A Brain-Inspired Approach for Compliant Force Interaction Control under Predefined Tasks

Force-controlled robots usually rely on preset models for interaction, and it is difficult to achieve compliant interaction without accurate environment models. In contrast, human brain can interact deftly and compliantly without a model. Therefore, this paper proposes a bionic compliant control approach based on the motor cortex-cerebellum to simulate the joint function of the human brain to control limbs for compliant interaction and improve the force compliant interaction effect of the robot when it has predefined tasks. The approach includes a motor cortex compliant trajectory planner constructed with an echo state network (ESN) and a cerebellum joint controller based on a spiking neural network (SNN). We applied and verified this approach to the simulated interactive control of a 7-DoF robotic manipulator. The results show that this approach can adaptively adjust the trajectory when performing predefined tasks, reduce the interaction force, and improve the naturalness and safety of the interaction.