Force-Based Control of UGVs by Using Physical Human-Robot Interaction

Abstract

Humans apply directional forces to move wheeled objects in various everyday applications such as maneuvering a loaded cart, a lawnmower, or moving a stretcher. The heavier the object, the higher the magnitude of force required by the human to move it. This thesis proposes to use similar interactions on mobile robots to move bigger and heavier vehicles at the command of a human operator that physically pushes the vehicle. At the same time, the objective is to have the human operator perceive the vehicle as being much lighter to extend his/her ability when working with heavy loads.

A Husky A200 mobile robot, a differential drive robot by Clearpath Robotics, and Ibex, a custom-built land robot with four individually articulated tracks, were used to test the control system for such an interaction. A cantilever beam equipped with strain gauges and four FSRs were used to measure the forces applied by a human operator. For a natural motion, an admittance control model was used to generate velocity commands for the vehicle by using the measured forces as inputs. The preliminary design used a fixed admittance model. After running experiments and analyzing the results, an adaptive admittance model was developed using Probabilistic Neural Networks to adaptively change the admittance parameters in response to a change in the operating surface environment.