Do You Need a Hand? -- a Bimanual Robotic Dressing Assistance Scheme

TL;DR

This work tackles the challenging problem of robotic dressing by introducing a bimanual framework in which an interactive robot guides the human arm via hand-holding while a dressing robot performs the garment task. A key insight is that the elbow angle $\psi$ governs dressing difficulty, motivating an optimal stretch policy and a posture-dependent dressing coordinate for learning from demonstrations. The method combines vision-free arm posture estimation, impedance-based guidance, and GMM/GMR-based imitation learning to achieve adaptable dressing across arm lengths and clothing types, validated by extensive experiments and ablations. The approach offers a practical paradigm shift toward two-robot-to-one-arm dressing, with potential impact on caregiving by reducing manual burden and enabling safer, more comfortable assistance.

Abstract

Developing physically assistive robots capable of dressing assistance has the potential to significantly improve the lives of the elderly and disabled population. However, most robotics dressing strategies considered a single robot only, which greatly limited the performance of the dressing assistance. In fact, healthcare professionals perform the task bimanually. Inspired by them, we propose a bimanual cooperative scheme for robotic dressing assistance. In the scheme, an interactive robot joins hands with the human thus supporting/guiding the human in the dressing process, while the dressing robot performs the dressing task. We identify a key feature that affects the dressing action and propose an optimal strategy for the interactive robot using the feature. A dressing coordinate based on the posture of the arm is defined to better encode the dressing policy. We validate the interactive dressing scheme with extensive experiments and also an ablation study. The experiment video is available on https://sites.google.com/view/bimanualassitdressing/home

Figures (15)

• Figure 1: The bimanual robot setup for the cooperative dressing. We employ two Franka Emika robots. The dressing robot grips the cloth and executes the dressing policy. The interactive robot is equipped with a SoftHand from qbrobotics that provides human-like hand-holding. It guides the human in the dressing process.
• Figure 2: The schematic diagram of cooperative dressing framework. The framework contains three blocks: a posture estimation, a dressing motion generator, and an optimal stretch controller. The posture estimation takes in an initial posture of the human arm and the current hand position and outputs the estimated current arm posture. The posture is then fed into the dressing motion generator together with the current dressing robot position to derive the next step dressing policy. The optimal stretch controller takes the initial human arm posture and the current hand posture yields the stretching policy for the interactive robotics arm.
• Figure 3: Illustration of the inner and outer path of the armscye. We consider the arm plane defined by the hand, elbow, and shoulder. The arm plane is divided into two parts: the inner arm area with an angle less than $\pi$ and the outer area with an angle more than $\pi$.
• Figure 4: The optimal direction that maximizes the increase of the elbow angle $\psi$. The arm posture is given by the shoulder, elbow, and hand joint positions: $\bm{P} = \{\bm{p}_{s}, \bm{p}_{e}, \bm{p}_{h}\}$. The optimal direction $\delta\bm{d}^*$ is aligned with the direction that connects the hand and the shoulder and marked with a red solid arrow. The triangle inequality theorem is demonstrated with another random direction $\delta\bm{d}$ and the resulting shoulder-to-hand distance is marked with a black dash line.
• Figure 5: A schematic diagram showing the inputs and outputs of the human posture estimation.