Development of a paradigm for systematic evaluation of robot dynamic transparency

Robots designed for physically interacting with humans must be compliant and reactive to ensure safety and effective performance. In the field of rehabilitation robotics, dynamic transparency, which is the measure of the parasitic interaction force a robot imparts on its end user, is a key metric for the quality of assessment and therapy that can be accomplished with the robot. The improvement of dynamic transparency is an area of ongoing research. In this work, a novel paradigm for the systematic evaluation of dynamic transparency is developed, implemented, and tested. This new paradigm builds upon previous works by using a robot with multiple degrees of freedom to reliably and fully assess a rehabilitation robot's transparency. Optimized B-splines are used to evenly excite the full range of the rehabilitation robot's dynamics, and interaction measurements are used to calculate a transparency metric. A case study is presented that employs the paradigm to automate the control gain-tuning process and minimize the interaction forces imparted on an end-user without increasing vibrations. The method was able to discern differences in average interaction forces over a trajectory as low as 0.2 N caused by changes in controller gains and use those differences reduce the average interaction force by 35% from hand-tuned starting gains. Through the use of this method, the dynamic transparency for a robot can be systematically evaluated and tuned to improve the quality of physical human-robot interaction.