Sensorless Hybrid Normal-Force Controller With Surface Prediction

Abstract

In this paper, an improved sensorless hybrid controller for constant force control along normal direction is proposed for applications including polishing, milling and deburring. No additional sensors are involved. External joint torques of all joints can be calculated from their electric current, dynamic model and friction model. With the help of dynamically consistent generalized inverse matrix, they can be converted to external force/torque at the end-effector. The underlying force control strategy is the integration of impedance control model and explicit force control. The novel improvement is the real-time prediction algorithm of surface's shape profile and normal direction without any prior knowledge. So, this force controller has great adaptiveness to arbitrary unknown surfaces. Experiments were performed on a 7 degrees-of-freedom (DOFs) robot to test the controller's capability and utility on an inclined plane and curved surface. Results prove the credibility of external force estimation. The normal force tracking accuracy is adequate for targeted applications. Real-time prediction is functional as the robot adjusts its orientation accordingly.