Mathematical Models for Human-Robot Systems in Assistive Robotics: Perception, Inference, and Assistance

Jain, Siddarth

doi:https://doi.org/10.21985/n2-w82r-1d87

Work

Mathematical Models for Human-Robot Systems in Assistive Robotics: Perception, Inference, and Assistance

Public

Download PDF

Download All Files (.zip)

Assistive robotics focuses on human-robot systems that provide physical support and assistance to the elderly and people with motor-impairments. While assistive machines, such as the powered wheelchair, can significantly enhance the functional independence of individuals, many users are challenged by their direct operation, the manner in which such systems are currently operated by the users. Moreover, as assistive machines become more capable, they often become more complex to control. This means paradoxically the more severe a person’s motor-impairment, the more challenging it is for them to operate the very assistive machines meant to aid them. An enduring goal is to address this discrepancy by incorporating robotics autonomy and intelligence into assistive machines to help reduce the control burden on the user. Such human-robot systems emphasize proximate interaction, forming personal and collaborative relationships, and sharing control with human personnel. Thus, the design, sensing, control, and assessment of such systems becomes more sophisticated due to having a human-in-the-loop. To succeed in their role of providing assistance in shared autonomy, the robotics autonomy must be capable of: autonomously perceiving the potential goals of the human user, inferring their intentions, and sharing control with the user for providing assistance in a manner that is acceptable to the human and at the same time is efficient for task executions. These behaviors engage a number of disciplines including computer vision, machine learning, autonomous control, human psychology, and cognitive science. In this thesis, we focus on building mathematical models and algorithms for autonomous perception, inference, and assistance in human-robot systems for assistive robotics. Specifically, we investigate the computational perception of navigation goals involving wheelchair docking at table and desk structures and manipulation goals for assistive robotic arms involving the detection of semantic grasp types on novel household objects. We investigate human intent recognition in shared autonomy by Bayesian filtering and the modeling of human actions in a probabilistic behavior model. We also introduce an intent-driven optimization that adapts the model to each individual user. For assistance personalization and adaptation, we investigate a mathematical framework for human-in-the-loop optimization of shared autonomy. Furthermore, we also investigate and present a novel application of body-machine interface in human-robot systems, which engage users in sustained physical activities with the aim to support partial recovery of movement skills. We validate all contributed algorithms and techniques in this thesis on real-hardware; using either a wheelchair robot or a robotic arm platform. We conduct human subject experiments in a variety of shared-autonomy settings and report our findings. This thesis contributes to and across various aforementioned disciplines, providing a greater understanding of the computational and human requirements for successful human-robot systems in the assistive domain.

Last modified