Simulating Safe Bite Transfer in Robot-Assisted Feeding with a Soft Head and Articulated Jaw

TL;DR

The paper tackles safe bite transfer in robot-assisted feeding by introducing a MuJoCo-based framework that models a soft-body head tightly coupled to a rigid skull via tendon-driven skinning. It reconstructs a deformable head from an RGB image using DECA, converts it to a tetrahedral mesh, and enables mandible actuation to simulate mouth opening. Bite transfer is parameterized by insertion depth $d$, entry angle $\alpha$, exit depth $e$, and exit angle $\beta$, allowing systematic exploration of contact forces $f_t$ to optimize user comfort. The main finding is that a straight-in, straight-out strategy with $\alpha = 90^{\circ}$, $d = 70\,\mathrm{mm}$, $\beta = 90^{\circ}$, $e = 10\,\mathrm{mm}$ minimizes both total and peak contact forces, suggesting practical guidelines for safe robot-assisted feeding; the work enables safer, simulation-based evaluation before real-world trials.

Abstract

Ensuring safe and comfortable bite transfer during robot-assisted feeding is challenging due to the close physical human-robot interaction required. This paper presents a novel approach to modeling physical human-robot interaction in a physics-based simulator (MuJoCo) using soft-body dynamics. We integrate a flexible head model with a rigid skeleton while accounting for internal dynamics, enabling the flexible model to be actuated by the skeleton. Incorporating realistic soft-skin contact dynamics in simulation allows for systematically evaluating bite transfer parameters, such as insertion depth and entry angle, and their impact on user safety and comfort. Our findings suggest that a straight-in-straight-out strategy minimizes forces and enhances user comfort in robot-assisted feeding, assuming a static head. This simulation-based approach offers a safer and more controlled alternative to real-world experimentation. Supplementary videos can be found at: https://tinyurl.com/224yh2kx.

Figures (6)

• Figure 1: Physical human-robot interaction simulation of bite transfer with contact forces between the spoon and the soft-body mouth shown by the blue and green spheres. The opacity of the head and skull is decreased to show the contact points.
• Figure 2: Overview of our approach to model a head with soft-body dynamics. (Top) A flexible head model is generated from an RGB image through surface mesh reconstruction, tetrahedral meshing, and loading into MuJoCo as a flexible element. (Bottom) The flexible head model is combined with a rigid skeleton through rigging, and skinning is applied to bind the two. Contacts are detected and marked with yellow cylinders (bottom right), after which sites are created at the detected points, visualized using red and green balls (bottom center). Tendon elements then connect the sites to form a fully rigged and actuated head model. This setup enables realistic deformation, mandible actuation, and mouth movement simulations.
• Figure 3: Bite transfer parameters modified in our experiments
• Figure 4: Bite transfer in simulation: The top row shows screen captures of the bite transfer process, with contact forces visualized as coloured spheres, showing increased contacts in (b). The bottom row shows the corresponding contact forces over time. Note that the y-axis scales are different for clarity.
• Figure 5: Force Analysis Results -- Entry Phase