Hybrid Control Strategies for Safe and Adaptive Robot-Assisted Dressing

TL;DR

Addressing safety in robot-assisted dressing, the paper proposes hazard-driven low-level control strategies to mitigate garment snags and user discomfort in real time. It combines real-time force monitoring with a Rasa chatbot for bi-directional interaction, enabling autonomous snag recovery, user intervention, and safe task abortion. Through human-human dressing trials to calibrate force thresholds and interaction strategies, the approach demonstrates improved safety, task continuity, and user trust. The work advances RAD toward responsive, personalized HRI systems by blending autonomous interventions with user-driven adaptation in real-world tasks.

Abstract

Safety, reliability, and user trust are crucial in human-robot interaction (HRI) where the robots must address hazards in real-time. This study presents hazard driven low-level control strategies implemented in robot-assisted dressing (RAD) scenarios where hazards like garment snags and user discomfort in real-time can affect task performance and user safety. The proposed control mechanisms include: (1) Garment Snagging Control Strategy, which detects excessive forces and either seeks user intervention via a chatbot or autonomously adjusts its trajectory, and (2) User Discomfort/Pain Mitigation Strategy, which dynamically reduces velocity based on user feedback and aborts the task if necessary. We used physical dressing trials in order to evaluate these control strategies. Results confirm that integrating force monitoring with user feedback improves safety and task continuity. The findings emphasise the need for hybrid approaches that balance autonomous intervention, user involvement, and controlled task termination, supported by bi-directional interaction and real-time user-driven adaptability, paving the way for more responsive and personalised HRI systems.

Figures (8)

• Figure 1: A typical Robot-Assisted Dressing scenario showing the robot's workflow for assisting a seated user in donning a jacket. The sequence includes receiving a dressing request, detecting the user's pose, locating and gripping the garment, and guiding it in real-time along an adaptive trajectory.
• Figure 2: RAD System architecture. The ROS network connects the Chatbot PC, Sensor PC, and Robot PC, enabling real-time communication and dynamic updates to the dressing trajectory. The ZED camera captures the user's pose for skeleton tracking, while the Franka Emika robot executes the dressing task using trajectory updates and real-time feedback.
• Figure 3: Workflow for Rasa-ROS integration in the RAD system. User input is processed by Rasa to classify intents and trigger appropriate ROS commands, which control actions such as pausing, adjusting speed, or resolving snags.
• Figure 4: Force (N), potential snags (15N-35N), escalated snags ($>$35N), and velocity magnitude over time in a trial with human intervention. When force exceeded 35N (red dashed line), the robot paused and requested user assistance. Successful intervention allowed dressing to continue.
• Figure 5: Autonomous control strategy for garment snagging during robot-assisted dressing. The figure shows force, recovery attempts, and task outcomes, highlighting escalated snags ($>$35N), compliance mode activation, recovery attempts, timeout conditions, and transitions to safe and home positions.