Georgia Institute of Technology
Robotic Caregivers -- Simulation and Capacitive Servoing for Physical Human-Robot Interaction
Abstract
dc:description.abstractPhysical human-robot interaction and robotic assistance presents an opportunity to benefit the lives of many people, including the millions of older adults and people with physical disabilities, who have difficulty performing activities of daily living (ADLs) on their own. Robotic caregiving for activities of daily living could increase the independence of people with disabilities, improve quality of life, and help address global societal issues, such as aging populations, high healthcare costs, and shortages of healthcare workers. Yet, robotic assistance presents several challenges, including risks associated with physical human-robot interaction, difficulty sensing the human body, and complexities of modeling deformable materials (e.g. clothes). We address these challenges through techniques that span the intersection of machine learning, physics simulation, sensing, and physical human-robot interaction. Haptic Perspective-taking: We first demonstrate that by enabling a robot to predict how its future actions will physically affect a person (haptic perspective-taking), robots can provide safer assistance, especially within the context of robot-assisted dressing and manipulating deformable clothes. We train a recurrent model consisting of both a temporal estimator and predictor that allows a robot to predict the forces a garment is applying onto a person using haptic measurements from the robot's end effector. By combining this predictor with model predictive control (MPC), we observe emergent behaviors that result in the robot navigating a garment up a person's entire arm. Capacitive Sensing for Tracking Human Pose: Towards the goal of robots performing robust and intelligent physical interactions with people, it is crucial that robots are able to accurately sense the human body, follow trajectories around the body, and track human motion. We have introduced a capacitive servoing control scheme that allows a robot to sense and navigate around human limbs during close physical interactions. Capacitive servoing leverages temporal measurements from a capacitive sensor mounted on a robot's end effector to estimate the relative pose of a nearby human limb. Capacitive servoing then uses these human pose estimates within a feedback control loop in order to maneuver the robot's end effector around the surface of a human limb. Through studies with human participants, we have demonstrated that these sensors can enable a robot to track human motion in real time while providing assistance with dressing and bathing. We have also shown how these sensors can benefit a robot providing dressing assistance to real people with physical disabilities. Physics Simulation for Assistive Robotics: While robotic caregivers may present an opportunity to improve the quality of life for people who require daily assistance, conducting this type of research presents several challenges, including high costs, slow data collection, and risks of physical interaction between people and robots. We have recently introduced Assistive Gym, the first open source physics-based simulation framework for modeling physical human-robot interaction and robotic assistance. We demonstrate how physics simulation can open up entirely new research directions and opportunities within physical human-robot interaction. This includes training versatile assistive robots, developing control algorithms towards common sense reasoning, constructing baselines and benchmarks for robotic caregiving, and investigating generalization of physical human-robot interaction from a number of angles, including human motion, preferences, and variation in human body shape and impairments. Finally, we show how virtual reality (VR) can help bridge the reality gap by bringing real people into physics simulation to interact with and receive assistance from virtual robotic caregivers.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Electrical and Computer Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Erickson, Zackory
- Advisor dc:contributor.advisor
-
- Kemp, Charles C.
- Committee members dc:contributor.committeemember
-
- Chernova, Sonia
- Liu, Karen C.
- Turk, Greg
- Abbeel, Pieter
Subjects
dc:subject × 5Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- http://hdl.handle.net/1853/65054