Surrogate based trajectory planning method for an unmanned electric shovel

Abstract

Trajectory planning is a crucial precondition for mechatronic systems to achieve precise and energy-saving working activities. However, conventional trajectory planning methods are restrained by curve types, which may not obtain the optimal trajectory adequately and explicitly. To address these challenges, a surrogate based trajectory planning (STP) method is developed in this study. Firstly, the time dependent variables of the optimal control problem are discretized using appropriate design of experiment (DoE) method. All the obtained training points for establishing the trajectory/speed surrogate model serve as the design variables. Thus, updating of the training points set is equal to updating the full-time trajectory/speed curve at every optimization iteration. By setting indispensable system dynamic constraints, the optimizer can gradually find the optimal trajectory. The proposed STP method is illustrated using an unmanned electric shovel (ES) example, which builds on the excavating trajectory planning problem by incorporating the excavating efficiency and energy consumption as the objectives. Numerical results show that the STP method can achieve a better excavating performance than other conventional trajectory planning methods. Besides, the high flexibility of STP is also demonstrated across different working conditions, including variable pile angles and complex pile surfaces.