Research On Electro-hydraulic Servo Control And Crop Information Optimization Method For Weeding Robot

Non-chemical weed control is a premise of not using of chemical herbicides and organically growing. Mechanical weed control operations with soil-engaging tools can increase soil aeration and water infiltration including related positive effects. Intra-row weeds which are mainly removed by human currently are more difficult to be eliminated than inter-row weeds due to their proximity to the crop or seed line. Manual weed control has the disadvantages of high cost and low efficiency. Intelligent intra-row weeding robot can differentiate and remove intra-and inter-row weeds simultaneously, and can navigate along crop rows autonomously, while identify and provide species information. In this paper, the hydraulic system and hydraulic servo control system was designed based on the mechanical unit of weeding robot. Machine vision data optimization method、manipulator motion control based on electro hydraulic servo technology and robot navigation method long crop rows were researched.(1) The hydraulic system of weeding robot which was mounted on the Oriental Red 804 tractor as supporting power has been designed. Hydraulic circuits of frame lateral movement, steering movement, steer wheel lifting, hydraulic motor speed control and other functions were designed respectively. Besides, the parameters of hydraulic cylinder and motor were calculated and verificated. Hydraulic control system of parallel lower computers which are based on embedded system has been designed.(2) A crop information optimization method based on sensor fusion was proposed, because the unstructured environment could degrade the performance of machine vision. The fuzzy corrector based on the fusion of odometry and computer vision was designed. The Mamdani fuzzy inference method was used to obtain the reliability and weighted value of 2 sensors, and then, the relatively accurate crop positioning data was gotten by weighting the reliability of 2 sensors. Ststic tests showed that the average error of crop positioning data reduced from 9.4mm to 5.9mm after correction.(3) Using odometry to compensate machine forward displacement in the image processing time could realize to solve the problems of the system real time lag. A method which used a certain odometry to estimate and interpolate the next moment crop positioning data was proposed when the time consuming of image processing was serious. Dynamic test shows that the average error of optimized crop positioning data was 4.3--8.1 mm at the forward speed of 1.1-1.7km/h. The sampling frequency of crop positioning data increased about 6 times.(4) The weeding manipulator continuous rotation control method was designed. The rotation speed of motor was real-time calculated according to phase feedback of motor and forward speed information, which make the motor rotate 360° between two crops. The influence of sensor data error on the speed of motor was analyzed, through the derivation and discretization of the speed formula. Simulation trajectory of weeding manipulator shows the relationship between covering area of manipulator between the crops and intra-row space, and diameter of manipulator. Field test shows that the control method is not suitable for the case of large changes in intra-row spacing. Weeding rate were 89.2% and 62.3% respectively in the 300mm and 600mm intra-row spacing. Extra rotations and missed rotations were not many. Less than 10% crops were damaged under 1-2km/h forward speed.(5) The weeding manipulator intermittent rotation control method was designed. The threshold spacing was used to switch modes of piecewise intermittent rotation control. The method of the single threshold crop positioning data was used for switch quantity control of the rotation of motor. Field test showed that the piecewise intermittent rotation control would cause unacceptable damage to crops, and the switch intermittent rotation control could adapt the working conditions of intra-row spacing changing. Weeding rate were 91.2% and 89.9% respectively in the 300mm and 600mm intra-row spacing. Missed rotations were slightly more compared to the other method. Less than 10% crops were damaged under 1-2km/h forward speed.(6) The crop rows following movement control which was divided into robot steering conrol and tractor driving assistance control was studied. Robot steering conrol was divided into the small row angle condition in which the row lateral deviation was taken as the control input and large row angle condition in which the current crop location was taken as control inputs. The method based on fusion of frame shifting deviation and row angle was used for tractor driving assistance control. Field comparison test showed that crop rows following control could increase the performance of weeding robot movement along rows. The root mean square error of row lateral deviation was 31.45mm,28.08mm less than before.