Non-impulsive Contact-Implicit Motion Planning for Morpho-functional Loco-manipulation

TL;DR

This work tackles loco-manipulation for a morpho-functional snake robot (COBRA) by formulating a non-impulsive implicit-contact path planning framework that integrates object interaction with locomotion. It develops a full-dynamics model, recasts contact interactions in local coordinates using the Delassus matrix $G = J_c M^{-1} J_c^T$, and solves a time-stepping optimization to compute external contact forces $f_{ext}$ and joint rates $u$ under dynamics and actuation constraints. The approach is demonstrated in high-fidelity Simulink simulations for flat-ground box manipulation across C-, S-, J-shaped lateral rolling and sidewinding gaits, revealing trade-offs in efficiency and energy use among gaits (e.g., S/J-shaped are more efficient, sidewinding moves more mass but at higher energy). The results validate the feasibility of non-impulsive contact-implicit loco-manipulation on a multi-joint snake robot and suggest practical insights for gait selection and real-world deployment with enhanced sensing and force estimation.

Abstract

Object manipulation has been extensively studied in the context of fixed base and mobile manipulators. However, the overactuated locomotion modality employed by snake robots allows for a unique blend of object manipulation through locomotion, referred to as loco-manipulation. The following work presents an optimization approach to solving the loco-manipulation problem based on non-impulsive implicit contact path planning for our snake robot COBRA. We present the mathematical framework and show high fidelity simulation results for fixed-shape lateral rolling trajectories that demonstrate the object manipulation.

Figures (9)

• Figure 1: Illustrates loco-manipulation problem concerning carrying a box with COBRA
• Figure 2: Full-dynamics model parameters in the object manipulation task considered in this paper
• Figure 3: Snapshots depicting simulated forward box push utilizing various gaits executed in Matlab. The J-shape gait is an asymmetric variation of the C-gait, which allows changing the direction of movement of the box by mirroring the gait.
• Figure 4: This image depicts the contact points and unilateral ground reaction forces during (a): C-shaped gaits, (b) S-Shaped Gait, (C) J-Shaped Gait, (d) Sidewinding gait, performed by the high-fidelity COBRA model simulated in the MATLAB environment. The contact forces consist of tangential forces ($\bm f_{T,i}$) and normal forces ($f_{N,i}$). L$x$ here refers to Link number $x$ on the robot, numbered from 1-10 starting from the head.
• Figure 5: This image depicts the contact points and unilateral ground reaction forces for contact interactions of the manipulated object with the robot and with the ground during (a): C-shaped gaits, (b) S-Shaped Gait, (C) J-Shaped Gait, (d) Sidewinding gait, performed by the high-fidelity COBRA model simulated in the MATLAB environment. The contact forces consist of tangential forces ($\bm f_{T,i}$) and normal forces ($f_{N,i}$). L$x$ here refers to Link number $x$ on the robot, numbered from 1-10 starting from the head.