Georgia Institute of Technology
Investigating perceived comfort dynamics and trajectory-level motion preferences in first-time human-robot interaction
Abstract
dc:description.abstractIn close-proximity human-robot interaction (CP-HRI), it is crucial for the robot to adapt its motion characteristics to user behavior and preferences to ensure user safety and comfort. Prior work has addressed this problem by jointly optimizing different aspects of a trajectory through multi-objective optimization. However, balancing these aspects of interaction is challenging and requires a deep understanding of user's subjective feelings and behavior preferences. In this work, we focus on first-time robot users. We investigate their perceived comfort dynamics over repeated interactions with a robot manipulator and their trajectory-level motion preferences using a range of objectives from prior works characterizing the safety, comfort and efficiency of the interaction. To take the first step towards understanding the perceived comfort of first-time robot users, we carry out a user study to investigate the connections between robot motion and users' perceived comfort dynamics. Specifically, we study the influence of workspace overlap, robot end-effector speed, and robot motion legibility on overall comfort and its evolution over repeated interactions. From statistical analysis, we identified that workspace overlap is the most influential motion characteristics compared to legibility and robot end-effector speed. We found that lower workspace overlap leads to higher perceived comfort with less fluctuations. We believe our findings can guide the design of robotic systems and their motion planning to maximize user comfort for the novice users. Based on the skeletal data collected from the user study, we investigate how users tend to prioritize and balance several interaction objectives from robot motion planning including: legibility, joint distances, visibility, smoothness, goal pose distance and deviation from nominal trajectory. We extract a weighting over these objectives using an inverse reinforcement learning algorithm with a receding horizon planning formulation. Our inferred weight vector suggests that two most prominent objectives optimized by users are reaching the goal position and maintaining legible motion. This insight can be utilized in robot motion planning strategies to perform human-like motions that can interact more naturally with users, which is important for social and service robots in healthcare or education.
Degree
thesis:*- Level thesis:degree_level
- Masters
- Department dc:contributor.department
- Computer Science
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Liu, Yifan
- Advisor dc:contributor.advisor
-
- Ravichandar, Harish
- Committee members dc:contributor.committeemember
-
- Chernova, Sonia
- Pradalier, Cedric
Subjects
dc:subject × 2Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/72064