SCCRUB: Surface Cleaning Compliant Robot Utilizing Bristles

Scrubbing surfaces is a physically demanding and time-intensive task. Removing adhered contamination requires substantial friction generated through pressure and torque or high lateral forces. Rigid robotic manipulators, while capable of exerting these forces, are usually confined to structured environments isolated from humans due to safety risks. In contrast, soft robot arms can safely work around humans and adapt to environmental uncertainty, but typically struggle to transmit the continuous torques or lateral forces necessary for scrubbing. Here, we demonstrate a soft robotic arm scrubbing adhered residues using torque and pressure, a task traditionally challenging for soft robots. We train a neural network to learn the arm's inverse kinematics and elasticity, which enables open-loop force and position control. Using this learned model, the robot successfully scrubbed burnt food residue from a plate and sticky fruit preserve from a toilet seat, removing an average of 99.7% of contamination. This work demonstrates how soft robots, capable of exerting continuous torque, can effectively and safely scrub challenging contamination from surfaces.