联合收获机单神经元PID导航控制器设计与试验

针对联合收获机在田间直线跟踪作业中在维持高割幅率条件下易产生漏割的问题,设计了一种基于单神经元PID(Proportion Integration Differentiation)的联合收获机导航控制器。以轮式联合收获机为平台,通过对原有液压转向机构进行电控液压改装,搭载相关传感器构建了导航硬件系统。开展了常规PID控制和单神经元PID控制的仿真以及实地对比试验,仿真结果表明单神经元PID控制具有超调小和进入稳态快等特点;路面试验表明,当收获机速度为0.7 m/s时,单神经元PID控制最大跟踪偏差为6.10 cm,平均绝对偏差为1.21 cm;田间试验表明,收获机速度为0.7 m/s时,单神经元PID控制田间收获最大跟踪偏差为8.14 cm,平均绝对偏差为3.20 cm。试验表明所设计的联合收获机导航控制器能够满足自动导航收获作业要求,为收获作业自动导航提供了技术参考。

Design and experiment of the single-neuron PID navigation controller for a combine harvester

Abstract: Improving the intelligence level of combined harvesting machinery can improve harvest efficiency, harvest quality, extend operation time, reduce labor cost and labor intensity. In this study, combined harvester automatic navigation software and hardware system was designed to improve the intelligence of the combined harvester, and a navigation controller of combine harvester based on single-neuron PID was designed for the problem of no leakage of straight-line tracking operation in the field environment of combine harvester under the condition of maintaining high cutting width rate. This study used the combine harvester as the research object, and its steering system was modified by electronic hydraulic. The combined harvester navigation hardware system consisted of RTK positioning module, angle sensor, electric hydraulic steering mechanism, computer, navigation and control box (data acquisition card, proportional amplifier, power supply). Navigation software developed based on Windows 10 operating system. The control terminal realized the reception of high-precision BeiDou positioning data, coordinate transformation, heading deviation and distance deviation calculation, navigation control decision, steering angle monitoring and send control commands. The entire navigation system workflow was that the control terminal first collects target path information to determine the target tracking path, converted the current position information of the harvester to Gaussian projection into plane coordinates and calculated the lateral deviation and heading angle in real-time. The obtained deviation information was filtering processed for deviation construction strategy decision and the obtained deviation amount was used as the input of a single-neuron PID controller. The single-neuron PID controller calculated the output target steering angle and sent a control command to steering PD controller, then steering controller calculated and output analog to control the opening of the electro-hydraulic proportional valve and the direction of working fluid flow, achieved rear steering wheel control and tracking target line. In this study, commonly used PID control was adopted. In order to make up for the shortcoming that the traditional PID controller could not adjust parameters online, the single-neuron was introduced to adjust parameters online. The single-neuron was the most basic control component in neural networks, the single-neuron network had only one layer of neurons, the output was obtained by the input according to a certain functional relationship, through the self-learning of the single-neuron, the connection strength between neurons was modified so that the acquired knowledge structure could adapt to the changes of the surrounding environment. Combining the single-neuron network with traditional PID controllers could achieve online parameter adjustment and optimization of PID parameters. The single-neuron PID control had self-adaptation and self-learning capabilities and had a simple structure and a small amount of algorithm calculation. The single-neuron PID control could meet the real-time requirements of the system and make up for the shortcomings of the traditional PID controllers. Matlab simulation experiments were performed on the designed conventional PID controller and single-neuron PID controller. The simulation results showed that the single-neuron PID control had the characteristics of fast response, small overshoot and fast steady-state. The PD steering controller was designed for the electronically controlled hydraulic steering mechanism. The steering controller tracking error obtained by the square wave tracking test was 0.5°. The results proved the steering control performance of the designed steering controller. The road and field comparison tests were performed on the conventional PID controller and the single-neuron PID controller. The road test showed that the average absolute deviation of navigation and tracking in the road at 0.7 m/s was 1.21 cm, the maximum tracking deviation was 6.10 cm. The field test showed that the average absolute deviation of navigation and tracking in the field at 0.7 m/s was 3.20 cm, the maximum tracking deviation was 8.14 cm, the standard deviation was 2.82 cm. The experiments showed that the designed single-neuron PID navigation controller was superior to the conventional PID control, and could achieve a certain control accuracy and meet the requirements of combine harvester field operations.