CAREER: Human/Machine Collaborative Learning and Control of Contact-Rich Dynamics

点击翻译按钮获取中文摘要

This Faculty Early Career Development (CAREER) grant seeks to discover fundamental engineering principles that enable humans and machines to collaboratively learn and control the complex dynamics that arise when the human or machine intermittently contacts objects or uneven terrain in the environment. Although humans successfully learn to control complex machine dynamics when they fly planes, drive cars, or perform robot-assisted surgery, these applications avoid the abrupt changes in dynamics encountered when one's feet hit the ground, or when one's hands initially grasp objects. Despite the prevalence of contact-rich dynamic interactions in daily life and work, state-of-the-art robots struggle with making and breaking contact whether or not there is a human in-the-loop. In many applications of current and future interest -- remotely-operated robots, active prosthetics and exoskeletons, and brain/machine interfaces, to name a few -- humans and machines will collaboratively learn to control legs and arms as they make and break contact with the environment. The project will advance the NSF mission to promote the progress of science and to advance national health, prosperity, and welfare by advancing understanding of how humans learn to control contact-rich dynamics and how machines can adapt to ensure safety and improve performance of the human/machine system. In the long term, the results of this project will human/robot teams to perform complex tasks involving dynamic interaction with the world. Since many jobs are becoming increasingly automated, and since many people will experience impaired movement at some point in their lives, the goals of this research have tangible benefits for work and/or health of many people. To ensure these benefits are shared equitably, this project includes three evidence-based education and outreach programs designed to broaden participation in STEM fields.The research goal of this project is to test the hypotheses that humans can learn dynamic models of complex contact-rich machine dynamics, that they can and invert those models to control machine interactions with the environment, and that machines can exploit these facts to adapt their behavior to assist the human to perform desirable tasks. This goal will be pursued through three objectives. The first is to mathematically derive and computationally approximate inverse models for contact-rich dynamics. To achieve this objective, the PI will derive conditions that ensure that a forward model involving contact dynamics is invertible (exactly or approximately) and create algorithms that compute representations for the (approximate) inverse model. The second objective is to experimentally test whether humans control contact-rich dynamics as if they learned and inverted forward models, and to discern how sensorimotor pathways are integrated to implement the human’s controller. To do so, the project team will conduct human subject experiments using a novel teleoperation testbed to study trajectory tracking tasks involving intermittent contact events. The third objective will mathematically derive and experimentally test performance of human/machine co-adaptation algorithms based on game theory. The education goal of this project is to broaden participation in STEM fields by engaging high school and college students from underrepresented groups in hands-on research and design experiences focused on human/machine interaction.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.