A machine learning based energy efficient trajectory planning approach for industrial robots

Abstract

Towards an energy efficient trajectory planning of industrial robot (IR), this paper proposes a machine learning based approach. Within the context, the IR’s movements are digitalised in joint space first, which allows using data attributes to represent IR’s trajectories. Moreover, a set of designed trajectories which can address IRs workspace are followed by the IR, and meanwhile, the energy consumption is measured. Then data sets are generated by combining the trajectory data and measured energy consumption data, and they are used to train a machine learning model. On top of that, the trained model provides a fitness function to evolution based or swarm-intelligence based algorithms to obtain a near-optimal or optimal trajectory. Finally, a simplified case study is demonstrated to validate the proposed method. The method provides a direct connection between joint control and energy efficiency objective, by which the solution space can be obviously relaxed, compared to the existing methods.