Full-coverage path planning method for high-clearance sprayer under irregular convex field block based on particle swarm optimization

Abstract

The invention discloses a full-coverage path planning method for a high-clearance sprayer under an irregular convex field block based on a particle swarm algorithm. The method is characterized by comprising the following steps: acquiring boundary data of a farmland area to obtain a boundary contour of an irregular convex field block; secondly, on the basis of a traditional U-shaped turning mode, an included angle between each operation row and the boundary of the field is introduced, and the connection path principle between the operation rows is elaborated in detail; the method comprises the following steps of: performing parallel line offset on a straight line passing through a central point of an irregular convex region, generating a full-coverage operation row with a random direction angle, giving different weights of three parameters, namely total length of an agricultural machine operation path, turning times and operation coverage rate, optimizing the direction angle of the operation row through an improved particle swarm optimization algorithm, and finally obtaining a full-coverage operation row with a random direction angle. According to the scheme, the energy consumption of the unmanned agricultural machine is effectively reduced, the working efficiency is improved, and the basic requirement of agricultural production for full-coverage path planning can be met.

Description

Technical Field

[0001] The invention belongs to the technical field of agricultural control machinery, and in particular relates to an irregular convex field block full coverage path planning algorithm based on an improved particle swarm algorithm. Background Art

[0002] Agricultural machinery navigation path planning is one of the basic and important supporting technologies in smart agriculture. The navigation path generated by combining agronomic requirements and mechanical performance can significantly improve the standardization and regularization of agricultural production methods and promote the efficient production of smart agriculture. One of the main challenges in achieving agronomic operations such as sowing, spraying, tilling, harvesting and crushing is to find a path that best covers the entire farmland. This process is usually called complete coverage path planning (CCPP). As one of the "enabling technologies" for automatic navigation and driving operations of agricultural machinery, the quality of the path planned for global coverage operations largely determines the operating efficiency, operation quality and energy consumption of the unit.

[0003] There are usually two methods for full coverage path planning of regular boundary plots: First, based on the method of baseline translation, a baseline parallel to the plot edge (usually the longer side) is used to translate at equal intervals according to the width of the operation to generate multiple straight line operation paths until the entire plot is covered, and then the appropriate headland turning strategy is selected according to the actual operation situation. This method only requires pre-measurement of the coordinates of the two points of the generated baseline, and the calculation process is simple, but the headland turning requires the use of sensors for headland detection or manual assisted turning. Second, the local coordinate system method is used to pre-measure the coordinates of the four vertices of the plot, calculate all navigation path feature points according to the predetermined operation mode and the headland turning sequence, and then generate a complete operation path that traverses the entire plot according to the operation sequence.

[0004] Particle swarm optimization (PSO) is an evolutionary computation technique proposed by Dr. Eberhart and Dr. Kennedy in 1995, which originated from the study of bird flock predation behavior. The algorithm was originally inspired by the regularity of bird flock activities, and then a simplified model was established using swarm intelligence. Based on the observation of animal flock activities, the particle swarm algorithm uses the information sharing of individuals in the group to make the movement of the entire group evolve from disorder to order in the problem-solving space, thereby obtaining the optimal solution. The advantage of the PSO algorithm is that it is simple and easy to implement and does not require many parameter adjustments. It has been widely used in function optimization, neural network training, fuzzy system control and other application fields of genetic algorithms. Summary of the invention

[0005] Based on the above analysis, this paper proposes a full coverage path planning algorithm for irregular convex fields based on an improved particle swarm algorithm. At the same time, an environmental simulation model is established based on irregular convex fields to accurately obtain the boundary data and area data of the fields, which improves the universality of the algorithm. The simulation results of actual fields show that the algorithm can reduce the driving distance of unmanned agricultural machinery, reduce the consumption of energy and fuel, and obtain good path planning results.

[0006] The technical solution of the present invention is:

[0007] A method for planning a full coverage path of a high-gap sprayer under irregular convex fields based on a particle swarm algorithm comprises the following steps:

[0008] Step 1: Establish an Ackerman steering model based on the steering structure of the high-clearance sprayer;

[0009] Step 2: Establish the kinematic model of the high clearance sprayer;

[0010] Step 3: Collect and organize farmland boundary data, convert the data into coordinate rows, and establish a simulation map of irregular field boundaries;

[0011] Step 4: Based on the farmland boundary characteristics and the angle between the boundary row of the field and the operation row of the unmanned agricultural machine, select an appropriate operation row path connection strategy;

[0012] Step 5: Call the irregular convex field full coverage planning algorithm to generate the initial operation row;

[0013] Step 6: Based on the direction angle of the initial operation line generated in step 5, the inertia weight and learning factor of the particle swarm algorithm are improved, and the new fitness function is modified to obtain the direction angle of the operation line with the shortest path and the least energy consumption, and generate the optimal full coverage planning path.

[0014] Further, the Ackerman steering model described in step 1 is:

[0015]

[0016] Where V is the speed of the sprayer, V1:V4 are the speeds of the four tires of the sprayer, W is the width of the vehicle body suspension, α and β are the front and rear wheel steering angles, and a and b are the front and rear vehicle body lengths.

[0017] Further, the kinematic model of the high ground clearance sprayer in step 2 is specifically expressed as follows:

[0018]

[0019] Where X and Y are the longitudinal and lateral positions of the high clearance sprayer in the global coordinate system. is the heading angle of the high clearance sprayer, v x is the longitudinal speed of the high gap sprayer, δ f is the front wheel turning angle of the high clearance sprayer, and l is the body length of the high clearance sprayer.

[0020] Furthermore, the irregular field farmland boundary modeling process in step 3 is as follows:

[0021] The geometric method is selected as the modeling method, which can not only accurately describe the real environment, but also meet the needs of full coverage path planning; first, the data of the working area is collected through the GIS system, including the longitude and latitude coordinates of the boundaries of irregular convex fields, and they are classified and stored according to the field number; secondly, the stored data is linearly expanded. At this time, the data is the longitude and latitude coordinates in the geodetic coordinate system. After converting it into the Cartesian coordinates in the geocentric coordinate system, the first data is introduced as the starting point of the boundary, and the Cartesian coordinates are converted into the local northeast celestial coordinate system. Finally, a polygon object is created through the coordinate data stored in the list, and a simulation map of the irregular convex field is fitted. The simulation modeling of irregular convex fields is realized through the above process.

[0022] Furthermore, the operation line path connection strategy in step 4 specifically includes:

[0023] The reserved field turning area has an angle with the extension direction of the working line, that is, the straight line where the vehicle is driving on the working line has a certain angle with the boundary straight line of the field turning area; the OM direction is the driving direction of the agricultural machinery in the working line area, and the MN direction indicates both the driving direction of the agricultural machinery in the field turning area and the extension direction of the plot boundary;

[0024] (1): When and When the angle γ is 90°, the connection path between the working rows is The reserved distance R of the land, where |P1P2|=(s*L)-2*R, The OM direction is the driving direction of the agricultural machinery in the working row area, and the MN direction indicates both the driving direction of the agricultural machinery in the turning area at the end of the field and the extension direction of the boundary of the plot; γ is the angle between the OM and MN lines; is the first arc turning path, |P1P2| is the straight path when turning at the end of the field, is the second arc turning path, R is the minimum turning radius of the unmanned agricultural machinery, s is the number of fields crossed, and L is the width of the working row;

[0025] (2): When and When the angle γ is less than 90°, the connection path between the working rows is In the formula, |P5P1|=2Rtanγ, in the formula, R+Rsinγ+|P2P3|cosγ+R-Rsinγ=l1, the reserved distance at the end of the field is |R+Rsinγ|, in the formula, l1 is the distance between the two working rows, and |P2P3|=(l1-2R) / cosγ, |P5P1| is the first straight line path, is the first arc path, |P2P3| is the second straight line path, is the second arc path;

[0026] (3): When and When the angle γ is greater than 90°, the connection path between the working rows is In the formula |P2P3|=(l1-2R) / cosγ, |P4P5|=|2Rtanγ|; |P4P5| is the second straight line path, where s is the number of fields crossed, and the above-mentioned angles are all in radians.

[0027] Further, the step 5 is specifically as follows:

[0028] After obtaining the environmental simulation model of the irregular convex field, the irregular convex field full coverage planning algorithm is called to generate the direction angle for the operation row. When the particle swarm algorithm is run to initialize the particles, the pos value of the particle will be randomly generated, that is, the k value representing the direction angle of the operation row; when the k of the straight line passing through the center point P0 (x0, y0) of the irregular convex area is specified, the equation of the straight line passing through the P0 point can be expressed as:

[0029] y-y0=k0(x-x0) (1)

[0030] y-y0=k i (x-x0),i∈(1,2,3,...,N) (2)

[0031] Where N is the number of particles, k0 is the tangent value of the angle between the working direction at point P0 and the horizontal direction; k i It is the tangent value of the angle between the working direction and the horizontal direction, and -90° <arctan(k)<90°。

[0032] Further, the step 6 specifically includes:

[0033] The improved particle swarm algorithm is used to optimize the direction angle of the operation row; for a single operation area of ​​an irregular convex field, the unmanned agricultural machinery full coverage path planning traversal model based on the improved PSO algorithm is solved as follows:

[0034] (1) Input the parameters of the irregular convex field operation area and the unmanned agricultural machinery operation parameters to perform environmental modeling;

[0035] (2) Setting algorithm parameters, generating and initializing the search group, randomly generating the initial position and velocity of the particle group, and generating the initial direction angle of the operation line;

[0036] (3) Calculate the objective function value of each particle in the group: ① Calculate the total length of the path traversed by the unmanned agricultural machinery each time (total objective function); ② If the objective function value of a particle is less than the individual extreme value of the particle, set the individual extreme value as the position of the particle, and then update the global extreme value of the group;

[0037] (4) Update the change speed of boundary points between dimensions: Setting the inertia weight and learning factor to decrease nonlinearly with the increase of iteration number is conducive to reducing the algorithm optimization time and effectively taking into account the global search degree and fine search degree of the algorithm;

[0038] (5) Compare the fitness value of each particle position after the update with the fitness value of the best position it has visited. If the fitness value of the current position is smaller, it is retained as the best position; update the best position of the entire particle group;

[0039] (6) Determine whether the current number of iterations has reached the maximum number of iterations; if so, the algorithm terminates and obtains the optimal k value, which is the optimal operation row direction angle, and outputs the optimal operation path; otherwise, go to step (4);

[0040] The traditional PSO algorithm is difficult to limit the size of particles. Using the particle swarm algorithm to optimize the direction angle of the operation line is prone to fall into the local optimum, and the planning result is poor. The improvements are as follows:

[0041] 1): Improve inertia weight

[0042] The inertia weight w reflects the influence of the previous generation of particles on the velocity of the current generation of particles and is an important parameter of the PSO algorithm. In order to find the global optimal solution, the dynamically changing w has a better optimization effect than the static value. A nonlinear decreasing inertia weight is used. As the number of iterations increases, the inertia weight w decreases continuously, so that the particles have a stronger local convergence ability in the later stage. The expression is:

[0043]

[0044] Where T max is the maximum number of iterations of the algorithm, k is the current number of iterations; where w is the inertia weight, which reflects the influence of the current velocity of the particle on the next update; w max is the maximum value of the inertia weight, w min is the minimum value of the inertia weight; cos() is the cosine function; tan() is the tangent function; Tmax is the maximum number of iterations of the algorithm; k is the current number of iterations;

[0045] 2): Improved learning factor

[0046] The learning factor determines the probability of a particle approaching the best position visited by itself and the group. The algorithm sets the learning factor as the trigonometric function of the proposed nonlinear decreasing inertia weight, which is expressed as:

[0047] c1=sinw (10)

[0048] c2=1-sinw (11)

[0049] Where w is the inertia weight, which reflects the influence of the current speed of the particle on the next update; c1 is the individual acceleration factor, which reflects the dependence of the particle on its own historical best position; c2 is the social acceleration factor, which reflects the dependence of the particle on the best position of the group, reflecting the group effect; sin is the sine function.

[0050] The present invention has the following technical effects:

[0051] The feasibility of the algorithm was verified through simulation experiments. Compared with the traditional method, the algorithm proposed in this paper has a stable full coverage traversal path spacing, accurate coverage range, reduced number of turns, and greatly reduced energy consumption. In four actual fields of different sizes, compared with the traditional method, the total traversal distance of the method of the present invention in fields 1, 3, and 4 was reduced by 9.01m, 23.25m, 8.71m, and 14.32m, and the number of turns was reduced by 5.3% to 61.5%, respectively, and the additional coverage ratio was reduced by 0.45% to 2.56%, effectively reducing duplication and omission phenomena. Therefore, the irregular convex field full coverage traversal path planning algorithm can effectively reduce energy loss and increase the sustainable operation time of unmanned agricultural machinery while ensuring full coverage of the operation.

[0052] The algorithm of this study is applicable to most irregular convex fields including rectangular fields and convex polygonal boundary fields. However, in a few convex fields, the algorithm of the present invention still needs to be improved. For example, in field 3, the total path length is reduced by 23.25m, but the additional coverage ratio is increased. At the same time, for irregular concave boundary fields, further in-depth research is needed. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A flowchart of environmental modeling is presented;

[0054] Figure 2 A schematic diagram showing the connection between the paths of the operating rows;

[0055] Figure 3 Generate schematic diagrams for adjacent job rows;

[0056] Figure 4 It is the connection diagram of adjacent operation rows;

[0057] Figure 5 Flowchart for optimizing the direction angle of the job row for the particle swarm algorithm;

[0058] Figure 6 There are four actual plots; (a) Plot 1; (b) Plot 2; (c) Plot 3; (d) Plot 4;

[0059] Figure 7 The full coverage planning algorithm for irregular convex fields is Figure 8 The full coverage path map generated after running in the four plots; (a) Plot 1; (b) Plot 2; (c) Plot 3; (d) Plot 4;

[0060] Figure 8 Iterative relationship diagram of the direction angle of the working row; (a) plot 1; (b) plot 2; (c) plot 3; (d) plot 4; DETAILED DESCRIPTION

[0061] Figure 1 The flow chart of environmental modeling is presented, and the process of implementing field data simulation modeling in the MATLAB Simulink environment after obtaining the longitude and latitude data of the field boundary using Google Earth is described in detail, thereby generating an environmental simulation model of actual farmland data. The specific steps are as follows: First, the longitude and latitude data of the farmland boundary are sorted into a list form; then, the Insert_points function is called to expand the data and increase the data density through interpolation; then, the numerical data of the map data is converted into a vector data type by calling the importMap function and the importEdge function; then, the longlat2meter function is used to further convert the data into a rectangular coordinate system. After that, the coordinate data is corroded by calling the corrosion function, and the standard form of the irregular area of ​​the farmland is generated by the getPoly function. Finally, the Polygon function is used to draw the simulation model of the farmland. The above process realizes the efficient conversion from geographic data to simulation model, providing technical support for the simulation modeling of the actual farmland environment.

[0062] Figure 2The figure shows the schematic diagram of the path connection between the working rows. When the field is divided into multiple working rows, the figure describes three path connection methods of the high-clearance unmanned sprayer between the working rows. In the figure, represents the driving direction of the agricultural machinery in the working row area, and represents the extension direction of the boundary of the plot. When the intersection angle with is greater than 90°, less than 90°, and equal to 90°, the unmanned sprayer adopts different path connection methods. For detailed description of the specific content, please refer to Section 2.2 of this article.

[0063] Figure 3 The diagram is a schematic diagram of the generation of adjacent operation rows. It illustrates the method of generating other operation rows through a straight line passing through the center point of an irregular field. The equation of the straight line passing through the center point of an irregular convex area is y-y0=k0(x-x0). As can be seen from the figure, a straight line y is generated along the positive direction of the y axis and the negative direction of the y axis. Ⅰ and Ⅱ .

[0064] Figure 4 This is a diagram of the connection between adjacent rows. It describes the connection between rows directly, which is related to the angle between L1 and A1B1. In the figure, the curved boundary MN is replaced by the straight boundary A1B1, and the angle between L1 and A1B1 corresponds to Figure 4 In and The intersection angles correspond to three different path connection methods. Figure 4 The three connection modes described in .

[0065] Figure 5 The flowchart of the particle swarm algorithm for optimizing the direction angle of the operation line. The specific process of the algorithm is as follows: First, the particle swarm algorithm randomly assigns an initial value to each particle in the group. Each initial value corresponds to a fitness function value, and the size of its fitness function value is calculated by formula (12). Then, it is judged whether the current fitness value meets the iteration condition; if the preset number of iterations is not reached, the speed and position of the particle are updated by the speed and position update formula. After the update, the fitness value of each particle's new position is compared with the fitness value of the historical best position it has visited. If the fitness value of the new position is better, it is used as the current best position of the particle, and the global best position of the entire particle group is updated at the same time.

[0066] After each iteration, the algorithm determines whether the current number of iterations has reached the maximum number of iterations. If so, the algorithm terminates and outputs the direction angle corresponding to the global optimal fitness value, which is the optimal direction angle of the optimized working line. This process effectively optimizes the direction angle of the working line and provides accurate parameter support for the path planning of the unmanned sprayer.

[0067] Figure 6There are four actual plots of land. All four plots are in Songjiang District, Shanghai, with an area of ​​802.22m 2 、1823.54m 2 、9670.31m 2 and 3655.12m 2 Plot 1 is a relatively small ordinary quadrilateral plot, Plot 2 is a relatively small ordinary rectangular plot, Plot 3 is an irregular convex plot with typical characteristics, and Plot 4 is an irregular convex plot with typical characteristics.

[0068] Figure 7 The full coverage planning algorithm for irregular convex fields is Figure 8 The full coverage path map generated after running in the four plots. In plots 1 and 2, the basic parameters of the agricultural machinery are set to R = 0.8m for the turning radius and L = 1.8m for the working width. In plots 3 and 4, the turning radius of the agricultural machinery is set to R = 2.2m and the working width is set to L = 5m.

[0069] Figure 8 It is an iterative relationship diagram of the direction angle of the work row, which reflects the optimization process of the total path length as the number of iterations of the particle swarm algorithm increases. It can be seen from the figure that the total path length gradually decreases with the increase of the number of iterations, indicating that the algorithm has a good optimization effect. When the rate of decrease of the total path length tends to be stable, the iterative process reaches the termination condition, indicating that the particle swarm algorithm has obtained the optimal solution. The above results verify the effectiveness and convergence of the algorithm in optimizing the direction angle of the work row and path planning.

[0070] 2) List the specific values ​​of the parameters of the algorithm part in the specific implementation method using the operating parameters of the four-wheel independent self-steering electric drive unmanned agricultural machinery

[19] , vehicle wheelbase L = 1580mm, operating width d = 2500mm, turning radius R = 2200mm. Considering the adaptability of the algorithm, in plots 1 and 2, the turning radius of the vehicle is set to 800mm and the operating radius is set to 900mm. In the particle swarm algorithm, the initial particle population size P is set size =200, number of iterations T max =500, maximum weight w max =1.2, minimum weight w min =0.2, k1=0.8, k2=0.1, k3=0.1.

[0071] 1. Iteration number T max =500, maximum weight w max =1.2, minimum weight w min =0.2. The nonlinear decreasing inertia weight used in this study is:

[0072]

[0073] 2. Minimum weight w min =0.2, k1=0.8, k2=0.1, k3=0.1. The fitness function used in this study is:

[0074] f fitness =(k1l+k2n t +k3ε)

[0075] 3. Vehicle wheelbase L = 1580mm, working width d = 2500mm, turning radius R = 2200mm. Figure 2 In Section 2.2, |P1P2|=(s*L)-2*R, This type of parameters is used in |P2P3|=(l1-2R) / cosγ, etc., where s is the number of fields spanned.

[0076] When the algorithm is executed, the field boundary information of the farmland is first collected, and the field contour simulation map is obtained after data expansion and corrosion. At the same time, the configuration parameters of the unmanned agricultural machinery are used as standard input, and the irregular convex field full coverage planning algorithm is called to generate the initial operation row direction angle, which is then passed into the particle swarm algorithm as the parameter to be optimized for optimization. Finally, the optimal operation row direction angle is obtained through algorithm iteration and optimization, thereby generating a full coverage operation path for irregular convex areas. A full coverage path planning algorithm for irregular convex fields is proposed. The algorithm can plan the operation path with the shortest travel distance, the highest coverage rate and the least number of turns for unmanned agricultural machinery under the premise that the operation area is fully covered, providing theoretical support for the development of path planning technology for unmanned agricultural machinery.

[0077] The first step of full coverage path planning is to model the environment. At present, the modeling methods of plot boundaries are all based on regular plots, and the traditional modeling methods are less applicable to irregular plots. Although the general particle swarm algorithm is relatively simple to implement and does not require complex derivative calculations, the algorithm converges slowly and requires a long number of iterations. At the same time, different initial parameters may lead to different results. This paper improves it in these two aspects. First, the number of GPS points in the plot boundary is increased. Through the dense equidistant dot microelement method, when the number of points is large enough, it can be fitted into an irregular polygon boundary. Secondly, the inertia weight and learning factor of the particle swarm algorithm are improved so that in the next iteration experience, it has more experience with its own learning weight, and the weight from other particles is larger, which has a better convergence effect.

[0078] The feasibility of the algorithm was verified through simulation experiments. Compared with the traditional method, the algorithm proposed in this invention has a stable full coverage traversal path spacing, accurate coverage range, reduced number of turns, and greatly reduced energy consumption. In four actual fields of different sizes, compared with the traditional method, the total traversal distance of this method in fields 1, 3, and 4 was reduced by 9.01m, 23.25m, 8.71m, and 14.32m, and the reduction rate of the number of turns was reduced by 5.3% to 61.5%, respectively, and the additional coverage ratio was reduced by 0.45% to 2.56%, effectively reducing duplication and omission phenomena. Therefore, the irregular convex field full coverage traversal path planning algorithm can effectively reduce energy loss and increase the sustainable operation time of unmanned agricultural machinery while ensuring full coverage of the operation.

[0079] The algorithm of this study is applicable to most irregular convex fields including rectangular fields and convex polygonal boundary fields. However, in a few convex fields, the algorithm of the present invention still needs to be improved. For example, in field 3, the total path length is reduced by 23.25m, but the additional coverage ratio is increased. At the same time, for irregular concave boundary fields, further in-depth research is needed.

Claims

1. A method for planning full coverage path of high-gap sprayer under irregular convex fields based on particle swarm algorithm, characterized in that: The following steps are involved: Step 1: Establish an Ackerman steering model based on the steering structure of the high-clearance sprayer; Step 2: Establish the kinematic model of the high clearance sprayer; Step 3: Collect and organize farmland boundary data, convert the data into coordinate rows, and establish a simulation map of irregular field boundaries; Step 4: Based on the farmland boundary characteristics and the angle between the boundary row of the field and the operation row of the unmanned agricultural machine, select an appropriate operation row path connection strategy; Step 5: Call the irregular convex field full coverage planning algorithm to generate the initial operation row; Step 6: Based on the direction angle of the initial operation line generated in step 5, the inertia weight and learning factor of the particle swarm algorithm are improved, and the new fitness function is modified to obtain the direction angle of the operation line with the shortest path and the least energy consumption, and generate the optimal full coverage planning path.

2. A method according to claim 1, characterized in that The Ackerman steering model described in step 1 is: Where V is the speed of the sprayer, V1:V4 are the speeds of the four tires of the sprayer, W is the width of the vehicle body suspension, α and β are the front and rear wheel steering angles, and a and b are the front and rear vehicle body lengths.

3. A method according to claim 1, characterized in that The kinematic model of the high clearance sprayer described in step 2 is specifically expressed as follows: Where X and Y are the longitudinal and lateral positions of the high clearance sprayer in the global coordinate system. is the heading angle of the high clearance sprayer, v x is the longitudinal speed of the high gap sprayer, δ f is the front wheel turning angle of the high clearance sprayer, and l is the body length of the high clearance sprayer.

4. A method according to claim 1, characterized in that The irregular field farmland boundary modeling process in step 3 is as follows: The geometric method is selected as the modeling method, which can not only accurately describe the real environment, but also meet the needs of full coverage path planning; first, the data of the working area is collected through the GIS system, including the longitude and latitude coordinates of the boundaries of irregular convex fields, and they are classified and stored according to the field number; secondly, the stored data is linearly expanded. At this time, the data is the longitude and latitude coordinates in the geodetic coordinate system. After converting it into the Cartesian coordinates in the geocentric coordinate system, the first data is introduced as the starting point of the boundary, and the Cartesian coordinates are converted into the local northeast celestial coordinate system. Finally, a polygon object is created through the coordinate data stored in the list, and a simulation map of the irregular convex field is fitted. The simulation modeling of irregular convex fields is realized through the above process.

5. A method according to claim 1, characterized in that: The operation path connection strategy in step 4 specifically includes: The reserved field turning area has an angle with the extension direction of the working line, that is, the straight line where the vehicle is driving on the working line has a certain angle with the boundary straight line of the field turning area; the OM direction is the driving direction of the agricultural machinery in the working line area, and the MN direction indicates both the driving direction of the agricultural machinery in the field turning area and the extension direction of the plot boundary; (1): When and When the angle γ is 90°, the connection path between the working rows is The reserved distance R of the land, where |P1P2|=(s*L)-2*R, The OM direction is the driving direction of the agricultural machinery in the working row area, and the MN direction indicates both the driving direction of the agricultural machinery in the turning area at the end of the field and the extension direction of the boundary of the plot; γ is the angle between the OM and MN lines; is the first arc turning path, |P1P2| is the straight path when turning at the end of the field, is the second arc turning path, R is the minimum turning radius of the unmanned agricultural machinery, s is the number of fields crossed, and L is the width of the working row; (2): When and When the angle γ is less than 90°, the connection path between the working rows is In the formula, |P5P1|=2Rtanγ, in the formula, R+Rsinγ+|P2P3|cosγ+R-Rsinγ=l1, the reserved distance at the end of the field is |R+Rsinγ|, in the formula, l1 is the distance between the two working rows, and |P2P3|=(l1-2R) / cosγ, |P5P1| is the first straight line path, is the first arc path, |P2P3| is the second straight line path, is the second arc path; (3): When and When the angle γ is greater than 90°, the connection path between the working rows is In the formula |P2P3|=(l1-2R) / cosγ, |P4P5|=|2R tanγ|; |P4P5| is the second straight line path, where s is the number of fields spanned, and the above-mentioned angles are all in radians.

6. A method according to claim 1, characterized in that The step 5 is specifically as follows: After obtaining the environmental simulation model of the irregular convex field, the irregular convex field full coverage planning algorithm is called to generate the direction angle for the operation row. When the particle swarm algorithm is run to initialize the particles, the pos value of the particle will be randomly generated, that is, the k value representing the direction angle of the operation row; when the k of the straight line passing through the center point P0 (x0, y0) of the irregular convex area is specified, the equation of the straight line passing through the P0 point can be expressed as: y-y0=k0(x-x0) (1) y-y0=k i (x−x0),i∈(1,2,3,…,N) (2) Where N is the number of particles, k0 is the tangent value of the angle between the working direction at point P0 and the horizontal direction; k i For The tangent value of the angle between the industry direction and the horizontal direction, and -90° <arctan(k)<90°。 7. A method according to claim 1, characterized in that The step 6 specifically includes: The improved particle swarm algorithm is used to optimize the direction angle of the operation row; for a single operation area of ​​an irregular convex field, the unmanned agricultural machinery full coverage path planning traversal model based on the improved PSO algorithm is solved as follows: (1) Input the parameters of the irregular convex field operation area and the unmanned agricultural machinery operation parameters to perform environmental modeling; (2) Setting algorithm parameters, generating and initializing the search group, randomly generating the initial position and velocity of the particle group, and generating the initial direction angle of the operation line; (3) Calculate the objective function value of each particle in the group; (4) Update the change speed of boundary points between dimensions: Setting the inertia weight and learning factor to decrease nonlinearly with the increase of iteration number is conducive to reducing the algorithm optimization time and effectively taking into account the global search degree and fine search degree of the algorithm; (5) Compare the fitness value of each particle position after the update with the fitness value of the best position it has visited. If the fitness value of the current position is smaller, it is retained as the best position; update the best position of the entire particle group; (6) Determine whether the current number of iterations has reached the maximum number of iterations; if so, the algorithm terminates and obtains the optimal k value, which is the optimal operation row direction angle, and outputs the optimal operation path; otherwise, go to step (4); The traditional PSO algorithm is difficult to limit the size of particles. Using the particle swarm algorithm to optimize the direction angle of the operation line is prone to fall into the local optimum, and the planning result is poor. The improvements are as follows: 1): Improve inertia weight The inertia weight w reflects the influence of the previous generation of particles on the velocity of the current generation of particles and is an important parameter of the PSO algorithm. In order to find the global optimal solution, the dynamically changing w has a better optimization effect than the static value. A nonlinear decreasing inertia weight is used. As the number of iterations increases, the inertia weight w decreases continuously, so that the particles have a stronger local convergence ability in the later stage. The expression is: Where T max is the maximum number of iterations of the algorithm, k is the current number of iterations; where w is the inertia weight, which reflects the influence of the current velocity of the particle on the next update; w max is the maximum value of the inertia weight, w min is the minimum value of the inertia weight; cos() is the cosine function; tan() is the tangent function; T max is the maximum number of iterations of the algorithm; k is the current number of iterations; 2): Improved learning factor The learning factor determines the probability of a particle approaching the best position visited by itself and the group. The algorithm sets the learning factor as the trigonometric function of the proposed nonlinear decreasing inertia weight, which is expressed as: c1=sin w (10) c2=1-sin w (11) Where w is the inertia weight, which reflects the influence of the current speed of the particle on the next update; c1 is the individual acceleration factor, which reflects the dependence of the particle on its own historical best position; c2 is the social acceleration factor, which reflects the dependence of the particle on the best position of the group, reflecting the group effect; sin is the sine function.

Images