Research On Key Technologies Of Mobile Robot Autonomous Navigation And Target Grasping

As an important part of the robot field,the theory and application of mobile robots have always been a hot topic and received wide attention.Different from the structured environment of industrial robots,the application environment of mobile robots has the characteristics of complexity,dynamics and uncertainty.The autonomous navigation of mobile robots has become the most urgent problem to be solved.On the other hand,robotic grasping is the basic task of robotic manipulation task,and the research of robotic target grasping has also attracted the close attention of many scholars.Therefore,this study takes mobile robot as the research object,and completes the key technology research on autonomous navigation and target grasping of mobile robot from three aspects: path planning,path tracking,target recognition and grasping position detection.(1)In view of the difficulties of the basic bat algorithm in balancing exploration and exploitation,which limit its efficiency in addressing global path planning,this paper proposes a hybrid bat algorithm combining bat algorithm and reinforcement learning to solve the global path planning problem.The optimization performance of the bat algorithm is determined by two parameters: the loudness attenuation coefficient and the pulse emissivity enhancement coefficient .In the hybrid bat algorithm,the Q-learning strategy is selected during the training phase to learn the optimal parameter values of each individual in the bat population,and then these trained optimal parameter values are applied to complete the global path planning task in the execution phase,so as to improve the optimization precision and convergence speed of bat algorithm.The proposed algorithm is compared to bat algorithm and particle swarm optimization in three different simulation environments.The simulation results show that the hybrid bat algorithm has satisfactory path planning effect and can be considered as a crucial choice for dealing with global path planning problems.(2)After obtaining the planned global optimal path,the next step is to handle the path tracking problem,in which the performance of path tracking control will directly determine whether the autonomous navigation problem of mobile robot can be effectively solved.Firstly,the kinematics and dynamics model of the two-wheel differential driving mobile robot is established,and the path tracking control law of the mobile robot is designed by backstepping method to calculate the expected values of the control variables.Afterwards,a PID control optimized by grey wolf algorithm is proposed,in which the grey wolf optimization algorithm is used to self-tune the PID control parameters to improve the dynamic performance of the control system.Eventually,the simulation and experimental results show that the proposed control algorithm can effectively and steadily control the mobile robot to complete the path tracking task.(3)In this paper,the deep learning method is applied to the grasping task of mobile robot,and a multi-task convolutional neural network is put forward.The network can deeply mine the visual features related to the robot grasping operation in the image,and can simultaneously accomplish the tasks of object recognition and grasping position detection.The experimental results show that there is a certain internal relationship between the object recognition and the grasping position detection.Moreover,the multitasking network has better performance than the single task network.The multi-task convolutional neural network is verified by the actual mobile robot experimental platform,which can successfully identify the target and detect its grasping position,finally guide the mobile robot to complete the target grasping operation.