Reactive Task Adaptation of a Dynamic System With External Disturbances Based on Invariance Control and Movement Primitives

This article proposes a general control framework for robot physical contact tasks. To satisfy the tradeoff between safety and performance in the interaction between robots and the environment, multipriority control is regarded as a preemptive strategy for constraint management when unknown disturbances exist. The equality and inequality constraints related to robot safety are enforced by invariance control, while reference profile tracking is accommodated by dynamic movement primitives without violating the higher priority constraints. With this framework, we complete the unification of force control and motion control in robot task execution. The robot thus acquires the capability for strong disturbance rejection and transferrable intelligence between similar tasks. At the same time, a variant linear-quadratic regulator (LQR) is integrated into the framework, which enables the robot to achieve exponential convergence of the tracking error. The proposed approach is tested and evaluated with two types of physical contact tasks, showing a superior control effect and faster convergence than the existing methods.