Research On The Selection Of Hexapod Robot Motion Mode Based On Adaptive Neural Network And Learning

The hexapod robot with a wide range of applications can be adaptive to the change of terrain under the complex field environment,and has a certain obstacle avoidance ability.Therefore,the research on adaptive control methods of hexapod robots is becoming more and more important.Nevertheless,it is difficult for the current behavior mode control methods of hexapod robots to maintain stable and efficient walking in the field environment.In response to this problem,this paper launches a research method for adaptive control of multi-legged robots based on neural networks and learning feedback to achieve stable and low-energy consumption walking of salticidae hexapod robots in the complex field environment.The main research contents are as follows:A model of the salticidae hexapod robot is built according to the biological mechanism of salticidaes.With the analysis of the forward and inverse kinematics of the single leg with redundant degrees of freedom,the foot trajectory of the salticidae robot is planed.Using a CPG-based gait generator.The CPG-BC signal is employed as the driving signal for the foot trajectory of a six-legged robot controlled by Hopf harmonic oscillators in six different phases.A neural network framework for adaptive behavior mode selection is established,along with the hidden layer of the neural network,which consists of the terrain category layer,the surface layer and the attitude angle layer.In the terrain category layer,a terrain classification system containing six typical terrains is established,the surface layer recognizes the undulation and slope of the terrain,and the real-time feedback of the pitch angle and roll angle of the attitude angle layer.Establish an adaptive learning module based on energy consumption.The aim of learning is to reduce the energy consumption and adjust the synaptic weight of each hidden layer.Based on neural network and adaptive learning module,a behavior mode selection control method for spider-jumping robot is proposed.An experimental prototype platform of the salticidae hexapod robot is built to verify the adaptive behavior pattern selection algorithm.Experiments show that the behavior mode selection control based on neural network and adaptive learning module is effective for speed and gait experiment of spider-jumping robot in complex field environment.