Motion Control Simulation Of Hexapod Robot

In the process of human's exploring the world,there are many places that can't be reached.Because of its strong carrying capacity,stability and adaptability to various types of landforms,biped robot has become the research focus.Compared with quadruped robot,hexapod robot possesses better stability,and the mechanism is simpler than octopod.Therefore,a simpler motion control strategy can be applied to hexapod robots,which is of great significance in applications.In this paper,a six legged robot model is designed and constructed,and the motion control strategy is studied and simulated from three motion control methods: toe trajectory planning,central pattern generator and deep reinforcement learning.There are the main contents of this paper:1.In order to verify the effectiveness of the control algorithm in the simulation environment,the mechanism of the hexapod robot is constructed,and the gait of the hexapod robot is analyzed.Firstly,due to the large number of mathematical model parameters of the existing robot toe trajectory planning scheme and its uncertainty,a toe trajectory planning scheme is proposed.Secondly,the forward and inverse kinematics solutions of the robot are analyzed.Finally,the robot motion control is realized by combining the trajectory planning of robot toe and the inverse kinematics solution of robot leg.2.With the central pattern generator(CPG)as the research object,a CPG model for robot rhythmic motion control is constructed by using Matsuoka and Hopf oscillators.In order to achieve a variety of gait movements,the Hopf oscillator of hexapod robot is improved to control the hexapod robot.Then the parameters of the oscillator model are analyzed.The CPG network structure is constructed to control the hexapod robot,and the output signal of the oscillator is processed through the joint mapping function to control the thigh and tibia joints of the robot.Finally,the CPG network is constructed based on the two oscillators in the simulation environment to realize the robot motion control,and the robot motion is analyzed quantitatively.The two oscillators are compared in two aspects:robot motion control effect and parameter complexity.3.Based on the comparison between LQR control system and reinforcement learning intelligent system,the state value,action value and reward function of intelligent system,which are related to robot motion control strategy training and learning are determined.The robot is trained to use the robot motion control strategy and learns it based on ddpg algorithm.In order to improve the training efficiency,an early termination module in the process of intelligent system movement is designed.When the robot moves to a dangerous state,the training round is terminated in advance,which improves the training efficiency of intelligent system.Finally,the motion control strategies are compared by setting different terrain structures,including plane,slope,gully and so on.And,the full text is summarized and prospected..