Research On Linear Auxiliary Driving System Of Paddy Management Machine Based On Seedling Belt Recognition

This study uses a four-wheel independent steering high-gap paddy field management machine that can meet the needs of paddy field management operations as a research platform.The goal is to reduce the crushing damage of the driving wheel train to the rice in the process of paddy field management.Through seedling belt identification,equivalent steering angle prediction,and MPC predictive control,low-injury seedling straight line assisted driving is finally realized.The research uses lidar to identify seedling belts and obtains the trajectory of the drivable seedling gap.Aiming at the problems of the steering structure's swimming gap and initial angle difference,the BP neural network equivalent steering angle prediction algorithm extracted from the feature information of the body sensor is used to realize the body steering Angle prediction,based on the acquisition of the driving trajectory and steering angle,combined with the MPC model predictive control algorithm,for linear navigation control,so as to realize the low-injury straight-line assisted driving of the high ground clearance paddy field management machine.The research mainly includes the following:1)This paper proposes a paddy field seedling band recognition algorithm which based on iterative slope virtualization mean clustering.On the basis of the conventional mean clustering algorithm,the seedling belt trajectory recognition is realized through slope virtualization,partition iterative clustering,and linear fitting.Compared with the ordinary clustering algorithm,the algorithm is based on the lidar detection of rice morphological characteristics,slope conversion and pixelization of the detected distance data,and combined with partition iterative clustering,so as to achieve the trajectory of the seedling belt,from a total of four field experiments have shown that the maximum median deviation is 24 mm,and the maximum transverse deviation is 42 mm in the front-view range of1 m,which has been proved that it basically meets the linear auxiliary navigation needs of the paddy filed.2)An equivalent steering angular BP neural network prediction algorithm based on the characteristics of steering angle spectrum is proposed.Aiming at the problem of steering clearance between four-wheel independent steering highland gap water field management machine,the spectrum characteristics of steering wheel steering angle,body acceleration,angular velocity and other attitude parameters during driving are deeply analyzed,and the regular parameters of the peak power spectrum frequency of the steering wheel is 0.2Hz are obtained,which are based on this method.For the training set periodic factor of BP neural network,the effect of acceleration,angular velocity and other equivalent steering angles is comprehensively evaluated.Finally,the conclusion is obtained that the Z-axis angle velocity and left and right front wheel steering angle parameters are obtained as effective training factors.By testing the test set,the MSE value of the equivalent steering angle predicted by the BP neural network is controlled in 0.44,and the mean variance error is controlled within0.66°,which verifies the effectiveness of the algorithm.3)In view of the narrow gap between seedlings when the the paddy field management machine is driving,the MPC model prediction control algorithm is used to carry out linear auxiliary navigation control,the MPC algorithm based on the bicycle model is modeled and analyzed,combined with the situation of paddy fields,the model is simplified and inferred,combined with the actual technical parameters and constraints of the paddy field management machine Through simulation in MATLAB environment,when the speed is0.25m/s,the steering control noise is ±1°,the equivalent steering angle error is ±0.6° and the track noise is ±4cm,the maximum trajectory deviation of tracking by equivalent steering angle input is 6.2cm,and the standard deviation is 3.39 cm,which meets the requirements of low-damage seedling trajectory tracking driving and verifies the feasibility of linear assisted navigation of the water field management machine based on the MPC model predictive control algorithm.4)In order to verify the feasibility of the theory and algorithm,field experiments are carried out for three different growth periods of rice in a season.Through the three-stage static seedling trajectory detection,the algorithm effectively identifies the characteristics of multiple seedling belt and wheel track.In the detection range,the midpoint deviations of the seedling belt in the early and middle periods are 11 mm and12mm respectively,and the median deviation of the post-term rotational track recognition is24 mm.In the characteristic recognition state of the seedling band,the clarity of the gap between seedlings affects the trajectory recognition accuracy.It can be seen from the wheel track detection data that the consistency of rice morphology affects the accuracy of track recognition.In the linear-assisted driving test,the mean variance error value corresponding to the theoretical trajectory of the intermediate test lidar detection trajectory is 1cm,the maximum error in the detection field is 4cm,the average variance error value of navigation control is1.05 cm,and the maximum error value is 4.1cm.The mean variance error value corresponding to the theoretical trajectory of the later test lidar is 1.3cm,the maximum error in the detection field is 5.5cm,the mean variance error value of navigation control is 1.4cm,and the maximum error value is 6.2cm.The MPC navigation control algorithm using equivalent steering angle input basically meets the needs of low injury seedling tracking in paddy field.In a word,through the above work,the linear auxiliary driving system of low-injury seedlings in the paddy field based on lidar has been studied in depth.From the seedling belt recognition algorithm,the equivalent steering angle prediction of steering wheel to the linear auxiliary navigation control,the corresponding solution strategy is proposed in this paper.Through the method proposed in this paper,the highland gap water field management machine can recognize the driving trajectory,accurately predict the forward wheel steering angle and linear auxiliary navigation at different growth periods of rice,which provides a technical basis for the low-injury seedling auxiliary driving of the highland gap water field management machine in the rice field.