Research On Biped Robot CPG Control Method Based On Joint Coupling And Layered Reflexes

As a dynamic system, legged robot has the characteristics of nonlinear and variable structures. In current control methods, the model-based approach needs accurate modeling of body and environment, whose complicated measuring and calculating lead to the weak real-time performance. Due to the lack of global planning and upper control, the behavior-based approach has a low level of intelligence and flexibility. With its simple structure, good internal coupling mechanism and adaptability to the environment, central pattern generator is being used to control various types of robots by researchers. But the CPG robot motion control system is still lack of rapidity, coordination and adaptability, which leads to a large gap with the control effect of biological CPG.In this paper, in order to improve rapidity, coordination and adaptability of CPG motion control system, the mathematical model of multi-layer CPG neural network is established. With reflection mechanisms based on biological reflexes, the CPG motion control system achieves the bipedal robot walking on simple structured terrain.Firstly, the joints coupling rules is refined from joint movements of human walking, and the coupling form of CPG network output signal is established, which provide the basis for the CPG network output target. On this basis, the modified Van der Pol oscillator is used to build the single neuron model, and the layered chain network of CPG which has rhythm generator, pattern generator and motor neurons is built by using the contract theory. In order to generate the pattern with environment adaptability, the coupling rules is used to change the coupling form of CPG network, which realizes the rapidity. And the stability, coordination is proved by simulation experiment.Secondly, with the three-layer-reflection model of animal studied, the layered reflection mechanism of multilayer CPG motion control system is built. The reflection mechanism division of robot is built with the layered reflection mechanism. On this basis, in an order of the necessity of robot walking, the corresponding reflection model of biped robot is established according to the reflection strategy of human walking. The output signal of multilayer CPG network is adjusted through phase adjustment, posture reflection, which realized an interaction among control system, robot and environment. The stability and environmental adaptability of the robot rhythm movement is improved, and the effectiveness of each reflection mechanism is verified by simulations.Finally, with physical model of biped robot built, myotatic reflex and crossed extensor reflex fused, the joint simulation method of ADAMS and MATLAB is used to simulating biped robot walking with instant impact, period impact, on slope and on stair. The experimental results validate the effectiveness and feasibility of multilayer CPG network, which is fused with layered reflexes, on generation of bipedal robot rhythmical gait.