The full text of this item is not available at this time because the student has placed this item under an embargo for a period of time. The Libraries are not authorized to provide a copy of this work during the embargo period, even for Texas A&M users with NetID.
Robot Adaptation in Human-Robot Collaborative Workspace Using Reinforcement Learning
Abstract
Collaborative robotics allows for automating complex manufacturing tasks in dynamic environments that require robust uncertainty handling by human operators. This has led to the need for developing efficient human-robot collaborative (HRC) systems. An important consideration for developing efficient HRCs is to account for pertinent human factors (HFs) such as cognitive fatigue, trust, mental workload and situation awareness. Cognitive fatigue is a common human factor in manufacturing applications that leads to poor task performance and increased task difficulty perceptions. In this thesis, a robot adaptation model under operator fatigue states is formulated and its implementation is described by employing a manufacturing use case of robotic surface polishing.
The problem of robot adaptation under operator cognitive fatigue is formulated using two models: 1) Hidden Markov Models (HMMs) with the objective of predicting and adapting to the pertinent human errors under cognitive fatigue, and 2) Markov Decision Processes (MDPs) with the objective of cognitive fatigue recovery and task performance optimization. The underlying MDP was solved using Q-learning, a reinforcement learning (RL) method. The implementation issues of Q-learning in this type of human-robot shared operation are discussed along with their implications on Q-learning convergence. A user study was conducted with eight participants to validate the performance of the adaptation model. The user study results showed improvements in fatigue recovery as analyzed from heart rate variability (HRV) features and fatigue perceptions, as well as on the task performance metrics such as accuracy and time efficiency. This indicates that such model formulation can be employed to develop effective robot adaptation strategies to account for fatigue-related vulnerabilities.
Citation
Shah, Jay Kalpeshbhai (2022). Robot Adaptation in Human-Robot Collaborative Workspace Using Reinforcement Learning. Master's thesis, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /198047.