Adaptive physical human-robot interaction (PHRI) with a robotic nursing assistant.

Human-robot interaction;nursing assistant robot;base sensor control;neuroadaptive controller;sensor simulation;technology acceptance model

Sumit Kumar Das

University:University of Louisville

Department:Electrical and Computer Engineering

关键词: Human-robot interaction;    nursing assistant robot;    base sensor control;    neuroadaptive controller;    sensor simulation;    technology acceptance model;   

Others  :  https://ir.library.louisville.edu/cgi/viewcontent.cgi?article=4497&context=etd

美国|英语

来源: The Universite of Louisville's Institutional Repository

【 摘 要 】

Recently, more and more robots are being investigated for future applications in health-care. For instance, in nursing assistance, seamless Human-Robot Interaction (HRI) is very important for sharing workspaces and workloads between medical staff, patients, and robots. In this thesis we introduce a novel robot - the Adaptive Robot Nursing Assistant (ARNA) and its underlying components. ARNA has been designed specifically to assist nurses with day-to-day tasks such as walking patients, pick-and-place item retrieval, and routine patient health monitoring. An adaptive HRI in nursing applications creates a positive user experience, increase nurse productivity and task completion rates, as reported by experimentation with human subjects. ARNA has been designed to include interface devices such as tablets, force sensors, pressure-sensitive robot skins, LIDAR and RGBD camera. These interfaces are combined with adaptive controllers and estimators within a proposed framework that contains multiple innovations. A research study was conducted on methods of deploying an ideal HumanMachine Interface (HMI), in this case a tablet-based interface. Initial study points to the fact that a traded control level of autonomy is ideal for tele-operating ARNA by a patient. The proposed method of using the HMI devices makes the performance of a robot similar for both skilled and un-skilled workers. A neuro-adaptive controller (NAC), which contains several neural-networks to estimate and compensate for system non-linearities, was implemented on the ARNA robot. By linearizing the system, a cross-over usability condition is met through which humans find it more intuitive to learn to use the robot in any location of its workspace, A novel Base-Sensor Assisted Physical Interaction (BAPI) controller is introduced in this thesis, which utilizes a force-torque sensor at the base of the ARNA robot manipulator to detect full body collisions, and make interaction safer. Finally, a human-intent estimator (HIE) is proposed to estimate human intent while the robot and user are physically collaborating during certain tasks such as adaptive walking. A NAC with HIE module was validated on a PR2 robot through user studies. Its implementation on the ARNA robot platform can be easily accomplished as the controller is model-free and can learn robot dynamics online. A new framework, Directive Observer and Lead Assistant (DOLA), is proposed for ARNA which enables the user to interact with the robot in two modes: physically, by direct push-guiding, and remotely, through a tablet interface. In both cases, the human is being “observed” by the robot, then guided and/or advised during