Research On Self-grown Based CPG Local Network

With the development of technology and related disciplines, researches onrobot have gotten a wider range of technical methods. Different from thetraditional design concepts, the bionics method of copying and recreating theform, function and control mechanisms of real animals through observing,learning and imitating the animals’ behavior of interacting with their livingenvironment have brought new ideas to research on robot. Over the years,imitation and replication of biological structure has brought great suc cess to thefield of robotic research. And also, with the deepening of neurobiology andbionic mechanism, biomimetic control method based on CPG network isbecoming the hot topic of the foot robot. In the research field of CPG networkbased control method, the topology and gait generation method of CPG networkis the focus of common concern. The theory and process of structure formation isthe bases of constructing CPG network, and it’s on the starting stage, research inthis field has important significance.In this paper, the research is based on development of biological neuralnetworks, and study the method of constructing CPG local network.Firstly, statistic features and density of neurons distributed in the brain wereanalyzed. The statistics laws of multiple neurons distributed randomly in alimited square field with equal probability were studied. The basic environmentfor the development of multi-neuron network was builded.Secondly, the biological mechanisms of neuronal development, axonalgrowth and formation of local neural network were analyzed. The simplifiedmodels of neurite initiation and differentiation, neurite elongation, axon pathfinding were builded. By studying neuron axon growth features under theconstraints of the gravitational field, the rules and mathematical treatment ofmulti-neuron axon growth in the plane were study. By analyzing results of a largenumber of network growth experiments, the mapping function between theconnected length of the neurons and their mutual phase relat ionship wasobtained.Finally, according to the network growth method designed before, thenetwork self-growth experiments were implemented and got the phaserelationship corresponding to two common gait respectively. The experimentsverified the effectiveness of the network self-growth algorithm designed basedon the gravitational-field constraints. By the use of Open Dynamics Engine(ODE), the quadruped robot motion control simulation platform wasbuilded to complement the robot rhythmic walking. By integrating the method ofnetwork growth and leg motion controller, the quadruped robot rhythm motioncontrol subsystem was builded. The experiment achieved rhythm walk ofquadruped robot based on self-growth network.