPM). Normalizing the path constraints equalizes the magnitudes of path constraints, thereby significantly reducing the computational complexity of solving the OCPs. In contrast to traditional approaches that separate control from path planning, the optimal control algorithm employed in this paper integrates the control variables of unmanned vessels into the planning process of berthing paths. This integration allows the planned berthing paths to reflect the motion characteristics and constraints experienced by the unmanned vessel. It uses the KCS ship model to construct the Maneuvering Modeling Group ship motion model, and using the power spectrum given in the NORSOK standard, it establishes a slow, time-varying wind field model. The hp-NRPM algorithm is used to adjust the control inputs within the constraint region, thereby deriving a berthing route that optimizes the performance function. For validating the method, Qingdao Port was used as a simulation environment, and multiple sets of experiments were performed. The experimental results confirmed that MT-OCBM is robust and effective against disturbances and various obstacles encountered during berthing. • Pseudospectral method can effectively solve the optimal control berthing problem. • Using Delaunay triangulation for map segmentation can mitigate problem complexity. • Multitasking method allows algorithm focusing on different goals at various stages. • Normalizing path constraints can expedite the convergence speed of berthing problem. [ABSTRACT FROM AUTHOR]