EAGER: TaskDCL: Personalized Robotic Assistance: Developing Safe, User-Taught Functionalities for Diverse Needs

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

In the era of advanced robotics and AI, there is a notable gap in leveraging these technologies for personalized human assistance, especially in settings that require nuanced understanding. Current robotic systems, despite their impressive locomotion capabilities, often lack the flexibility needed to adapt to the diverse needs of users, as their interactions are typically restricted to predefined functionalities. This limitation is particularly evident in scenarios where humans teach robots to perform highly personalized tasks within shared physical spaces, which may lead to discomfort and safety concerns among users. This EArly-concept Grant for Exploratory Research (EAGER) project will fund research that attempts to address the challenging task of a quadrotor to pick up/drop off a small package from/onto a human's outstretched hand following the human's instructions. The challenge of the task comes from the quadrotor's close proximity to a human, which can trigger stress due to noise and movement, potentially undermining task completion. The research effort seeks to address the above-mentioned task utilizing a framework that enables anyone to safely instruct robots in customized tasks. This research is crucial for the exploration of harnessing the intelligence enabled by machine learning to expand the capabilities of robots toward humans’ needs, especially in the industries that urge rapidly growing robot participation and coordination with humans, e.g., manufacturing, logistics, transportation, and national defense. This research aims to attain the objective of allowing quadrotors to safely interact with people for the chosen task of direct hand-to-hand package delivery, ultimately leading to robots that can genuinely adapt to and meet individual needs for more personalized and safe human-robot interactions. The researched effort incorporates the human's cognitive state into the quadrotor's decision-making and action processes, fostering a bidirectional sensorimotor interaction and allowing both the human and the quadrotor to sense and influence each other's decisions and actions while the quadrotor conducts the task in close proximity to the human. Two interconnected research thrusts will be pursued: (i) planning and control for safe human-robot interaction with models for human cognitive states and (ii) iterative learning for enhanced performance towards efficient and safe human-robot interactions. The research framework builds upon the advances made by the PIs in control theory/engineering and psychology and is expected to make important contributions to the society of the future, in which humans and robots behave and interact safely and effectively while occupying shared spaces. This EAGER award has been co-funded by the Dynamics, Controls, and System Diagnostics and the Mind, Machine, and Motor Nexus Programs. 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.