Zhang Longmei, Liu Gangwei, Qi Yandong, Yang Tengxiang, Jin Chengqian

(Nanjing Institute of Agricultural Mechanization, Ministry of Agriculture and Rural Affairs, Nanjing 210014, China)

Abstract: Agricultural machinery driverless system is the basis for realizing precision agriculture.Research and development of agricultural machinery driverless system is an effective way to alleviate the shortage of labor force and realize precision agriculture.This paper introduces the latest progress in the navigation and positioning technology, navigation path planning technology and automatic steering technology of agricultural machinery in China and abroad.The navigation and positioning technologies of agricultural machinery based on GPS positioning technology, Beidou positioning technology and visual navigation are expounded respectively.The global path planning algorithm based on full coverage path planning and global point-to-point path planning is expounded.Local path planning based on obstacle avoidance and tracking, and automatic steering technology based on motor power and electronic hydraulic control.These technologies have been relatively mature after years of development, and have been widely applied in agricultural production practice.However, there are still many problems, including the positioning technology is vulnerable to environmental interference, the utilization rate of path planning technology data is low and the dependence on environmental information is high, and the automatic steering technology is complex in the modification process, and it is easy to damage and difficult to maintain in harsh field operations.Based on the above, it is proposed that integrated navigation, multi-sensor fusion technology and modular automatic steering device are the future research directions of agricultural machinery unmanned driving system.

Keywords: navigation and positioning; path planning; automatic steering; obstacle avoidance; path tracking

0 Introduction

The progress of science and technology is promoting the rapid development of modern agricultural production to automation and precision.Agricultural machinery with automatic navigation function is increasingly applied to agricultural production[1].The key technology of agricultural machinery autonomous driving system will greatly improve the independent innovation ability and industrialization ability of agricultural machinery and equipment, break the monopoly of foreign agricultural machinery autonomous driving key technology, and promote the improvement of relevant technical level of intelligent agricultural equipment in China.The key technologies of agricultural machinery unmanned driving system mainly include navigation and positioning technology, navigation path planning technology, automatic steering technology[2].This paper analyzed the navigation and positioning technology at home and abroad from three angles of GPS positioning technology, Beidou positioning technology and visual navigation, and summarized the research progress of agricultural machinery navigation and positioning technology.This paper analyzed the navigation path planning technology at home and abroad from the perspectives of global path planning and local path planning, and summarized the research progress of agricultural machinery navigation path planning.The automatic steering technology at home and abroad was expounded and analyzed from the perspectives of motor power steering and electronically controlled hydraulic steering, and the research progress of agricultural machinery automatic steering technology is reviewed.Finally, the future research directions of positioning technology, path planning technology and automatic steering technology of unmanned system are prospected.

1 Positioning technology

1.1 GPS positioning

GPS global positioning system is a high-precision radio navigation positioning system based on artificial earth satellites, which can provide accurate geographical location, vehicle speed and accurate time information[3].

GPS positioning includes pseudorange single point positioning, carrier phase positioning and real-time differential positioning anywhere in the world.GPS system can accurately locate various information of farmland, including agricultural machinery navigation, arable land, precision sowing, spraying and fertilizing[4-5].

Li et al.[6]designed an automatic navigation system based on RTK-DGPS positioning and double closed-loop steering control.The system used RTK-DGPS and AHRS500GA to provide position information and auxiliary correction information respectively to achieve accurate positioning.The test results showed that the average deviation of GPS positioning data was reduced to 0.031 m after correction.

1.2 Beidou positioning

Beidou satellite navigation system is a satellite navigation system independently constructed and operated by China.It is an important national space infrastructure that provides all-weather, all-time, high-precision positioning, navigation and timing services for global users[7].Beidou satellite system has been continuously developed since its establishment.At present, Beidou-1 system, Beidou-2 system and Beidou-3 system have been built in China[8].Beidou-2 positioning accuracy was 10 m[8].Beidou-2 Positioning Principle is as Figure 1.

Figure 1 Positioning principle of Beidou-2

At present, Beidou system has been widely used in agricultural production.Beidou navigation system was mainly used in agricultural machinery, such as agricultural machinery automatic navigation and driving system, agricultural machinery precision operation, and agricultural machinery operation supervision service.With the opening of Beidou-3 satellite navigation system, it provided more accurate and intelligent services for agriculture.

Ding et al.[9]designed a Beidou-based positioning controller.The locator used three low-precision Beidou positioning modules to place the data in an equilateral triangle with a length of 1 m to obtain the center point positioning data.The average static positioning accuracy of the Beidou receiving module was improved from 2.06 m to 1.50 m, and the dynamic positioning accuracy was improved to less than 0.78 m.Xiong et al.[10]designed an automatic navigation control system of orchard dispenser based on BDS.The system used RTK-BDS positioning, horizontal positioning accuracy can reach 0.8 m, and vertical positioning accuracy can reach 1.5 m.

1.3 Visual navigation

Visual navigation consists of three parts, including information acquisition, information processing and positioning tracking[11].

The working principle of visual navigation was that the sensor was used to perceive the surrounding environment information, and the obtained information was processed by visual image.The target area in the image and the obstacles in the navigation path were located, the mobile platform can independently move according to the target path[11-12].

Zhai et al.[13]proposed a tractor binocular vision navigation test method based on virtual reality technology.The experiments showed that when the tractor tracked the crop row at a speed not more than 2 m/s, the absolute value of the average position deviation was not more than 0.072 m, and the standard deviation of the position deviation was not more than 0.141 m.The absolute value of average heading deviation was not greater than 2.622°, and the standard deviation of heading deviation was not greater than 4.462°.

Li et al.[14]proposed a positioning algorithm of agricultural machinery omnidirectional vision positioning system(Figure 2).The absolute position of the sensor relative to the identification coordinate system was estimated according to the azimuth angle of the identification.The calculation model of omnidirectional vision imaging system was shown in Figure 2.

Figure 2 Omnidirectional visual imaging system

Tian et al.[15]proposed a tractor trajectory prediction method based on trinocular vision, and predicted future motion vector changes by coordinate change method and grey model method.

2 Navigation path regularization

Navigation path planning is the foundation of agricultural machinery navigation and tracking control.Under the complex and changeable agricultural environment, path planning can not only provide accurate guidance information for agricultural machinery, but also help to improve the quality of agricultural machinery operation[16].According to the mastery of environmental information, path planning can be divided into global path planning and local path planning[16].

2.1 Global path planning

Global path planning is a path planning method based on known environmental information[17].This method takes improving the quality of work and reducing the cost of work as the planning goal to seek a better job path.According to the different job scenarios, global path planning can be divided into two types: global coverage path planning and global point-to-point path planning.

2.1.1 Complete coverage path planning

The principle of full coverage path planning is that the control algorithm plans a path for the executor to ensure that the executor traverses all points in the workspace, while avoiding obstacles in the map[18].The full coverage path planning can completely cover the working area due to its moving path, which meets the requirements of agricultural field operations and provides a route for agricultural machinery.In addition, full coverage path planning technology can significantly improve the straightness of machinery and reduce the path repetition rate during agricultural machinery operation, so it has been widely used in the field of agriculture.

2.1.2 Global point-to-point path planning

Global point-to-point path planning is a feasible collision-free path from the starting point to the target point under the premise of known environmental information of the working area.At present, the mature algorithms are A*algorithm, ant colony optimization algorithm, genetic algorithm, simulated annealing algorithm and particle swarm optimization algorithm[16-19].Comparison of main algorithms for path planning is as shown in Table 1.

Table 1 Comparison of path planning algorithms

2.2 Local path planning

Local path planning is a real-time path planning that obtains the information of agricultural machinery and environmental obstacles according to the sensor information when the environmental information is completely or partially unknown[20].Due to the complexity and randomness of agricultural operation environment information, it is difficult for single sensor to complete the established task in complex agricultural operation environment.It is necessary to accurately perceive environmental information through multi-sensor fusion technology to improve the reliability of autonomous navigation.At present, the common local path planning in agricultural environment includes obstacle avoidance path planning and local tracking path planning[21].

2.2.1 Obstacle avoidance path planning

Obstacle avoidance path planning is to obtain the orientation and size information of obstacles in the working environment through sensors.After calculation and analysis, a safe and collision-free path is planned in real time and dynamically.

According to the size of obstacles and the width of agricultural machinery operation, Taïx et al.[21]proposed path planning methods for large obstacles and small obstacles.When avoiding large obstacles, the region segmentation algorithm was used to plan the route one by one for the segmented sub-regions, and when avoiding small obstacles, the original straight line path was locally adjusted to plan the route.

Foka et al.[22]proposed a predictive navigation model method combining probabilistic programming with obstacle avoidance.The navigation model can predict the optimal path to avoid obstacles.Liu et al.[23]proposed single obstacle avoidance algorithm and double/multiple obstacle avoidance algorithm(Figure 3)under known working conditions.

Figure 3 Single/double obstacle avoidance algorithm

Xi et al.[24]used the third-order Bezier curve optimization method to form a continuous smooth obstacle avoidance path of agricultural machinery(Figure 4).The linear control model of agricultural machinery motion was established by chain control theory, and the PI controller was used to compensate the angle.The simulation of the control method and the plowing experiment were carried out.The experimental results showed that the obstacle avoidance accuracy of Bezier curve was 5.21 cm, and the accuracy of agricultural machinery to continue driving along a straight line after obstacle avoidance was 1.98 cm.

Figure 4 Shortest tangent obstacle avoidance path

2.2.2 Local tracking path planning

Local tracking path planning is that agricultural machinery approaches the path planning from the current position to the preset path in the path tracking process[25-26].Pure tracking model, which is widely used in local tracking path planning, is a geometric method to simulate artificial driving process.

Yao et al.[25]proposed a path tracking method based on UWB wireless positioning(Figure 5).The experimental results showed that the steady-state deviation of linear tracking was 5.4-8.4 cm, the average steady-state deviation was 6.3 cm, and the average deviation of rectangular tracking was 20.6 cm, which can meet the operating requirements.

Figure 5 Pure tracking model

Figure 6 Ground turn path tracking control chart

Liu et al.[26]developed an automatic navigation system based on the platform of Revo ZP9500 high clearance sprayer.They designed linear path tracking control algorithm and curve path tracking control algorithm with position deviation and heading deviation as state variables(Figure 6).

3 Automatic steering technology

Automatic steering system of agricultural machinery by actuator can be divided into motor power steering system and electronic control hydraulic steering system[27].The main advantage of the motor control steering system is the small change of the agricultural machinery steering system.The disadvantage is that the manual automatic switching clutch mechanism is complex, the installation accuracy of the transmission mechanism is high, and the adaptability to different models is poor[27].The electronically controlled hydraulic steering system is mainly composed of hydraulic pump, electronically controlled proportional steering valve and steering controller.It is necessary to change the original hydraulic steering oil path and structure.The main advantages of the electronically controlled hydraulic steering system are high control power and fast response speed[28].The disadvantage is that the modification process is complex.

3.1 Electric power steering

The working principle of the motor power steering is that the electronic control unit calculates the desired angle of the steering wheel and the required steering torque by analyzing the information collected by the sensor, and controls the motor to provide the steering torque for the steering wheel to realize automatic steering[28, 30].

Bak et al.[29]designed an agricultural robot(Figure 7).

Figure 7 Electric automatic steering

The robot can work in the target area according to the planned path, and has the characteristics of reducing wheel sideslip and wheel wear.

Noguchi et al.[30]developed an automatic steering system based on a small gasoline tractor.He et al.[31]studied the automatic steering control of rice transplanter based on Jingguan PZ-60 rice transplanter with electric steering wheel as steering actuator(Figure 8).The research was to adopt system identification test and PID nested steering control algorithm.The test showed that the average absolute errors of steering angle tracking of the transplanter were 0.354° and 0.663°, respectively, when the transplanter ran straightly in flat and uneven paddy fields under the mud floor.

(a)Flat paddy fields

Yin et al.[32]developed an automatic driving system.when the operating speed was 3.6 km/h, the maximum heading angle was 7.86° and 2.48° under the two modes of remote control operation and automatic navigation, and the maximum root mean square error of lateral deviation was 7.47 cm and 4.66 cm(Figure 9).

Figure 9 Installation of steering motor

Yang et al.[33]designed an automatic steering mechanism and its electronic control system based on direct connection of DC motor and hydraulic steering gear(Figure 10).The system was to adopt the wheel angle closed-loop control algorithm and the wheel steering angle signal tracking test.

Figure 10 Steering drive schematic

3.2 Electrical controlled hydraulic steering

Electronically controlled hydraulic steering is sent by the control system signal to the hydraulic steering actuator, through the electro-hydraulic proportional directional valve to control the flow direction and velocity of hydraulic oil, so as to control the expansion of the steering hydraulic cylinder to drive the steering wheel deflection to the desired angle[34-35].

Zhang et al.[34]designed an electro-hydraulic steering control system ,which used steering simulator for steering motion analysis.Zhang et al.[35]designed a friction wheel steering drive(Figure 11).The device was to adopt simulation model verification test, genetic algorithm parameter optimizer performance comparison test and drive system performance test.The test result showed that the average absolute steady-state error of the 20° step response of the drive system was 0.197°.

(a)Lateral view (b)Stereogram

Li et al.[36]designed a steering controller based on double closed-loop control method(Figure 12).The controller was to adopt double closed-loop control method with angular velocity control as inner loop and steering angle control as outer loop[36].

Figure 12 Electronically controlled hydraulic steering system

4 The existing problems and future prospect

4.1 The existing problems

The key technologies of agricultural machinery unmanned driving system mainly include positioning technology, path planning technology and automatic steering technology[37].These key technologies have been relatively mature after years of development and have been widely used in agricultural production practice, but there are still many problems that need further study.

At present, the main problems existing in the key technologies of agricultural machinery unmanned driving system[38-41]include as follows.

(1)In terms of positioning technology: At present, the research of positioning technology mainly focuses on single positioning technology, and the research on combined positioning technology was relatively less[42-44].In the complex unstructured environment of farmland, the terrain, light and climate conditions are changeable.The single positioning technology has certain limitations, such as signal loss caused by occlusion, inevitable drift of data over time, and the influence of external light and shadow.

(2)In terms of path planning technology: At present, global path planning has the problems of repeated measurement planning of the same plot and low data utilization rate.Local path planning has a high degree of dependence on environmental information, which mainly has the problems of single sensor and insufficient median pose resolution ability in complex environments.

(3)In terms of automatic steering technology: automatic steering technology includes motor power steering and electronically controlled hydraulic steering.The motor power steering has the problems of complex switching clutch mechanism, high installation accuracy of transmission mechanism and poor adaptability to different models.Electronically controlled hydraulic steering has the problems of complex modification process, easy damage and difficult maintenance in harsh field operations.

4.2 Future prospect

In recent years, visual navigation and automatic obstacle avoidance have achieved remarkable results in theoretical algorithms, but they are still in the laboratory research stage.Automatic steering technology research started late, the existing automatic steering device integration was low, was not conducive to maintenance.Based on the above analysis, the development of agricultural machinery unmanned driving system of new ideas includes as follows.

(1)In the aspect of positioning technology : in the future, integrated navigation technology combining satellite positioning and machine vision can be studied to collect the real-time dynamic environmental information of farmland and improve the ability of autonomous decision-making and automatic obstacle avoidance of agricultural machinery.

(2)In terms of path planning technology: in the future, multi-sensor fusion technology can be used to accurately obtain environmental information and improve the analytical ability of the environment by improving the optimization algorithm.In addition, further research can be carried out in combination with the characteristics of agricultural machinery to increase the adaptability of planning path to agricultural machinery, so as to improve navigation accuracy and operation efficiency.

(3)In terms of automatic steering technology: in the future, it can develop better compatibility, simpler structure and easy maintenance of automatic steering system.