面向蛇形机器人的三维步态控制的层次化联结中枢模式生成器模型

A Hierarchical Connectionist Central Pattern Generator Model for Controlling Three-dimensional Gaits of Snake-like Robots

提高蛇形机器人的三维运动控制能力是提高蛇形机器人环境适应能力的关键之一.虽然联结中枢模式生成器(Connectionist central pattern generator,CCPG)模型具有复杂度小、适合硬件实现等优点,但是目前的CCPG模型难以生成相位协调的多自由度运动的控制信号,从而限制了它的三维步态控制能力.本文根据生物CPG机制的分层结构和运动神经元的功能,提出一个有层次化结构的CCPG(Hierarchical CCPG,HCCPG)模型.HCCPG模型由基本节律信号生成层、模式形成层、运动信号调整层这三个部分组成.运动信号调整层的运动神经元能够独立地对模式形成层的输出信号的幅值、相位等进行调整,从而较好地解决了CCPG模型难以生成相位协调的多自由度运动控制信号的问题.HCCPG模型具有步态控制能力强、复杂度小、有良好的扩展性等优点,从而适合用于控制三维步态.在HCCPG模型的基础上提出一个三维步态控制方法.仿真验证了这个控制方法的有效性.

A key to promoting the snake-like robot＇s adaptability is improving its 3-D locomotion ability. Although the connectionist central pattern generator models （CCPG） have merits such as low complexity, appropriate for the hardware implementation, the current CCPG models have difficulties in generating the phase-coordinated control signals for the multi-degrees-of-freedom motions. Consequently, the CCPG^s ability to generate the 3-D gaits of snake-like robots are seriously restricted. According to the layered structure of the biological CPG mechanism and the functions of the motoneuron, a hierarchical CCPG （HCCPC） model is proposed. The HCCPC is composed of three layers, namely the basic rhythm generation （RG） layer, the pattern formation （PF） layer, and the motion modulation （MM） layer. The motoneurons of the MM layer can independently modulate the amplitude and phase of the PF layer＇s output so it overcomes the difficulty faced by the current CCPG models. The HCCPC model has merits such as strong gaits adjustment ability, small complexity, and good expendability, which make it appropriate for generating the 3-D gaits. Based on the HCCPG model, a 3-D gait control method is proposed. Simulations have validated this gait control method.