Unmanned loader driving path planning method

Abstract

The invention is suitable for the technical field of loaders, and provides an unmanned loader driving path planning method. According to the invention, the unmanned driving function of the articulated loader can be realized, and the dependence on a driver is avoided, so that the working efficiency of the articulated loader is obviously improved, and the occurrence rate of driving accidents is effectively reduced. Aiming at the characteristics of complex structural characteristics and severe working environment of an articulated loader, compared with a traditional single-target shortest path planning method, the multi-target path planning strategy under constraint conditions is adopted. The innovation can better adapt to the requirements of the articulated loader in a severe working environment, and the practicability and adaptability of path planning are improved. According to the method, a learning function is introduced, and the adaptive Kriging Monte Carlo dynamics proxy model is successfully constructed. The introduction of the model effectively solves the problem of long time consumption of simulation calculation, so that the acquisition of the vehicle characteristic parameters is quicker and more efficient.

Description

Technical Field

[0001] The invention belongs to the technical field of loaders, and in particular relates to a driving path planning method for an unmanned loader. Background Art

[0002] With the rapid development of science and technology, the trend of intelligentization and unmanned operation in the field of engineering machinery is becoming increasingly prominent. As an efficient engineering operation equipment, loaders are widely used in non-structural environments such as construction sites and mines. Unmanned loaders can realize the automated transportation of materials and uninterrupted operation throughout the day, but in non-structural environments, due to the unleveled road surface, the loader may have large posture changes during driving and even face the risk of tipping over. Therefore, planning a safe, efficient and executable path based on the non-structural environment map is the basis for ensuring the unmanned and autonomous operation of the loader.

[0003] In response to this challenge, path planning for loader autonomous driving is particularly important. Traditional methods usually simplify three-dimensional maps into two-dimensional maps, ignoring the actual impact of ground height fluctuations on loader driving, and thus require that the driving road surface must be a structured environment. For structured terrain, due to its flat terrain and regular boundaries, the 2D occupancy grid can effectively plan the optimal path, and the generated path only needs to consider avoiding obstacles. However, for unstructured environments, the boundaries may be curved or incomplete, the ground is uneven, and there are a large number of obstacles of different sizes and shapes. In addition, the terrain instability caused by soil composition and the impact of bad weather must be considered, which makes the path planning method in structured environments cannot be directly applied to unstructured environments.

[0004] At present, commonly used path planning methods include artificial potential field (APF), search-based methods, sampling-based methods, etc. However, they still have limitations when facing the complexity of unstructured environments. Especially in the case of large terrain undulations, the loader is prone to overturn due to instability. In view of this, the present invention proposes a driving path planning method for an unmanned loader. Summary of the invention

[0005] The purpose of the present invention is to provide a method for planning a driving path of an unmanned loader, aiming to solve the problems raised in the above-mentioned background technology.

[0006] The purpose of the present invention is achieved through the following technical solutions:

[0007] A method for planning a driving path for an unmanned loader comprises the following steps:

[0008] Step S1, constructing a dynamic adaptive model to provide dynamic characteristic data of the articulated loader for subsequent path planning;

[0009] Step S2, establishing a working environment: simplifying the vehicle working environment into a three-dimensional map to determine the initial position and target position of the path;

[0010] Step S3, constructing the objective function and determining the constraint conditions;

[0011] Step S4, using the planner to output the optimal path: extract the top 10 non-inferior solutions of the non-dominated sorting after the iteration of the multi-objective genetic algorithm, perform weighted comparison, and select the best solution as the optimal path output by the algorithm after comprehensive consideration;

[0012] Step S5: Check whether the output optimal path is reasonable through joint simulation; if it is reasonable, output the path; if it is unreasonable, return to step S4 for further optimization.

[0013] Furthermore, the specific process of step S1 is as follows:

[0014] Based on the Kriging model and Monte Carlo simulation, a learning function is introduced to construct an adaptive Kriging Monte Carlo dynamics proxy model. In the dynamics proxy model, the learning function index U(x) is associated with the lower confidence boundary lcb function for optimization. lcb is defined as:

[0015]

[0016] In the formula, lcb(x) represents the confidence lower limit function, represents the model prediction mean, Represents the model prediction deviation; take b as 2, calculate and minimize lcb to 0, then U(x) and b are equal, corresponding to the model error probability Φ(-2) = 0.023. After training, the dynamic agent model provides the dynamic characteristic data of the articulated loader for path planning.

[0017] Furthermore, the specific process of step S2 is as follows:

[0018] The three-dimensional environment data is simplified into coordinate values ​​x, y, and z, and the coordinate values ​​are stored in the map environment matrix; the design variables are the coordinates of each path point (x i ,y i ,z i ), set the starting point S coordinate to (x 1 ,y 1 ,z 1 ), the coordinate of the end point G is (x n ,y n ,z n ); The map is evenly divided into grids in the xy plane, and the remaining trajectory path points between the starting point and the end point are discretely distributed at the grid vertices, expressed as (x i ,y i,z i ); Finally, connect the points in order to form a complete trajectory.

[0019] Furthermore, the specific process of step S3 is as follows:

[0020] Step S31, constructing an objective function;

[0021] The sum of the spatial distances between adjacent path points is used as the path length cost, that is, the path travel length cost function F L Defined as:

[0022]

[0023] In the formula, n represents the number of path segments in the path, l i Represents the spatial distance between the path points at both ends of the i-th path, specifically expressed as:

[0024]

[0025] The total travel time of each path segment is used as the path time cost, that is, the path travel time cost function F T Defined as:

[0026]

[0027] In the formula, n represents the number of path points contained in a path, t i It represents the travel time between two adjacent path points in the i-th segment of the path, which can be specifically expressed as:

[0028]

[0029] In the formula, l i represents the spatial distance between two adjacent path points in the i-th path, v i and v i-1 Represents the safe speed value calculated by the dynamics agent model at two path points;

[0030] The smoothness cost F of the path is measured by evaluating the change in the angle between adjacent straight line segments on each path. S :

[0031]

[0032] In the formula, n represents the number of path points contained in a path. and For q i-2 ,q i-1 ,q i These three consecutive path points form two adjacent vectors:

[0033]

[0034] In the formula, x i is the x coordinate of the i-th path, y i is the y coordinate of the i-th path, z i is the z coordinate of the i-th path, x i-1 is the x coordinate of the i-1th path, y i-1 is the y coordinate of the i-1th path segment, z i-1 is the z coordinate of the i-1th path, x i-2 is the x coordinate of the i-2th path, y i-2 is the y coordinate of the i-2th path, z i-2 is the z coordinate of the i-2th path;

[0035] Step S32, determining constraint conditions;

[0036] The constraints include the maximum articulation angle, the maximum pitch angle, and the maximum roll angle. The articulation angle constraint is ±30°, the pitch angle constraint is ±30°, and the roll angle constraint is ±20°.

[0037] Waypoint q i (x i ,y i ,z i ) has a ground contact point q il (x il ,y il ,z il ), the obstacle crossing point is q ir (x ir ,y ir ,z ir ), then the roll angle φ i for:

[0038]

[0039] The maximum roll angle constraint is:

[0040] φ i ≤φ max

[0041] In the formula, φ ma x is the maximum roll angle, L is the body width of the articulated loader;

[0042] Assume that the path point q i The angle between the path segment and the x-axis in the xy plane is β, then q il and q ir In the xy plane, q i Calculated and expressed as q il (x i-Δx,y i -Δy) and q ir (x i +Δx,y i +Δy), where Δx and Δy are calculated as follows:

[0043]

[0044] On a certain planned path segment, the starting point q i-1 and end point q i Composition vector The pitch angle θ of the path segment is represented by the angle of the vector in space i :

[0045]

[0046] The maximum pitch angle constraint is:

[0047] θ i ≤θ max

[0048] In the formula, θ max is the maximum pitch angle;

[0049] The i-th path point q i The hinge angle is defined as ψ i , for three adjacent path points q i-2 ,q i-1 ,q i , the hinge angle at the path point is described by the projection angle of the corresponding vectors of the adjacent two end path segments on the horizontal plane;

[0050]

[0051] The maximum articulation angle constraint is:

[0052] ψ i ≤ψ max

[0053] In the formula, ψ max is the maximum articulation angle;

[0054] Based on the constructed objective function and constraints, the mathematical model of path planning is obtained:

[0055] minF(x)=[F L (x),F L (x),F S (x)] T

[0056]

[0057] Where n represents the number of path points contained in a path, F(X) represents the multi-objective function, X represents the set of decision variables, and T represents the inversion operation.

[0058] Furthermore, the process of step S4 is as follows:

[0059] A multi-objective genetic algorithm is used to iterate the algorithm in combination with the objective function and constraints. After the iteration is completed, the top 10 non-inferior solutions are extracted from the obtained non-dominated sorting results. Then, the difference between the same objective function values ​​in different non-inferior solutions is calculated, and a weight is assigned to each difference. By comprehensively considering the difference and weight of the objective function values, an optimal solution is selected as the optimal path output by the algorithm. Finally, the path point coordinates and speed values ​​of the optimal path are stored.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] 1. The present invention can realize the unmanned driving function of the articulated loader, get rid of the dependence on the driver, thereby significantly improving the working efficiency of the articulated loader and effectively reducing the incidence of driving accidents.

[0062] 2. In view of the complex structural characteristics and harsh working environment of articulated loaders, the present invention adopts a multi-objective path planning strategy under constraints compared to the traditional single-objective shortest path planning method. This innovation can better meet the needs of articulated loaders in harsh working environments and improve the practicality and adaptability of path planning.

[0063] 3. The present invention introduces a learning function and successfully constructs an adaptive Kriging Monte Carlo dynamics proxy model. The introduction of this model effectively solves the problem of long simulation calculation time, making the acquisition of vehicle characteristic parameters faster and more efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 The figure is a flow chart of path planning of the method of the present invention.

[0065] Figure 2 Environment graph for path planning.

[0066] Figure 3 Schematic diagram of the roll angle of an articulated loader when overcoming an obstacle; (a) is the change of the front body of the articulated loader when overcoming an obstacle, (b) is the change of the rear body of the articulated loader when overcoming an obstacle, and (c) is the calculation of the roll angle.

[0067] Figure 4 This is the path planning framework diagram.

[0068] Figure 5 It is an environment map in the form of a three-dimensional grid.

[0069] Figure 6 Schematic diagram of joint simulation.

[0070] Figure 7 The results of trajectory planning for flat terrain; (a) is the trajectory visualization diagram, (b) is the length convergence curve, (c) is the time convergence curve, and (d) is the smoothness convergence curve.

[0071] Figure 8 These are the results of trajectory planning for conventional obstacle terrain; (a) is the trajectory visualization diagram, (b) is the length convergence curve, (c) is the time convergence curve, and (d) is the smoothness convergence curve.

[0072] Fig. 9 The trajectory planning results for extreme obstacle terrain; among them, (a) is the trajectory intuitive diagram, (b) is the length convergence curve, (c) is the time convergence curve, and (d) is the smoothness convergence curve. DETAILED DESCRIPTION

[0073] In order to have a clearer understanding of the technical features, purposes and beneficial effects of the present invention, the technical solution of the present invention is now described in detail below, but it should not be construed as limiting the applicable scope of the present invention.

[0074] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0075] An embodiment of the present invention provides a method for planning a driving path for an unmanned loader, the flow chart of which is as follows: Figure 1 As shown, the method comprises the following steps:

[0076] Step S1, constructing a dynamic adaptive model to provide dynamic characteristic data of the articulated loader for subsequent path planning;

[0077] Step S2, establishing a working environment: simplifying the vehicle working environment into a three-dimensional map to determine the initial position and target position of the path;

[0078] Step S3, constructing the objective function and determining the constraint conditions;

[0079] Step S4, using the planner to output the optimal path: extract the top 10 non-inferior solutions of the non-dominated sorting after the iteration of the multi-objective genetic algorithm, perform weighted comparison, and select the best solution as the optimal path output by the algorithm after comprehensive consideration;

[0080] Step S5: Check whether the output optimal path is reasonable through joint simulation; if it is reasonable, output the path; if it is unreasonable, return to step S4 for further optimization.

[0081] The details are as follows:

[0082] (i) Step S1: construct a dynamic adaptive model to provide dynamic characteristic data of the articulated loader for subsequent path planning.

[0083] Based on the Kriging model and Monte Carlo simulation, a learning function is introduced to construct an adaptive Kriging Monte Carlo dynamics surrogate model (AK-MCS). In the dynamics surrogate model, the learning function index U(x) is associated with the lower confidence boundary lcb function for optimization. lcb is defined as:

[0084]

[0085] In the formula, lcb(x) represents the confidence lower limit function, represents the model prediction mean, Represents the model prediction deviation; take b as 2, calculate and minimize lcb to 0, then U(x) and b are equal, corresponding to the model error probability Φ(-2) = 0.023. After training, the dynamic agent model provides the dynamic characteristic data of the articulated loader for path planning.

[0086] (ii) Step S2, establishing a working environment: simplifying the vehicle working environment into a three-dimensional map to determine the initial position and target position of the path.

[0087] like Figure 2 As shown in the figure, the three-dimensional environment data is simplified into coordinate values ​​x, y, z, and the coordinate values ​​are stored in the map environment matrix; the design variables are the coordinates of each path point (x i ,y i ,z i ), set the starting point S coordinate to (x 1 ,y 1 ,z 1 ), the coordinate of the end point G is (x n ,y n ,z n ); The map is evenly divided into grids in the xy plane, and the remaining trajectory path points between the starting point and the end point are discretely distributed at the grid vertices, expressed as (x i ,y i ,z i ); Finally, connect the points in order to form a complete trajectory.

[0088] (III) Step S3: constructing the objective function and determining the constraints.

[0089] (1) Constructing the objective function;

[0090] Since the planned path consists of a series of three-dimensional coordinate points, the sum of the spatial distances between adjacent path points is used as the path length cost, that is, the path travel length cost function F L Defined as:

[0091]

[0092] In the formula, n represents the number of path segments in the path, l i Represents the spatial distance between the path points at both ends of the i-th path, specifically expressed as:

[0093]

[0094] The total travel time of each path segment is used as the path time cost, that is, the path travel time cost function F T Defined as:

[0095]

[0096] In the formula, n represents the number of path points contained in a path, t i It represents the travel time between two adjacent path points in the i-th segment of the path, which can be specifically expressed as:

[0097]

[0098] In the formula, l i represents the spatial distance between two adjacent path points in the i-th path, v i and v i-1 Represents the safe speed value calculated by the dynamic agent model at two path points.

[0099] The core of the path smoothing objective function is to ensure that the vehicle drives smoothly to facilitate path tracking, while avoiding safety accidents caused by excessive steering angles. Since the motion path is composed of a series of spatial coordinate points, the line between each two points forms a straight line segment. In order to ensure smooth driving, it is necessary to maintain a smooth transition between straight lines as much as possible. Therefore, the path smoothness cost F can be measured by evaluating the change in the angle between adjacent straight lines on each path. S :

[0100]

[0101] In the formula, n represents the number of path points contained in a path. and For q i-2 ,q i-1 ,q i These three consecutive path points form two adjacent vectors:

[0102]

[0103] In the formula, x i is the x coordinate of the i-th path, y i is the y coordinate of the i-th path, z iis the z coordinate of the i-th path, x i-1 is the x coordinate of the i-1th path, y i-1 is the y coordinate of the i-1th path segment, z i-1 is the z coordinate of the i-1th path, x i-2 is the x coordinate of the i-2th path, y i-2 is the y coordinate of the i-2th path, z i-2 is the z coordinate of the i-2th path;

[0104] (2) Determine the constraints;

[0105] In view of the characteristics of non-structural environments and in combination with the characteristics of engineering vehicles, the present invention incorporates the constraints of the maximum articulation angle, the maximum pitch angle, and the maximum roll angle into the constraint conditions. Among them, the articulation angle constraint is intended to ensure that the turning angle of the loader is strictly controlled within a safe range during the turning process to avoid the risks brought by excessive turning. The present invention sets it as ±30°. In addition, the pitch angle constraint reflects the maximum angle that the loader can reach when overcoming obstacles. It is a key indicator for evaluating the off-road performance of the vehicle. The present invention selects ±30°. It is worth noting that different loaders also have different critical roll angles for overcoming obstacles under different speeds, pitch angles, and articulation angles. The present invention sets it as ±20°.

[0106] Assume that the vehicle is at path point p i When crossing an obstacle, the right front wheel contacts the obstacle first, causing the front vehicle to tilt to the left. Figure 3 At this time, the posture of the rear vehicle body is as follows Figure 3 As shown in (b), under the action of the swing bridge, the entire vehicle body tilts but the rear axle does not tilt. Therefore, in the constraint conditions of the present invention, the roll angle mentioned specifically refers to the tilt angle of the front vehicle body, which is represented by φ. Figure 3 As shown in (c), in addition, the path point q i (x i ,y i ,z i ) has a ground contact point q il (x il ,y il ,z il ), the obstacle crossing point is q ir (x ir ,y ir ,z ir ), then the roll angle φ i for:

[0107]

[0108] The maximum roll angle constraint is:

[0109] φ i ≤φmax

[0110] In the formula, φ max is the maximum roll angle, L is the body width of the articulated loader;

[0111] Assume that the path point q i The angle between the path segment and the x-axis in the xy plane is β, then q il and q ir In the xy plane, q i Calculated and expressed as q il (x i -Δx,y i -Δy) and q ir (x i +Δx,y i +Δy), where Δx and Δy are calculated as follows:

[0112]

[0113]

[0114] On a certain planned path segment, the starting point q i-1 and end point q i Composition vector The pitch angle θ of the path segment is represented by the angle of the vector in space i :

[0115]

[0116] The maximum pitch angle constraint is:

[0117] θ i ≤θ max

[0118] In the formula, θ max is the maximum pitch angle;

[0119] The i-th path point q i The hinge angle is defined as ψ i , for three adjacent path points q i-2 ,q i-1 ,q i , the hinge angle at the path point is described by the projection angle of the corresponding vectors of the adjacent two end path segments on the horizontal plane;

[0120]

[0121] The maximum articulation angle constraint is:

[0122] ψ i ≤ψ max

[0123] In the formula, ψ max is the maximum articulation angle;

[0124] Based on the constructed objective function and constraints, the mathematical model of path planning is obtained:

[0125] minF(x)=[F L (x),F T (x),F S (x)] T

[0126]

[0127] Where n represents the number of path points contained in a path, F(X) represents the multi-objective function, X represents the set of decision variables, and T represents the inversion operation.

[0128] (iv) Step S4, using the planner to output the optimal path: extract the top 10 non-inferior solutions of the non-dominated sorting after the iteration of the multi-objective genetic algorithm, perform weighted comparison, and select the best solution as the optimal path output by the algorithm after comprehensive consideration;

[0129] See also Figure 4 , using a multi-objective genetic algorithm (NSGA-Ⅱ), combined with the objective function and constraints, the algorithm is iterated; after the iteration is completed, the top 10 non-inferior solutions are extracted from the obtained non-dominated sorting results; then the difference between the same objective function values ​​in different non-inferior solutions is calculated, and a weight is assigned to each difference; by comprehensively considering the difference and weight of the objective function values, an optimal solution is selected as the optimal path output by the algorithm; finally, the path point coordinates and speed values ​​of the optimal path are stored.

[0130] (V) Step S5: Check whether the output optimal path is reasonable through joint simulation; if it is reasonable, output the path; if it is unreasonable, return to step S4 for further optimization.

[0131] In order to test the performance of the method proposed in the present invention, we built a joint simulation platform based on MATLAB / Simulink and RecurDyn software.

[0132] First, EarthSculptor is used to draw and generate a terrain map of a three-dimensional unstructured environment, which is then imported into MATLAB in the form of a point cloud for simulation, as shown in Figure 5 shown.

[0133] Next, Simulink and RecurDyn are used for joint simulation to dynamically verify the planned path. Figure 6 shown.

[0134] In a relatively flat road environment, such as Figure 7 As shown in (a), the generated trajectory always travels along the road with less undulations, and when approaching an obstacle, the off-road characteristics of the articulated loader are fully utilized to successfully climb over the obstacle and reach the end point. Figure 7 As shown in (b) to (d), with the increase in the number of iterations, the path length, path time, and path smoothness gradually converged, and convergence was completed before the iteration number 40. Specifically, the path length converged to within 131.129m, and the minimum value converged to 114.735m; the path time converged to within 26.090s, and the minimum value converged to 23.008s; the path smoothness converged to within 18°, and the minimum value converged to 4.483°.

[0135] This result shows that the proposed method can effectively adapt to complex terrain conditions and plan a driving path for the articulated loader that meets its characteristics. As the number of iterations increases, the convergence of path length, time and smoothness indicates the stability and reliability of the algorithm.

[0136] Path planning is performed for vehicle driving conditions where the starting point is far away from conventional obstacles and the end point is surrounded by positive and negative obstacles, such as Figure 8 As shown in (a). As the number of iterations increases, Figure 8 As shown in (b) to (d), the path length, path time, and path smoothness gradually converged, and convergence was completed before the number of iterations reached 50. Specifically, the path length converged to within 118.254m, and the minimum value converged to 112.999m; the path time converged to within 38.033s, and the minimum value converged to 29.495s; the path smoothness converged to within 11°, and the minimum value converged to 4.142°.

[0137] This result shows that the proposed method can effectively realize path planning and optimization when considering complex driving scenarios, and has good convergence performance. The proposed method shows good stability and reliability when dealing with complex situations during vehicle driving. The effectiveness and practicality of the algorithm are further verified through the convergence analysis of indicators such as path length, time and smoothness, providing a feasible solution for the application of articulated loaders in actual work.

[0138] Path planning is performed for vehicle driving conditions where the starting point is far from an extreme obstacle and the end point is near the obstacle. Fig. 9 As shown in (a), the path planning successfully avoids extreme obstacles and reaches the end point smoothly after crossing regular obstacles. As the number of iterations increases, Fig. 9As shown in (b) to (d), the path length, path time, and path smoothness gradually converged and completed convergence before the iteration number 30. Specifically, the path length converged to within 107.448m, and the minimum value converged to 105.787m; the path time converged to within 27.207s, and the minimum value converged to 23.078s; the path smoothness converged to within 12°, and the minimum value converged to 2.599°.

[0139] This result shows that the proposed method successfully avoids extreme obstacles and selects a relatively short, short-time and smoother path. The improvement of path smoothness helps to reduce the bumps and shakes of the vehicle during driving, and improves the driving comfort and stability. At the same time, the effective control of path length and path time also reduces the energy consumption during driving and improves driving efficiency. These results further verify the feasibility and practicality of the proposed method in complex terrain.

[0140] The above are only preferred embodiments of the present invention. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These should also be regarded as the protection scope of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for planning a driving path for an unmanned loader, characterized in that: The following steps are involved: Step S1, constructing a dynamic adaptive model to provide dynamic characteristic data of the articulated loader for subsequent path planning; Step S2, establishing a working environment: simplifying the vehicle working environment into a three-dimensional map to determine the initial position and target position of the path; Step S3, constructing the objective function and determining the constraint conditions; Step S4, using the planner to output the optimal path: extract the top 10 non-inferior solutions of the non-dominated sorting after the iteration of the multi-objective genetic algorithm, perform weighted comparison, and select the best solution as the optimal path output by the algorithm after comprehensive consideration; Step S5: Check whether the output optimal path is reasonable through joint simulation; if it is reasonable, output the path; if it is unreasonable, return to step S4 for further optimization.

2. The method for planning a driving path for an unmanned loader according to claim 1, characterized in that: The specific process of step S1 is as follows: Based on the Kriging model and Monte Carlo simulation, a learning function is introduced to construct an adaptive Kriging Monte Carlo dynamics proxy model. In the dynamics proxy model, the learning function index U(x) is associated with the lower confidence boundary function lcb for optimization. lcb is defined as: In the formula, lcb(x) represents the confidence lower limit function, represents the model prediction mean, Represents the model prediction deviation; take b as 2, calculate and minimize lcb to 0, then U(x) and b are equal, corresponding to the model error probability Φ(-2) = 0.

023. After training, the dynamic agent model provides the dynamic characteristic data of the articulated loader for path planning.

3. The method for planning a driving path for an unmanned loader according to claim 2, characterized in that: The specific process of step S2 is as follows: The three-dimensional environment data is simplified into coordinate values ​​x, y, and z, and the coordinate values ​​are stored in the map environment matrix; the design variables are the coordinates of each path point (x i ,y i ,z i ), set the starting point S coordinates to (x1, y1, z1), and the end point G coordinates to (x n ,y n ,z n ); The map is evenly divided into grids in the xy plane, and the remaining trajectory path points between the starting point and the end point are discretely distributed at the grid vertices, expressed as (x i ,y i ,z i ); Finally, connect the points in order to form a complete trajectory.

4. The method for planning a driving path for an unmanned loader according to claim 3, characterized in that: The specific process of step S3 is as follows: Step S31, constructing an objective function; The sum of the spatial distances between adjacent path points is used as the path length cost, that is, the path travel length cost function F L Defined as: In the formula, n represents the number of path segments in the path, l i Represents the spatial distance between the path points at both ends of the i-th path, specifically expressed as: The total travel time of each path segment is used as the path time cost, that is, the path travel time cost function F T Defined as: In the formula, n represents the number of path points contained in a path, t i It represents the travel time between two adjacent path points in the i-th segment of the path, which can be specifically expressed as: In the formula, l i represents the spatial distance between two adjacent path points in the i-th path, v i and v i-1 Represents the safe speed value calculated by the dynamics agent model at two path points; The smoothness cost F of the path is measured by evaluating the change in the angle between adjacent straight line segments on each path. S : In the formula, n represents the number of path points contained in a path. and For q i-2 ,q i-1 ,q i These three consecutive path points form two adjacent vectors: In the formula, x i is the x coordinate of the i-th path, y i is the y coordinate of the i-th path, z i is the z coordinate of the i-th path, x i-1 is the x coordinate of the i-1th path, y i-1 is the y coordinate of the i-1th path segment, z i-1 is the z coordinate of the i-1th path, x i-2 is the x coordinate of the i-2th path, y i-2 is the y coordinate of the i-2th path, z i-2 is the z coordinate of the i-2th path; Step S32, determining constraint conditions; The constraints include the maximum articulation angle, maximum pitch angle and maximum roll angle constraints. The articulation angle constraint is ±30°, the pitch angle constraint is ±30°, and the roll angle constraint is ±20°. Waypoint q i (x i ,y i ,z i ) has a ground contact point q il (x il ,y il ,z il ), the obstacle crossing point is q ir (x ir ,y ir ,z ir ), then the roll angle φ i for: The maximum roll angle constraint is: f i ≤φ max In the formula, φ max is the maximum roll angle, L is the body width of the articulated loader; Assume that the path point q i The angle between the path segment and the x-axis in the xy plane is β, then q il and q ir In the xy plane, q i Calculated and expressed as q il (x i -Δx,y i -Δy) and q ir (x i +Δx,y i +Δy), where Δx and Δy are calculated as follows: On a certain planned path segment, the starting point q i-1 and end point q i Composition vector The pitch angle θ of the path segment is represented by the angle of the vector in space i : The maximum pitch angle constraint is: i i ≤θ max In the formula, θ max is the maximum pitch angle; The i-th path point q i The hinge angle is defined as ψ i , for three adjacent path points q i-2 ,q i-1 ,q i , the hinge angle at the path point is described by the projection angle of the corresponding vectors of the adjacent two end path segments on the horizontal plane; The maximum articulation angle constraint is: ψ i ≤ψ max In the formula, ψ max is the maximum articulation angle; Based on the constructed objective function and constraints, the mathematical model of path planning is obtained: minF(X)=[F L (x),F T (x),F S (x)] T Where n represents the number of path points contained in a path, F(X) represents the multi-objective function, X represents the set of decision variables, and T represents the inversion operation.

5. The method for planning a driving path for an unmanned loader according to claim 4, characterized in that: The process of step S4 is as follows: A multi-objective genetic algorithm is used to iterate the algorithm in combination with the objective function and constraints. After the iteration is completed, the top 10 non-inferior solutions are extracted from the non-dominated sorting results. Then, the difference between the same objective function values ​​in different non-inferior solutions is calculated, and a weight is assigned to each difference. By comprehensively considering the difference and weight of the objective function values, an optimal solution is selected as the optimal path output by the algorithm. Finally, the path point coordinates and speed values ​​of the optimal path are stored.

Images