A Decentralized Bayesian Approach for Snake Robot Control

Snake robot control usually adopts a centralized framework using high dimensional states, which suffers from exponentially increased computational cost with module number. Although decentralized control schemes have been proposed by using central pattern generators, the dynamical correlation among snake modules and also the interaction with environment have not been fully investigated. In this letter, we proposed a Bayesian decentralized framework for snake robot control, which can exploit efficient parallel computing, one agent per module for the snake robot. There are two primary contributions: 1) A probabilistic propagation rule is derived to model the uncertainty during locomotion in cluttered scenario with obstacles; 2) An inter-module likelihood and a module-based virtual external force density are introduced to simulate the corresponding interaction among modules and with environment. Our experimental results show promising performance of the proposed method on real world data.