Motion Control Method Research Of Legged Robots With Partial Elastic Joint

Insects are one of the most diverse populations in the world,and the hexapod robots is one of the products inspired by their biomimetics.The hexapod robot is used for complex environments with low moving speed requirements because of its numerous degree of freedom of the joint and highly stable during movement.For the research of hexapod robots,scholars at home and abroad have achieved rich results.In recent years,more and more heuristic algorithms and machine learning algorithms have been used in gait control of robots.The robot can move smoothly and quickly in many complex environments that were previously impossible.Robots can even learn how to stand or even walk from nothing.In this paper,the structure and control methods of hexapod robot are improved and innovated from three aspects: bionic joint design and its influence on the energy utilization of the robot with specific conditions.Solution method of terminal trajectory in the case of unstructured terrain with ground information.Feasibility of classical gait switching of hexapod robot based on feedforward neural network.Firstly,according to the requirements of the robot,the hexapod robot is designed as a structure that can be switched between the hexapod and the spherical.The details of the body,the legs,the upper shell are described in detail.Then,the bionic optimization of elastic joints is carried out,including the source of the bionics of elastic joints,the simplified model of elastic joints,and the equivalent model of joint.Then the hexapod robot is based on the DH coordinate method for the theoretical analysis of kinematics and dynamics.The kinematics and dynamics relationship between the body and the legs of in the swing phase or in the supporting phase are obtained.After analyzing the dynamics and energy consumption of elastic joints,the two optimal spring stiffness which can minimize the average output torque required for driving and minimize the energy output in one cycle are obtained under the specific joint rotation angle function.Compared with the fully elastic joint,the partial elastic joint can release the elastic potential energy when a certain joint of the robot needs to be accelerated and decelerated,and assist the acceleration and deceleration of the drive to reduce the necessary output of the drive.Since the required torque is a fixed value at the conditional timing,the more the auxiliary output of the spring,the less the drive output,but it will bring some instability.The construction of the robot control system will be explained later,and the intelligent control level of the robot will be explained.Then select the robot control system and hardware,including the description and selection of the central processor,sensor,servo drive,power system and other modules.Then,the two aspects of end trajectory planning and gait switching are studied: the unstructured terrain pavement with known ground conditions is determined by Lagrange multipliers under Karush-Kuhn-Tucher conditions.Under the condition,the generation of the end trajectory curve is transformed into the optimal loss function under the constraint condition.The loss function and the constraint condition include the influence on the organism due to the introduction of the elastic joint.In the case that the original problem is difficult to solve,it can be transformed into a convenient and streamlined solution to the dual problem.Based on the end trajectory,the feasibility of gait switching based on feed-forward neural network and the hyperparameters such as learning rate,depth,number of units and activation function type of the experimental network are considered according to the classical hexapod robot gait planning.Finally,the simulation and prototype walking experiment are carried out.The corresponding experimental platform is built in Adams and Matlab simulation software.The simulation experiment of energy consumption of partial elastic joints and the simulation of hexapod robot gait switching based on feedforward neural network are carried out.The results of partial elastic joint Adams simulation experiments show that the joint can effectively reduce the energy consumption of the legged robot when walking,and obtain the specific consumption reduction data under given conditions.The Matlab simulation experiment of gait switching verifies the feasibility of gait switching based on feedforward neural network,and obtains the time of process in gait switching and its ratio with gait cycle.