Date of Award

12-2023

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical Engineering

Committee Chair/Advisor

Yue Wang

Committee Member

John Wagner

Committee Member

Ardalan Vahidi

Committee Member

Ian Walker

Abstract

Human-robot cooperative manipulation (co-manipulation) is one of the most prominent human-robot collaboration (HRC) tasks, where humans and robots manipulate the same object. Trust in HRC is crucial in determining human acceptance of robots and, hence, robot utilization. A probabilistic dynamic Bayesian network (DBN) trust model that integrates a time-series trust model is presented in this thesis. The trust model is learned using a continuous and normalized Baum-Welch (BW) algorithm, devised to account for the continuous nature of trust evolution and the limitations of the classic parameter learning method. To ensure a good HRC in co-manipulation, a variable impedance control framework is proposed based on trust and human intention estimation from force, with trust-based obstacle avoidance and trust-based robot-level task hierarchy. A case study on human-robot co-transportation was carried out, and a comprehensive statistical analysis was conducted. Our trust-aware variable impedance control framework has shown its superiority in terms of efficiency, agreement, safety, physical human-robot interaction (pHRI), and social HRI (sHRI) factors when compared to other benchmark behaviors. Another important factor during human-robot co-manipulation is safety, i.e., passivity and stability, especially during the activation or deactivation of the nullspace tasks. Although the nullspace tasks and their activation and deactivation allow the robot to multitask with priorities, they naturally bring nonpassivity/instability to the system. Such an HRC system can be formed as an impulsive multi-dimensional switched system, where state changes may occur at the switching moments. To analyze the passivity and stability of such systems, the classic passivity theory is first extended using the transition-dependent average dwell time (TDADT) and multiple Lyapunov functions (MLFs). Then, we extend this technique to analyze the human-robot co-manipulation system with activation and deactivation of the nullspace tasks. With the designed conditions and switching law, weak passivity and exponential integral-input-to-state practical stability (eiISpS) of the HRC switched system can be guaranteed.

Author ORCID Identifier

0009-0007-3245-0101

Available for download on Tuesday, December 31, 2024

Share

COinS