Study On Control And Fault Detection Of Wheeled Mobile Picking Robots

Wheeled mobile picking robot is widely used in fruit and vegetable picking,and challenges in its navigation and fruit picking control have always been research hotspots.In terms of navigation,current studies generally fail to consider the limitations of complex agricultural environment and wheeled mobile picking robot’s own mobility in actual operation.In the face of complex environment,the above problems will lead to picking robot’s poor performance in mobile obstacle avoidance.Therefore,it is necessary to study the local planning and obstacle avoidance control of wheeled mobile picking robot in the complex agricultural environment.On the other hand,precise fruit and vegetable picking is the ultimate goal of the wheeled mobile picking robot,whose picking process and navigation work cooperatively.Therefore,it is also necessary to carry out research on fruit recognition and positioning of wheeled mobile picking robot and on planning and control of picking manipulator.In the navigation movement and picking process of wheeled mobile picking robot,its mobile chassis and picking mechanical arm may fail,thus affecting the work and safety of the robot.In view of this,it is necessary to further carry out the fault detection of the wheeled mobile picking robot,real-time monitoring of the working state of the wheeled mobile picking robot,to ensure that the robot can work smoothly.Therefore,the research on fault detection is of great significance in improving the safe autonomous work of wheeled mobile picking robot.The main work and contributions of this paper are as follows:1.The characteristics of the three walking modes of the picking robot mobile platform are elaborated.By comparison and adapting the actual situation of domestic plantations,the fourwheel mobile walking structure is preferred.This paper briefly describes the system composition of the wheeled mobile picking robot,and describes the experimental platform of the wheeled mobile picking robot built by the research group.Based on the four-wheel skidding steering mobile robot model,the motion chassis of the wheeled mobile picking robot was analyzed.Based on D-H coordinate system,the kinematics modeling of picking manipulator was carried out.2.Aiming at the problems of local planning and obstacle avoidance of wheeled mobile picking robot,a navigation control method based on improved artificial potential field was designed,which introduced the bacterial foraging and particle swarm optimization(BF-PSO)for off-line optimization of controller parameters and potential field function gain coefficients.A navigation algorithm based on interval type-2 fuzzy neural network(IT2FNN)and Q-learning was proposed to solve the navigation problem in unknown complex environment during picking process.This algorithm solves Q learning algorithm’s problem of mapping from state space to behavior space in unknown environment.The robot can complete path planning with fewer fuzzy rules and realize autonomous navigation with fewer steps in complex environment.The feasibility of the two algorithms in real environment is verified by joint global programming experiments.3.Taking pitaya fruit as the picking object,researches on the identification and positioning algorithm of the cutting point of branches and fruit of pitaya were carried out respectively.A class activation-faster regions with convolutional neural network(CA-Faster R-CNN)detection and localization algorithm is proposed for the cutting points of pitaya branches.This algorithm reduces the positioning error without reducing the recognition accuracy.For pitaya fruits,a pitaya-You only look once(P-YOLO)detection method was designed to effectively identify and locate pitaya fruits in natural environment.Finally,the motion planning method and control strategy of the picking manipulator were designed.Combined with the visual system,experiments on the motion planning control of the picking manipulator were carried out to verify the effectiveness of the motion planning control of the picking manipulator.4.For the sensor and actuator faults of the walking system of wheeled mobile picking robot,a study on model-based interval observer fault detection was carried out.Firstly,a fault detection method based on residual error generated by interval observer of the linear time-varying(LTV)system was designed,and the interval observer was applied to the fault detection of sensor of wheeled mobile robot.The fault detection tests of sensor and actuator were carried out on the experimental platform,which provided a new idea for the theory of robot fault detection.5.Picking manipulator can be considered as a kind of nonlinear system described by state and input whose bounds of unknown nonlinear function are unmeasurable.The existing interval observer method fails to effectively deal with the unknown nonlinear part with parameter uncertainty,so a neural network adaptive interval observer design method is proposed to detect faults in manipulator sensor and actuator.Finally,the effectiveness of the proposed neural network adaptive interval observer and the feasibility of the proposed fault detection method are verified by simulation.