Dynamic force sensing for high-speed robot manipulation using Kalman filtering techniques

An implementation was made of the theory of dynamic force sensing for high-speed robot manipulation which uses an optimal filtering technique in order to extract the external forces and moments on an end-effector from those measured by a wrist force sensor which are corrupted by the inertial forces and moments on the end-effector. The theory is summarized, and its implementation in its complete nonlinear style to provide for basic offline experiments is described. The experimental data show that the theory works well in extracting the external forces and moments. In a limited linearized style, the theory was applied to a real manipulation task: detecting an object on a planned trajectory by force information. The monitoring of the collision between the end-effector and the object using the force sensor, which usually requires the end-effector to move slowly to reduce the inertial forces and moments, was done successfully, even for fast motion of the end-effector.