A bio-inspired approach for regulating and measuring visco-elastic properties of a robot arm

This work focuses on interaction control of robot manipulators in unstructured environments, with special regard for situations of unpredictable contact/noncontact transitions. It is basically addressed to those environments where a high level of robot adaptability is required and no information on the geometry of the environment is available. By pointing out the main limitations of standard interaction control schemes in managing situations of contact/noncontact transitions, this paper proposes a new control solution that is inspired by the biological model of motor control in voluntary movements. It consists of a combination of a feedforward loop and a proportional-derivative plus gravity compensation control in the feedback loop. The control law is named coactivation-based compliance control in the joint space since a unique function, called coactivation function, is evaluated for regulating robot visco-elasticity in an unpredictably variable environment. It resumes the mechanism of adjustable visco-elastic properties acting on the agonist and antagonist muscles of a human arm. The work also proposes a methodology for evaluating performance of interaction control schemes that is based on stiffness graphical representation through ellipses. The method replicates the experimental setup used in neuroscience to measure stiffness in human limbs. It is regarded as a powerful tool for evaluating robot behavior over space and time, since it allows both a visual representation of stiffness variation during motion and a quantitative measure of robot performance. It is shown how the method can be used to evaluate a control scheme and how it can provide indications to improve a control law. In this paper, an application to the standard compliance control in the joint space and the coactivation-based compliance control is presented. © 2005 Wiley Periodicals, Inc.