Trajectory control of a wheeled robot using interaction forces for intuitive overground human-robot interaction

"Effective and intuitive physical human robot interaction (pHRI) requires an understanding of how humans communicate movement intentions with one another. It has been suggested that humans can guide another human by hand through complex tasks using force information only. However, no clear and applicable paradigm has been set forth to understand these relationships. While the human partner can readily understand and adhere to this expectation, it would be difficult for anyone to explain their intuitive motions with strict rules, algorithms, or steps. Uncovering such a procedural framework for the control of robotic systems to execute expected performance simply from force and torque inputs would be the necessary first step in realizing human-like pHRI. As such, this research is motivated to develop a procedural framework that uses human forces and torques, and this information alone, to guide a robot, with the goal that the robot's performance would be comparable with that of a human. To accomplish this goal, a process is presented for finding the signatures in human-human interaction force data that infers the appropriate robot velocities using neural network training. The results uncovers a number of signatures of physical interaction between humans. The importance of all available force inputs, in this case six degrees of freedom, is made clear through qualitative and quantitative data analysis. The relationship between the interaction forces and the robot velocities are shown to be specific to a particular participant and cannot be generalized to others. The effect of time delays present in human biomechanics is shown to manifest in the force-velocity mapping, such that increased mapping performance is observed for 50-100 ms delays. All of the results are demonstrated in a software robot simulation confirming the research s impact. Our work presents how human-generated interaction forces can be interpreted by a mobile robot for effective pHRI in the future"--Abstract, page iii.

Name thesis:degree_name

Ph. D. in Mechanical Engineering

Grantor

Missouri University of Science and Technology

Author dc:creator

• Holmes, George Leno, Jr.

• Human data neural networks
• Human like interaction
• Human robots
• Physical Human Robot Interaction
• Mechanical Engineering
• Robotics

Repository record dc:identifier

https://scholarsmine.mst.edu/doctoral_dissertations/2887

OAI identifier oai:identifier

oai:scholarsmine.mst.edu:doctoral_dissertations-3892