Adaptive impedance control

摘要

The objective of this research is to establish a method of controlling a robot that operates in an unstructured environment. This requires three fundamental objectives. First, the robot must be capable of constrained, as well as unconstrained manipulation. Second, the robot must be capable of learning not only where obstacles are in its workspace, but the characteristics of those obstacles as well. Finally, the robot must be capable of learning the characteristic of its environment without explicitly modifying its task definition. To achieve these goals, a new approach to robotic environment estimation is described. First, a discretized model of a robot's workspace provides for spatial and temporal variations in the environment. This operates in concert with a multiple input, multiple output recursive estimation algorithm. As the robot maneuvers about its workspace, measured tip force and position information provide the estimation routine states necessary for the identification of environment impedance. The updated parameters of the estimated environment impedance are stored in the robot's position dependent discretized workspace. As the robot learns the characteristics of its environment, the target impedance of the robot's controller adapts to this estimate based upon a specified performance criterion. The author describes five different methods to extract the estimated environment information and adapt the target impedance of the robot. Three techniques provide predictive behavior that enables adaptation prior to contact with the environment. This reduces the magnitude of contact forces generated during impact. In addition, two techniques provide robust initialization that ensure stability when operating in uncertain environments. Finally, the methods developed for robotic tasks are extended to a novel teleoperation platform. A new approach to force reflecting teleoperation of long reach flexible robots is described. The environment estimation method is employed for the estimation of a remote environment manipulated by a slave robot. The target impedance of the master robot, coupled to a human, adapts to variations in the remote environment. This ensures safe telemanipulation in hazardous or uncertain environments. Furthermore, the adaptive impedance controller is augmented with virtual fixtures to provide teleoperated obstacle avoidance.