Unmanned truck path planning method for mining area charging yard

Abstract

The invention relates to an unmanned truck path planning method for a mining area charging yard, belongs to the technical field of unmanned driving path planning, and aims at the characteristics of irregular shape, complex and diversified obstacles and the like of the mining area charging yard on the basis of a constructed planning scene artificial potential energy field. A heuristic search path planning algorithm is improved in cost evaluation function design, rapid generation of a vehicle driving path in a mining area charging yard is realized, the time of a vehicle from a starting node to a target node is reduced, the trafficability of the vehicle in a complex and changeable environment is improved, and the vehicle driving efficiency is improved. And a basis is provided for the automatic driving technology of the unmanned vehicle in the mining area.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned driving path planning, and in particular to an unmanned truck path planning method for a mining area loading yard. Background Art

[0002] With the intelligent development of production and operation processes in open-pit mines, unmanned vehicles are increasingly used in transportation operations in mines. Unmanned production operations in open-pit mines are mainly divided into three parts: mining, transportation, and unloading. Due to the large size of mining transport vehicles, their total mass and inertia increase when fully loaded, which increases braking loss and blind spots. At the same time, due to the inflexible steering and long braking distance of mining transport vehicles, it is difficult for human drivers driving vehicles in the loading yard to accurately judge the steering and U-turn capabilities, maintain a safe speed and boundary distance, which brings safety risks to production operations in the mining loading yard.

[0003] In addition, in order to improve the efficiency of unmanned mining and transportation in the mine loading yard, designing a planning method for quickly generating the driving trajectory of unmanned vehicles has become an important part of the application of unmanned driving technology in mining areas. In the open-pit mine loading yard, unmanned trucks need to perform accurate path planning in a complex unstructured environment. The driving trajectory of the transport vehicle is restricted by the vehicle's parking posture, U-turn and gear shift position, etc. It must be safe and efficient enough to ensure the normal operation of the mine production operation.

[0004] In an urbanized road environment, the roads are continuous and flat, so it is easier to implement unmanned driving path planning. However, the environment of a mining loading yard is highly complex and diverse. The site shape is often irregular and full of various potential obstacles and uncertainties. This complex environmental condition poses a severe technical challenge to the real-time path planning of unmanned trucks. Therefore, how to perform real-time path planning for unmanned trucks in mining loading yards to ensure the safe driving and efficient operation of unmanned trucks is an important issue that needs to be solved urgently.

[0005] In recent years, many path planning methods for autonomous vehicles have been proposed. They use heuristic search to complete global path planning from the starting point to the end point. However, they are mainly aimed at open areas, and do not take into account the dynamic changes in the size and shape of the site and the loading points of unmanned transport vehicles during the production operation of the mining area. In this scenario, traditional heuristic search will lead to low completeness of the planned path, slow algorithm solution speed, and inability to dynamically adapt to various loading operation scenarios, and cannot guarantee the applicability of its algorithm in the mining area loading yard. Summary of the invention

[0006] In view of the above problems, the present invention provides an unmanned truck path planning method for a mining area loading yard. In view of the characteristics of the irregular shape and complex and diverse obstacles in the mining area loading yard, on the basis of the constructed artificial potential field of the planning scene, the path planning algorithm of heuristic search is improved in the design of the cost evaluation function, so as to realize the rapid generation of the vehicle driving path in the mining area loading yard, reduce the time for the vehicle to travel from the starting node to the target node, improve the passability of the vehicle in the complex and changeable environment, and provide a basis for the automatic driving technology of unmanned vehicles in the mining area.

[0007] The present invention provides an unmanned truck path planning method for a mining area loading yard, including:

[0008] Step S1, obtaining the full-area map of the mining area loading yard where the vehicle is located, and performing rasterization processing to obtain a rasterized map;

[0009] Step S2, defining the obstacle area based on the rasterized map;

[0010] Step S3, constructing an artificial potential field for static obstacles;

[0011] Step S4, determining the empty lanes and heavy lanes in the mining area loading yard;

[0012] On the artificial potential field of static obstacles, add the influence of the empty lanes and heavy lanes on the artificial potential field to obtain the empty lane potential energy and heavy lane potential energy at any position respectively;

[0013] Step S5, determining the trajectory planning task in the mining area loading yard;

[0014] Step S6, based on the artificial potential field of static obstacles, empty lane potential energy and heavy lane potential energy at any position and the potential energy trajectory planning task, obtaining the complete potential field of the trajectory planning space;

[0015] Step S7, based on the complete potential field of the trajectory planning space, designing the driving cost to obtain the updated driving cost; designing the heuristic cost based on the number of gear shifts of the path to obtain the updated heuristic cost;

[0016] Based on the updated driving cost and the updated heuristic cost, obtaining the total cost;

[0017] Based on the total cost, improving the Hybrid A * algorithm to obtain the improved Hybrid A * algorithm;

[0018] Step S8, using the improved Hybrid A * algorithm to obtain the optimal feasible smooth path of the unmanned vehicle in the mining area loading yard from the starting point to the destination.

[0019] Optionally, the specific steps of defining the obstacle area based on the rasterized map include:

[0020] Step S21, extracting the irregular obstacle coverage area from the rasterized map as obstacle area 1;

[0021] Step S22, using an image processing algorithm to identify the boundary contour of the object in the image, and marking the area outside the boundary contour of the object in the image as obstacle area 2;

[0022] Step S23: merge obstacle area 1 and obstacle area 2 to define an obstacle area.

[0023] Optionally, the static obstacle artificial potential energy field is a static obstacle artificial potential energy field at any position, and the expression is:

[0024]

[0025] Among them, E static (P) is the artificial potential field of the static obstacle, E max is the upper limit of the maximum potential energy in the artificial potential energy field; is the distance between position P and obstacle point ξ; Ω is the set of all static obstacle points in the loading yard; ξ is an obstacle point in the set Ω; m static is the power function factor of the artificial potential energy field of the static obstacle; d0 is the escape distance, and the potential energy beyond this distance is 0.

[0026] Optionally, the empty lane is a lane for unloaded vehicles to enter the loading yard;

[0027] The heavy-load lane is the lane where the heavily loaded vehicle leaves the loading yard after completing the loading task.

[0028] Optionally, the first trajectory planning task is a parking entry trajectory planning task starting from an empty lane and stopping at a loading point;

[0029] The second trajectory planning task is to start from the loading stop and merge into the heavy lane to leave the site.

[0030] Optionally, the complete potential energy field of the trajectory planning space is obtained, expressed as:

[0031] E plan =k1E static +S(T)·k2E empty -S(T)·k3E load

[0032] Among them, k1 is the weight coefficient of static obstacle potential energy; k2 is the weight coefficient of empty lane potential energy; k3 is the weight coefficient of heavy lane potential energy, T is the time step, E static(P) is the potential energy of the empty lane, E load (P) is the potential energy of the heavy lane, and S(T) is the trajectory planning task at time step T.

[0033] Optionally, the specific steps of using the improved Hybrid A * algorithm to obtain the optimal feasible smooth path of the driverless vehicle in the mining area loading yard from the starting point to the destination include:

[0034] The specific steps of using the improved Hybrid A * algorithm to obtain the optimal feasible smooth path of the driverless vehicle in the mining area loading yard from the starting point to the destination include:

[0035] Step S8.1: Initialize the open list and the closed list;

[0036] Step S8.2: Let a = 1. When a = 1, it represents the first round of iteration;

[0037] Step S8.3: Check the status of the open list in the a-th round of iteration. If it is empty, terminate the calculation and determine that no valid path to the target node n t can be found;

[0038] If the open list in the a-th round of iteration is not empty, select the node a with the lowest cost value in the open list of the a-th round of iteration, and update the closed list and the updated list in the a-th round of iteration to obtain the updated closed list B a and the updated open list C a ; Use the updated closed list B a as the closed list in the (a + 1)-th round of iteration, and use the updated open list C a as the open list in the (a + 1)-th round of iteration;

[0039] Step S8.4: Determine whether the node a is the target node n t ;

[0040] If so, proceed to the next step; if not, let a = a + 1 and return to step S8.3;

[0041] Step S8.5: Use the Reed-Shepp curve generation algorithm to obtain the smooth path l t from the node a to the target node n n,a , and determine whether it collides with an obstacle. If so, let a = a + 1 and return to step S8.3;

[0042] If not, proceed to the next step;

[0043] Step S8.6: Based on the smooth path l from the node a to the target node n t ​n,a With node a as the center, the improved Hybrid A * algorithm is used to expand the child nodes, obtaining the expanded node a and several child nodes n c,a ; c = 1, 2, 3... D, where D represents the total number of child nodes of node a;

[0044] Step S8.7, Traverse all child nodes n c,a to obtain the final second-update open list C″ of node a a and proceed to the next step;

[0045] Step S8.8, Determine whether a is greater than or equal to A, where A represents the total number of iterations. If so, obtain the final second-update open list; based on the final second-update open list, obtain the optimal feasible smooth path. If not, set a = a + 1 and return to Step S8.3.

[0046] Optionally, the specific steps to obtain the final second-update open list C″ of node a a include:

[0047] Step S8.71, Set c = 1. When c = 1, it represents the first child node of node a;

[0048] Step S8.72, Determine whether the child node n c,A is in the updated closed list B a If not, proceed to the next step; if so, skip this child node without processing, set c = c + 1, and return to Step S8.82;

[0049] Step S8.73, Check whether the child node n c,a already exists in the updated open list C a If not, add it to the updated open list C a to obtain the second-update open list C′ a ; if it already exists, compare the travel cost of the child node n c,a with that of the previous child node n c-1,a If it is less, obtain the updated child node n′ c,a ;

[0050] Use the updated child node n′ c,a to update the updated open list C a to obtain the second-update open list C′ a and proceed to the next step; if it is greater than or equal to, set c = c + 1 and return to Step S8.82;

[0051] Step S8.74, Determine whether c is greater than or equal to D. If so, obtain the final second-update open list C″ of node a a, proceed to the next step; otherwise, let c = c + 1 and return to step S8.82;

[0052] Optionally, the specific steps of obtaining the updated closed list B a and the updated open list C a include:

[0053] Select the node a with the lowest cost value in the open list of the a-th iteration, transfer it to the closed list of the a-th iteration to obtain the updated closed list B a , and remove it from the open list to obtain the updated open list C a ; Use the updated closed list B a as the closed list of the (a + 1)-th iteration, and use the updated open list C a as the open list of the (a + 1)-th iteration.

[0054] Optionally, the specific steps of determining whether the smooth path l t from node a to the target node n n,a collides with an obstacle include:

[0055] Uniformly sample the smooth path l t from node a to the target node n n,a to obtain multiple sampling points n i,a ;

[0056] Map the vehicle body coverage range corresponding to each sampling point n i,a to the grid cells, and detect whether the vehicle pose at each sampling point n i,a collides with an obstacle.

[0057] Optionally, the specific steps of determining whether the child node n c,a is in the closed list B a include:

[0058] If the position coordinates and yaw angle of the child node n c,a are the same as the nodes in the closed list B a , it is the same node, indicating that the child node n c,a is in the closed list B a list, otherwise it is not in the closed list.

[0059] Compared with the prior art, the present invention has at least the following beneficial effects:

[0060] (1) The present invention fully considers the business requirements and scene characteristics of the mining area scene. By introducing an artificial potential field to quantify the relative relationship between trajectory points and obstacle points, it accurately describes the influence of the obstacle point set on the planned trajectory and avoids the risk of collision between the planned path and obstacles.

[0061] (2) Based on the constructed artificial potential field of static obstacles, the present invention adds the influence of the main road on the artificial potential field to ensure that the trajectory planning algorithm of the mining area loading yard fully considers the requirements of actual driving rules, and realizes the complete spatial modeling of the trajectory planning of the mining area loading yard.

[0062] (3) By improving the traditional Hybrid algorithm, the present invention incorporates the total potential energy of the path in the potential field and the number of gear shifts of the path into the design of the driving cost and the heuristic cost, so as to limit the number of gear shifts of the generated path result by the Hybrid algorithm, ensure that the obtained path result can meet the accurate tracking requirements of the lower-level control module as much as possible, and finally use the improved Hybrid algorithm to realize the rapid generation of the vehicle driving path in the mining area loading yard, improve the passability of the driverless truck in the complex and changeable environment, and provide a basis for the autonomous driving technology of the driverless truck in the mining area. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The drawings are only for the purpose of showing specific embodiments and are not considered to be a limitation of the present invention.

[0064] Figure 1 It is a schematic diagram of the process of the path planning of the driverless truck for the mining area loading yard in the embodiment of the present invention;

[0065] Figure 2 It is a schematic diagram of using the improved Hybrid A * algorithm to obtain the optimal feasible smooth path of the driverless vehicle in the mining area loading yard from the starting point to the destination;

[0066] Figure 3 (a)-(b) are the schematic diagram of static obstacles and the schematic diagram of the potential field of static obstacles in the embodiment of the present invention;

[0067] Figure 4 (a)-(c) are the schematic diagrams of the empty lane and the heavy lane and the corresponding enabling fields in the embodiment of the present invention;

[0068] Figure 5 (a)-(b) are the schematic diagram of the mining area loading yard and the schematic diagram of the trajectory planning space of the mining area loading yard in the embodiment of the present invention;

[0069] Figure 6 (a)-(b) are the schematic diagrams of the node expansion of the Hybrid A algorithm and the node expansion of the comparative example A * algorithm in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] To more clearly understand the above objects, features, and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. In addition, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0071] A specific embodiment of the present invention, such as Figure 1-6 , discloses an unmanned truck path planning method for a mining area loading yard, and the specific implementation steps are as follows:

[0072] Step S1: Obtain the full-area map of the mining area loading yard where the vehicle is located, and perform rasterization processing to obtain a rasterized map; the rasterized map includes a plurality of raster cells, and an attribute value is assigned to each raster cell to represent the corresponding semantic information;

[0073] Use an image processing algorithm to extract the boundaries of the raster cells corresponding to irregular obstacles;

[0074] Extract key nodes from the boundaries of the raster cells corresponding to irregular obstacles;

[0075] Use the key nodes as the vertices of the graph and the connection relationship between the nodes as the edges to construct an obstacle connectivity graph;

[0076] Obtain obstacle nodes based on the obstacle connectivity graph;

[0077] Optionally, the obstacle connectivity graph is used to describe obstacle information;

[0078] Optionally, the semantic information in step S1 includes drivable areas, obstacles, and / or loading points;

[0079] The key nodes include inflection points or intersection points;

[0080] Each raster cell represents a region of a fixed size;

[0081] Step S2: Define an obstacle area based on the rasterized map;

[0082] Optionally, the specific steps for defining an obstacle area based on the rasterized map include:

[0083] Step S21: Extract the area covered by irregular obstacles from the rasterized map as obstacle area one;

[0084] Step S22: Use an image processing algorithm to identify the object boundary contour in the image, and mark the area outside the object boundary contour in the image as obstacle area two;

[0085] Step S23: Merge obstacle area one and obstacle area two and define them as the obstacle area;

[0086] Step S3: Construct a static obstacle artificial potential field;

[0087] It can be understood that the static obstacle artificial potential field is used to quantify the relative relationship between path nodes / trajectory points and obstacle points;

[0088] Optionally, the static obstacle artificial potential field is the static obstacle artificial potential field at any position, and the expression is:

[0089]

[0090] where, E static (P) is the static obstacle artificial potential field, E max is the upper limit of the maximum potential energy in the artificial potential field. When the result obtained by summation is greater than this value, the summation is no longer continued to reduce the calculation pressure; is the distance between position P and obstacle point ξ; Ω is the set of all static obstacle points in the loading yard; ξ is an obstacle point in set Ω; m static is the power function factor of the static obstacle artificial potential field, which determines the convergence speed when increases; d0 is the escape distance, and the potential energy at positions beyond this distance is 0.

[0091] Step S4: Determine the empty lane and heavy lane of the loading yard in the mining area;

[0092] Add the influence of the empty lane and heavy lane on the artificial potential field to the static obstacle artificial potential field to obtain the empty lane potential energy and heavy lane potential energy at any position respectively;

[0093] It can be understood that the empty lane is the lane for vehicles to enter the loading yard empty;

[0094] The heavy lane is the lane for vehicles to leave the loading yard heavily after completing the loading task;

[0095] Optionally, the expressions for the empty lane potential energy and heavy lane potential energy at any position are respectively:

[0096]

[0097] where, E static (P) is the empty lane potential energy, E load (P) is the heavy lane potential energy, Г empty is the set of all trajectory points on the empty lane, Г load is the set of all trajectory points on the heavy lane; is set Г empty or set Гload One of the trajectory points in is the distance between the P-th position and the trajectory point ; m empty is the power function factor of the empty lane potential energy, m load is the power function factor of the heavy lane potential energy, which determines the convergence speed when increases; E limit is the maximum potential energy upper limit of "empty lane potential energy" and "heavy lane potential energy". When the sum result is greater than this value, the summation is no longer continued to reduce the calculation pressure; d0 is the escape distance, and the potential energy of positions beyond this distance is 0.

[0098] In the present invention, the influence of the driving trend of the main road on the trajectory generation is considered, that is, the driving trajectory trends of the vehicle entering and leaving the loading yard must be consistent with the driving trends of the two main roads in the yard, avoiding situations such as vehicle blocking and avoidance during the multi-vehicle transportation operation in the mining area loading yard, as shown in Figure 3 (a).

[0099] Step S5, determine the trajectory planning tasks in the mining area loading yard; the trajectory planning tasks in the mining area loading yard include Trajectory Planning Task 1 and Trajectory Planning Task 2;

[0100] The Trajectory Planning Task 1 is the parking-in trajectory planning task starting from the empty lane and stopping at the loading point;

[0101] The Trajectory Planning Task 2 is the leaving-yard trajectory planning task starting from the loading stop point and merging into the heavy lane.

[0102] It can be understood that when performing the parking-in trajectory planning, the empty lane plays a "guiding" role in the trajectory, and the heavy lane plays a "pushing-away" role in the trajectory; when performing the leaving-yard trajectory planning, it is the opposite.

[0103] Optionally, the expression of the trajectory planning task is:

[0104]

[0105] where S(T) is the trajectory planning task at the T-th time step.

[0106] Step S6, based on the static obstacle artificial potential field, empty lane potential energy, heavy lane potential energy at any position, and the potential energy trajectory planning task, obtain the complete potential field of the trajectory planning space, and the expression is:

[0107] E plan = k1E static + S(T)·k2E empty − S(T)·k3E load

[0108] Among them, k1 is the weight coefficient of the static obstacle potential energy; k2 is the weight coefficient of the empty lane potential energy; k3 is the weight coefficient of the heavy lane potential energy.

[0109] Step S7: Based on the complete potential energy field of the trajectory planning space, design the driving cost to obtain the updated driving cost; design the heuristic cost based on the number of gear shifts of the path to obtain the updated heuristic cost;

[0110] Based on the updated driving cost and the updated heuristic cost, obtain the total cost;

[0111] Based on the total cost, improve the Hybrid A * algorithm; obtain the improved Hybrid A * algorithm;

[0112] In the present invention, the potential energy field is used to describe the influence of the environment on the moving object; among them, a higher potential energy value exists near the obstacle, while a lower potential energy value exists in the area far from the obstacle, and a route passing through the low potential energy area is selected to reduce the potential collision risk.

[0113] Optionally, the expression of the total cost f(N) is:

[0114] f(N) = g(N) + h(N)

[0115] Among them, f(N) is the total cost of the obstacle node N, that is, the comprehensive priority value, g(N) is the actual driving cost from the starting point to the obstacle node N, and h(N) is the heuristic estimated cost from the obstacle node N to the end point;

[0116] Furthermore, the expression of the driving cost g(N) of the obstacle node N is:

[0117] g(N) = λ1d(N) + λ2Q(N) + λ3R(N)

[0118] Among them, d(N) is the driving time from the search starting point to the obstacle node N, which is defined by the ratio of the driving distance to the turning speed limit and the straight-line speed limit inside; Q(N) represents the cumulative effect of the potential energy on the path from the starting point to the obstacle node N, which is obtained by summing the potential energy values of each path point χ in the discretized sampling path point set Г(N); R(N) is the number of gear shifts of the path from the starting point to the obstacle node N, to avoid frequent steering wheel turning when the vehicle tracks the path; λ1 is the weight coefficient of the driving distance; λ2 is the weight coefficient of the potential energy integral value; λ3 is the weight coefficient of the number of steering changes.

[0119] Among them, in the present invention, the path from the starting point to the obstacle node N is discretely sampled to obtain the path point set Г(N);

[0120] By summing the potential energy values ​​of each path point χ, the cumulative effect Q(N) of the potential energy on the path from the starting point to the obstacle node N is obtained, and the expression is:

[0121] Q(N)=∑ χ∈Г(N) E plan (x)

[0122] The expression of the heuristic cost h(N) is:

[0123] h(N)=μ1max{d Astar (N),d RS (N)}+μ2E plan (pos(N))

[0124] Among them, d Astar (N) is the distance from obstacle node N to the search end point A * The estimated driving time of the search result is defined by the ratio of the driving distance to the turning speed limit and the straight speed limit; d RS (N) is the length of the Reed-Shepp curve from obstacle node N to the search end point; pos(N) is the position information of obstacle node N; E plan (pos(N)) is the potential energy value at the position of the obstacle node N; μ1 is the weight coefficient of the estimated driving distance; μ2 is the weight coefficient of the potential energy integral value.

[0125] Step S8: Using the improved Hybrid A * The algorithm obtains the optimal feasible smooth path from the starting point to the destination of the unmanned vehicle in the mine loading yard.

[0126] Optionally, the improved Hybrid A is used in step S8. * The specific steps of the algorithm to obtain the optimal feasible smooth path from the starting point to the destination of the unmanned vehicle in the mine loading yard include:

[0127] Step S8.1, initializing the open list Openlist and the closed list Closelist;

[0128] Determine the path starting node n s The position coordinates, heading angle and target node n t Position coordinates and heading angle;

[0129] Set the path starting node n s Add to the initial open list Openlist to obtain the open list Openlist;

[0130] Exemplarily, the open list stores all the nodes that are currently being considered but not yet fully explored; each node has a corresponding cost value f(n) in the open list.

[0131] The closed list is a set that records the nodes that have been fully explored; once a node is taken out of the open list and all its neighbors have been processed, this node will be added to the closed list.

[0132] Step S8.2: Let a = 1. When a = 1, it represents the first round of iteration.

[0133] Step S8.3: Check the status of the open list Openlist in the a-th round of iteration. If it is empty, terminate the calculation and determine that no valid path to the target node n t can be found.

[0134] If the open list Openlist in the a-th round of iteration is not empty, select the node a with the lowest cost value in the open list Openlist of the a-th round of iteration, transfer it to the closed list Closelist of the a-th round of iteration to obtain the updated closed list B a , and remove it from the open list Openlist to obtain the updated open list C a ; Use the updated closed list B a as the closed list for the (a + 1)-th round of iteration, and use the updated open list C a as the open list for the (a + 1)-th round of iteration.

[0135] Step S8.4: Determine whether the node a is the target node n t , that is, determine whether the yaw angle and distance between the node a and the target node n t are within the given threshold range.

[0136] If so, proceed to the next step; if not, that is, the distance between the node a and the target node n t is greater than the threshold, then let a = a + 1 and return to Step S8.3.

[0137] In the present invention, the set threshold range is that the threshold range of the yaw angle error is 5°, and the threshold range of the distance error is 2 meters.

[0138] Step S8.5: Use the Reed-Shepp curve generation algorithm to obtain the smooth path l from the node a to the target node n t ; n,a ;

[0139] Uniformly sample the smooth path l from the node a to the target node n t to obtain multiple sampling points n n,a ; i,a ;

[0140] Correspond the vehicle body coverage range corresponding to each sampling point n i,a to the grid cells, and detect whether the vehicle pose at each sampling point n i,a collides with an obstacle;

[0141] If there is no collision with the obstacle, then retain the smooth path l t from node a to the target node n n,a , and proceed to the next step;

[0142] If a collision occurs with the obstacle, then abandon the smooth path l t from node a to the target node n n,a , set a = a + 1, and return to step S8.3;

[0143] Step S8.6, Based on the smooth path l t from node a to the target node n n,a , with node a as the center, perform child node expansion through the improved Hybrid A * algorithm, and several child nodes n c,a with node a as the parent node and the expanded node a can be obtained; c = 1, 2, 3... C, where D represents the total number of child nodes of node a; Optionally, the expression of the expanded node a is:

[0144] a = (x, y, θ, g(a), f(a), *a p )

[0145] where, (x, y) represents the x-axis coordinate and y-axis coordinate of the node, θ represents the yaw angle of the node, g(a) is the driving cost from the starting point to the a-th node, f(a) is the total generation cost of the a-th node, that is, the comprehensive priority value, and *a p represents the pointer to the parent node;

[0146] In the present invention, the expansion method of the Hybrid A * algorithm is as Figure 5 shown. The Hybrid A * algorithm considers the driving characteristics of the vehicle when expanding nodes, introduces the basic concept of the Reeds-Shepp curve, and limits the transition process to the next position and orientation to one of 6 basic actions according to the current position and orientation state of the vehicle: straight forward (S+), straight backward (S-), left turn forward (L+), left turn backward (L-), right turn forward (R+), right turn backward (R-). Among them, the four expansion directions of left front, right front, left rear, and right rear are based on the bicycle model and are obtained according to the maximum front wheel steering angle. The expansion step length is the distance traveled by the vehicle at a fixed speed for 1 s, and the expanded end position is the next search node of the vehicle;

[0147] Step S8.7, determine the child node n c,a is in the closed list B a , if not, proceed to step S8.8;

[0148] If the child node n c,a has the same position coordinates and yaw angle as the node in the closed list B a , it is the same node, indicating that the child node n c,a is in the closed list B a list. Do not process the child node n c,a , let c = c + 1, and return to step S8.7;

[0149] Step S8.8, calculate the total cost value f(n c,a ) of the child node n c,a . The expression is:

[0150] where f(n c,a ) is the comprehensive priority value of node n c,a , g(n c,a ) is the cost function of traveling from the starting point to node n c,a , and h(n c,a ) is the heuristic function from node n c,a to the end point;

[0151] Step S8.9, check whether the child node n c,a already exists in the updated open list C a . If not, add it to the updated open list C a to obtain the secondarily updated open list C′ a ; if it already exists, compare the traveling cost g(n c,a ) of the child node n c,a with the traveling cost g(n c-1,a ) of the child node n c-1,a . If the traveling cost g(n c,a ) of the child node n c,a is less than the traveling cost g(n c-1,a ) of the child node n c-1,a , then update the comprehensive priority value f(n c,a ) of the child node n c,a to obtain the updated child node n′ c,a ;

[0152] Use the updated child node n′ c,a to update the updated open list C a to obtain the secondarily updated open list C′ a , and proceed to the next step; if the child node n c,aThe driving cost g(n c,a ) is greater than or equal to the driving cost of the child node n c-1,a The driving cost g(n c-1,a ). Let c = c + 1, and return to step S8.7;

[0153] Step S8.10: Determine whether c is greater than or equal to D. If so, obtain the final secondary updated open list C″ of node a a , and proceed to the next step; if not, let c = c + 1 and return to step S8.7;

[0154] Step S8.11: Determine whether a is greater than or equal to A, where A represents the total number of iterations. If so, obtain the final secondary updated open list; obtain the optimal feasible smooth path based on the final secondary updated open list. If not, let a = a + 1 and return to step S8.3.

[0155] As mentioned above, it is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. An unmanned truck path planning method for a loading yard in a mining area, characterized in that, include: Step S1, obtaining a rasterized map of the entire area map of the mining area loading yard where the vehicle is located; Step S2: defining obstacle areas based on the rasterized map; Step S3, constructing a static obstacle artificial potential energy field; Step S4, determining the empty lanes and the loaded lanes of the mining area loading yard; On the artificial potential energy field of static obstacles, the influence of empty lanes and heavy lanes on the artificial potential energy field is added to obtain the potential energy of empty lanes and heavy lanes at any position respectively; Step S5, determining the trajectory planning task in the mining area loading yard; Step S6, based on the artificial potential energy field of static obstacles at any position, the potential energy of empty lanes and the potential energy of loaded lanes, and the potential energy trajectory planning task, a complete potential energy field of the trajectory planning space is obtained; Step S7: designing a driving cost based on the complete potential energy field of the trajectory planning space to obtain an updated driving cost; Design the heuristic cost based on the number of gear shifts in the path and obtain the updated heuristic cost; Based on the updated exercise cost and the updated inspiration cost, the total cost is obtained; Improve Hybrid A based on the total cost * algorithm to obtain the improved Hybrid A * algorithm; Step S8: Use the improved Hybrid A * algorithm to obtain the optimal feasible smooth path of the driverless vehicle in the loading yard of the mining area from the starting point to the destination.

2. The unmanned truck path planning method for a mine loading yard according to claim 1 is characterized in that: The specific steps to define the obstacle area based on the rasterized map include: Step S21, extracting the irregular obstacle coverage area from the rasterized map as obstacle area 1; Step S22, using an image processing algorithm to identify the boundary contour of the object in the image, and marking the area outside the boundary contour of the object in the image as obstacle area 2; Step S23: merge obstacle area 1 and obstacle area 2 to define an obstacle area.

3. The unmanned truck path planning method for a mine loading yard according to claim 1 is characterized in that: The static obstacle artificial potential energy field is a static obstacle artificial potential energy field at any position, and the expression is: Among them, E static (P) is the artificial potential field of static obstacles, and E max is the upper limit of the maximum potential energy in the artificial potential field; is the distance between the position P and the obstacle point ξ; Ω is the set of all static obstacle points in the loading yard; m static is the power function factor of the artificial potential field of static obstacles; d0 is the escape distance, and the potential energy at positions beyond this distance is 0.

4. The unmanned truck path planning method for a mine loading yard according to claim 1 is characterized in that: The empty lane is a lane for unloaded vehicles to enter the loading yard; The heavy-load lane is the lane where the heavily loaded vehicle leaves the loading yard after completing the loading task.

5. The unmanned truck path planning method for a mine loading yard according to claim 1 is characterized in that: The first trajectory planning task is a parking trajectory planning task starting from an empty lane and stopping at a loading point; The second trajectory planning task is to start from the loading stop and merge into the heavy lane to leave the site.

6. The method for unmanned truck path planning for a mining area loading yard according to claim 3, wherein The complete potential energy field of the trajectory planning space is obtained, and the expression is: E plan = k1E static + S(T)·k2E empty − S(T)·k3E load Among them, k1 is the weight coefficient of the static obstacle potential energy; k2 is the weight coefficient of the empty lane potential energy; k3 is the weight coefficient of the heavy lane potential energy, T is the time step, and E static is the artificial potential energy of the static obstacle; E empty is the potential energy of the empty lane, E load is the potential energy of the heavy lane, and S(T) is the trajectory planning task at the T time step.

7. The unmanned truck path planning method for a mine loading yard according to claim 1 is characterized in that: The specific steps of obtaining the optimal feasible smooth path of the driverless vehicle in the loading yard of the mining area from the starting point to the destination by using the improved Hybrid A algorithm described in step S8 are as follows: * ​ Step S8.1, initializing the open list and the closed list; Step S8.2, let a=1. When a=1, it indicates the first round of iteration; Step S8.

3. Check the status of the open list in the a-th iteration. If it is empty, terminate the calculation and determine that no valid path leading to the target node n t can be found; If the open list in the a-th iteration is not empty, select the node a with the lowest cost value in the open list of the a-th iteration, and update the closed list and the update list in the a-th iteration to obtain the updated closed list B a and the updated open list C a ; use the updated closed list B a as the closed list in the (a + 1)-th iteration, and use the updated open list C a as the open list in the (a + 1)-th iteration; Step S8.4, determine whether node a is the target node n t ; If yes, proceed to the next step; if no, set a=a+1 and return to step S8.3; Step S8.

5. Use the Reed-Shepp curve generation algorithm to obtain a smooth path l from node a to the target node n t and determine whether it collides with an obstacle. If so, set a = a + 1 and return to step S8.3; n,a ​ If not, proceed to the next step; Step S8.

6. Based on the smooth path l from node a to the target node n t , with node a as the center, expand the child nodes through the improved Hybrid A n,a algorithm to obtain the expanded node a and several child nodes n * ; c = 1, 2, 3... D, where D represents the total number of child nodes of node a; c,a ​ Step S8.7, traverse all child nodes n c,a to obtain the final secondary updated open list C″ of node a a and proceed to the next step; Step S8.8, determine whether a is greater than or equal to A, where A represents the total number of iterations. If so, obtain the final secondary update open list; based on the final secondary update open list, obtain the optimal feasible smooth path. If not, set a=a+1 and return to step S8.

3.

8. The method for path planning of an unmanned truck for a loading yard in a mining area according to claim 7, wherein Separate the updated closed list B a and the updated open list C a The specific steps are as follows: Select the node a with the lowest cost value in the open list of the a-th iteration, transfer it to the closed list of the a-th iteration to obtain the updated closed list B a , and remove it from the open list to obtain the updated open list C a ; Use the updated closed list B a as the closed list for the (a + 1)-th iteration, and use the updated open list C a as the open list for the (a + 1)-th iteration.

9. The method for path planning of an unmanned truck for a mining area loading yard according to claim 7, wherein Determine the smooth path l from node a to the target node n t Specific steps for determining whether it collides with an obstacle include: n,a Whether it collides with an obstacle The smooth path l from node a to the target node n t is uniformly sampled to obtain multiple sampling points n n,a ; i,a ; For each sampling point n i,a The corresponding vehicle body coverage range is mapped to the grid cells, and it is detected whether the vehicle pose at each sampling point n i,a collides with obstacles.

10. The method for path planning of an unmanned truck for a mining area loading yard according to claim 7, wherein The judgment sub-node n c,a is in the closed list B a The specific steps are as follows: If child node n c,a has the same position coordinates and yaw angle as the nodes in the closed list B a , it is the same node, indicating that child node n c,a is in the closed list B a list; otherwise, it is not in the closed list.

Images