| Snake-shaped robots have attracted the attention of many scholars due to their flexible movement,small size,and various forms of movement.Compared with other types of mobile robots,snake-shaped robots have stronger ground adaptability.There is a broad application space in detection and other aspects.In this paper,a modular wheeled snake robot system based on orthogonal joints is designed,a snake robot simulation platform is built,and A* and deep reinforcement learning algorithms are used for path planning and environment mapping.The specific research contents are as follows:Firstly,a modular wheeled snake robot structure based on orthogonal joints is designed.The joint module and the torso module of the snake-like robot are connected to each other in the way of orthogonal connection;in order to meet the lightweight design,the material is upgraded and improved,and the nylon material with high strength,low density and strong toughness is selected as the structure of the robot materials,and completed the construction of the snake-like robot entity.Secondly,establish a three-dimensional model of the snake robot,and build a simulation platform for the snake robot in Gazebo.First create a snake-like robot URDF model,use xacro to optimize the URDF file describing the robot model,and then display the snake-like robot on the Gazebo and Rviz platforms.Then the motion of the snake-like robot is analyzed,and finally the snake-like robot simulation platform is used for experiments to verify the rationality of the snake-like robot structure and the integrity of the simulation platform.Further,based on Adaptive Monte Carlo Localization(AMCL)to locate the snake-like robot,use the SLAM(Simultaneous Localization and Mapping)algorithm based on 2D lidar and IMU fusion to enable the snake-like robot to realize map building Function.In order to realize the autonomous navigation of the snake-like robot,the A*algorithm and deep reinforcement learning are used for path planning on the navigation framework of move＿base,and the environment is constructed at the same time.Aiming at the problem that the path of the A* algorithm does not fit the motion of the snake-like robot,a path planning method for the snake-like robot based on DQN(Deep Q-Learning)is proposed.The simulation results in Gazebo show that the designed snake robot can realize effective path planning and environment mapping in environments with and without obstacles.The proposed algorithm improves the success rate of snake robot path planning.Obstacles have better obstacle avoidance ability and reduce the number of collisions.Finally,build the snake robot hardware platform,design the snake robot control system,and carry out the physical verification of the snake robot body.The hardware platform of the snake robot is built by using the upper computer,single chip microcomputer,steering gear control board and mobile power supply;the software platform of the control system of the snake robot is designed,including the actuator module,the mobile terminal module and the PC terminal module.Finally,the physical experiment of path planning and map building is carried out.The experimental results show that the designed snake robot can achieve effective autonomous obstacle avoidance and construct an environmental map.The path planning algorithm based on the DQN algorithm is superior to the traditional A* algorithm,and the success rate of reaching the target point is significantly improved. |
