Date on Master's Thesis/Doctoral Dissertation

12-2021

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Electrical and Computer Engineering

Degree Program

Electrical Engineering, PhD

Committee Chair

Popa, Dan

Committee Member

Naber, John

Committee Member

Welch, Karla

Committee Member

Roussel, Thomas

Author's Keywords

Robotics; physical Human-Robot Interaction; kinematic sensing; technology acceptance model; nursing assistant

Abstract

The use of mobile manipulators in service industries as both agents in physical Human Robot Interaction (pHRI) and for social interactions has been on the increase in recent times due to necessities like compensating for workforce shortages and enabling safer and more efficient operations amongst other reasons. Collaborative robots, or co-bots, are robots that are developed for use with human interaction through direct contact or close proximity in a shared space with the human users. The work presented in this dissertation focuses on the design, implementation and analysis of components for the next-generation collaborative human machine interfaces (CHMI) needed for mobile manipulator co-bots that can be used in various service industries. The particular components of these CHMI's that are considered in this dissertation include: Robot Control: A Neuroadaptive Controller (NAC)-based admittance control strategy for pHRI applications with a co-bot. Robot state estimation: A novel methodology and placement strategy for using arrays of IMUs that can be embedded in robot skin for pose estimation in complex robot mechanisms. User perception of co-bot CHMI's: Evaluation of human perceptions of usefulness and ease of use of a mobile manipulator co-bot in a nursing assistant application scenario. To facilitate advanced control for the Adaptive Robotic Nursing Assistant (ARNA) mobile manipulator co-bot that was designed and developed in our lab, we describe and evaluate an admittance control strategy that features a Neuroadaptive Controller (NAC). The NAC has been specifically formulated for pHRI applications such as patient walking. The controller continuously tunes weights of a neural network to cancel robot non-linearities, including drive train backlash, kinematic or dynamic coupling, variable patient pushing effort, or slope surfaces with unknown inclines. The advantage of our control strategy consists of Lyapunov stability guarantees during interaction, less need for parameter tuning and better performance across a variety of users and operating conditions. We conduct simulations and experiments with 10 users to confirm that the NAC outperforms a classic Proportional-Derivative (PD) joint controller in terms of resulting interaction jerk, user effort, and trajectory tracking error during patient walking. To tackle complex mechanisms of these next-gen robots wherein the use of encoder or other classic pose measuring device is not feasible, we present a study effects of design parameters on methods that use data from Inertial Measurement Units (IMU) in robot skins to provide robot state estimates. These parameters include number of sensors, their placement on the robot, as well as noise properties on the quality of robot pose estimation and its signal-to-noise Ratio (SNR). The results from that study facilitate the creation of robot skin, and in order to enable their use in complex robots, we propose a novel pose estimation method, the Generalized Common Mode Rejection (GCMR) algorithm, for estimation of joint angles in robot chains containing composite joints. The placement study and GCMR are demonstrated using both Gazebo simulation and experiments with a 3-DoF robotic arm containing 2 non-zero link lengths, 1 revolute joint and a 2-DoF composite joint. In addition to yielding insights on the predicted usage of co-bots, the design of control and sensing mechanisms in their CHMI benefits from evaluating the perception of the eventual users of these robots. With co-bots being only increasingly developed and used, there is a need for studies into these user perceptions using existing models that have been used in predicting usage of comparable technology. To this end, we use the Technology Acceptance Model (TAM) to evaluate the CHMI of the ARNA robot in a scenario via analysis of quantitative and questionnaire data collected during experiments with eventual uses. The results from the works conducted in this dissertation demonstrate insightful contributions to the realization of control and sensing systems that are part of CHMI's for next generation co-bots.

Share

COinS