3-DoF Snake Robot Head Stabilization for Visual Navigation Using a Coaxial Spherical Parallel Manipulator

Abstract

This letter proposes a three-degree-of-freedom (3-DoF) head stabilizer, defined as a compact camera-orientation stage mounted between the snake robot's head link and the onboard monocular camera, based on a coaxial spherical parallel manipulator (CSPM) to enable visual navigation. Visual navigation suits biologically-inspired robots with tight form-factor constraints because it can leverage datasets collected on other platforms and operate with a single monocular camera. However, locomotion-induced head motion in wheel-less snake robots causes camera-orientation oscillations and in-plane image rotations, making it difficult to obtain stable image sequences. To address this, we formulate a matrix-based camera-stabilization command from the head-mounted IMU and a PCA-based locomotion-direction estimate, and use an idealized 3-DoF active spherical joint to quantify the resulting stabilization demand. We then adopt an established coaxial-SPM model to evaluate the task-specific roll-pitch feasibility and select CSPM geometry with a singularity margin. The resulting CSPM-based head stabilizer is integrated into a snake robot simulation, where it attenuates locomotion-induced camera-orientation oscillations and reduces in-plane image rotation, producing steadier image sequences for monocular visual navigation. When coupled with a topological visual-navigation model trained on heterogeneous data, the stabilized image stream supports zero-shot monocular visual navigation in simulated wheel-less snake-robot scenarios.