Admittance Control With Unknown Location of Interaction

Admittance control is commonly used for controlling robots that physically interact with the environment or objects. Its goal is to realize a desired admittance property between the interaction wrench and robot motion. The original admittance control requires the interaction wrench and motion be converted to/from the space in which the desired admittance property is defined. In this letter, we first present the end-to-end admittance control formulation that represents the direct relationship between the generalized velocity and interaction wrench measurement when they satisfy the desired admittance property. Based on this formulation, we then develop a sampling-based algorithm for directly computing the robot reference motion without using the actual location of the interaction or the additional inverse kinematics step that is typically required for task-space admittance control. The algorithm is demonstrated in a number of simulation examples including collisions with a moving object and fixed obstacle.