Path Planning for Berthing Maneuvers of Space Serial Manipulators

Resumo

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2027.The path planning of serial manipulators in autonomous berthing is an important task that demands precise control sequences. This paper presents a comparative analysis of non-linear optimal control histories generated by machine learning (ML) strategies for the berthing maneuvers of space serial manipulators. A new framework for path planning is proposed, and the convergence characteristics of different ML control histories are evaluated. The methodology employs a non-linear optimal control formulation, followed by statistical evaluation of the output. Given a fixed time frame, a control history that drives the manipulator system states toward the target is obtained. Acceptable error margins are enforced. The study compares the convergence behavior in terms of end-effector positioning error and number of non-linear programming iterations in combination with Support Vector Machine, Random Forest, Extreme Gradient Boost and Gaussian Process strategies, under operational uncertainties. Results highlight variations in performance and provide insights for the selection of ML algorithms for space applications, ultimately contributing to the development of safer and more autonomous space robotic missions.

Palavras-chave