Simulation Research On Gait And Path Planning Of Quadruped Robot

Quadruped robots possess stronger terrain adaptability than wheeled robots,enabling them to traverse rough terrain and navigate complex environments,making them suitable for various fields such as search and rescue,exploration,and transportation,with high flexibility and broad prospects.Gait planning refers to determining the movement trajectories of the quadruped robot’s limbs in a given environment,enabling the robot to walk smoothly and efficiently.A reasonable gait control strategy is the basis for stable and efficient locomotion of quadruped robots in complex environments.Path planning refers to determining the robot’s movement path in a given environment,allowing the robot to reach its target location.By utilizing a reasonable path planning algorithm,it is possible to enhance the dependability and effectiveness of quadruped robots in real-world scenarios through the avoidance of collisions with obstacles or the need for detours.The main research content of this thesis includes:1.Based on the structure of mainstream domestic and foreign quadruped robots,a three-dimensional model of a quadruped robot was established in Solidworks,and a URDF file describing the structure and kinematic characteristics of the quadruped robot was exported,and forward and inverse kinematic analyses were conducted on the model.Providing models for subsequent simulation and laying the mathematical foundation.2.The gait control system of the quadruped robot was designed,including the gait pattern adjuster and foot trajectory design,enabling control of the robot’s motion.Apply the deep reinforcement learning algorithm SAC(Soft Actor-Critic)to quadruped robot gait control,optimizing the trajectory parameters through the deep reinforcement learning algorithm SAC,creating unique trajectories for the feet depending on the surrounding environment.,thus improving the adaptability and motion stability of the quadruped robot to different terrains.The above strategies were trained and tested in the pybullet simulation environment,verifying the efficiency of the gait management approach that relies on the SAC deep reinforcement learning algorithm.3.The RRT-Dijkstra global path planning algorithm,which combines RRT and Dijkstra,was proposed.Firstly,the RRT algorithm was improved and optimized by setting the target node sampling rate,dynamically setting the step length and extending new nodes,improving the search efficiency of the RRT algorithm and the path search ability in narrow spaces.The improved RRT algorithm was then combined with the Dijkstra algorithm to reduce the distance of the global path generated by the RRT algorithm.The dynamic window approach(DWA)local path planning algorithm was used to adjust the path evaluation function,reducing its path planning time.The optimized and improved DWA algorithm was then combined with the RRT-Dijkstra algorithm to adjust the smoothness of the global path.The fusion algorithm was compared and analyzed in Matlab,it shows the superiority of the algorithm and the dynamic obstacle avoidance function.The effectiveness of the fusion algorithm was verified by deploying it on a quadruped robot in a ROS-based simulation environment.4.A physical platform of a quadruped robot was established,including the design of the robot’s software and hardware systems.On this basis,the above gait control strategy and path planning control algorithm were applied to the platform,verifying the effectiveness of the deep reinforcement learning gait control strategy and the path planning fusion algorithm.