Active compliance force control based on an iterative learning strategy for multi-task manipulation of robots

For complex robot contact operation scenarios such as handling, assembly and grinding, how to achieve precise contact force control between the robot and the object or environment while ensuring the accuracy of the robot's working pose is still a challenging problem. This paper proposes a new unified contact force control strategy that combines the admittance control method and iterative learning strategy. This new strategy is used to integrate the manipulation of various primitives such as screwing, turning and pressing, which are usually included in the complex manipulation tasks of robots. Firstly, according to the characteristics of the multi-task operation, a fast-switching multi-function end-effector is designed, and the multi-task operation dataset which contains the contact pose and contact force information is collected through manual experiments, to construct a reference data-driven force control model. On this basis, a layered composite control method is designed for the learning and realization of the force control process. The bottom layer introduces the classical admittance control method to solve the actual force control. The upper layer performs the training on the control inputs through the iterative learning control strategy, so as to obtain high-precision and high-flexible control performance. Simulation and experimental studies are carried out for a variety of typical operation tasks. The results show that the method proposed in this paper can achieve a 96.63% improvement in force/torque tracking accuracy compared with the traditional admittance control method while maintaining the sub-millimeter positioning error, which validated the effectiveness, feasibility, and generalizability of this method.