Joining High-Level Symbolic Planning with Low-Level Motion Primitives in Adaptive HRI: Application to Dressing Assistance

Abstract

For a safe and successful daily living assistance, far from the highly controlled environment of a factory, robots should be able to adapt to ever-changing situations. Programming such a robot is a tedious process that requires expert knowledge. An alternative is to rely on a high-level planner, but the generic symbolic representations used are not well suited to particular robot executions. Contrarily, motion primitives encode robot motions in a way that can be easily adapted to different situations. This paper presents a combined framework that exploits the advantages of both approaches. The number of required symbolic states is reduced, as motion primitives provide "smart actions" that take the current state and cope online with variations. Symbolic actions can include interactions (e.g., ask and inform) that are difficult to demonstrate. We show that the proposed framework can adapt to the user preferences (in terms of robot speed and robot verbosity), can readjust the trajectories based on the user movements, and can handle unforeseen situations. Experiments are performed in a shoedressing scenario. This scenario is particularly interesting because it involves a sufficient number of actions, and the humanrobot interaction requires the handling of user preferences and unexpected reactions.

Personalized collaborative plans for robot-assisted dressing via optimization and simulation

“…Pignat and Calinon (2017) and Canal et al (2018) use learning from demonstration (LfD) to have a human teach the robot how to perform dressing tasks. Canal et al (2018) combines motion primitives with higher-level symbolic task planning to represent dressing tasks and demonstrate their method for putting a shoe onto a person's foot. A potential advantage of LfD is that individuals can directly teach the robot, which gives them the opportunity to communicate their preferences.…”

Learning to Collaborate From Simulation for Robot-Assisted Dressing

“…Work by Gao et al and Zhang et al used vision to construct a model of a user's range of motion and a dynamic trajectory plan to dress that user in a vest [7]- [9]. The I-dress project has proposed several robotic dressing assistance techniques including: a learning-from-demonstration approach applied to dressing a jacket sleeve and shoe [10], [11], dressing data analysis techniques for classifying dressing errors and distinguishing different underlying garment layers [12], and a dressing state machine controlled by user gestures and voice commands demonstrated by dressing a user in loose rubber shoes [13].…”

Probabilistic Real-Time User Posture Tracking for Personalized Robot-Assisted Dressing

“…Real-time updates of the trajectory also ensure a one-shot dressing, which avoids multiple trials and errors in [5,6,10,11] that may put the users at risk. The proposed method also avoids markers attached on users or clothes [11,12,45] that may cause inconvenience in daily dressing scenarios.…”