Method and device for determining distance between intelligent operation machines and electronic equipment

Abstract

The invention provides a method and device for determining the distance between intelligent operation machines and electronic equipment, and belongs to the technical field of automatic driving. The method comprises the following steps: acquiring a first distance between a first operation machine and a second operation machine and a second distance between the first operation machine and the second operation machine; inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result of the first operation machine and the second operation machine output by the target distance determination model; wherein the first distance is obtained based on the positions of the first operation machine and the second operation machine in the road network coordinate system; the second distance is obtained based on mutual perception between the first work machine and the second work machine. According to the embodiment of the invention, the accuracy of determining the distance between the intelligent operation machines is improved.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and specifically to a method for determining the distance between intelligent working machines, a device for determining the distance between intelligent working machines, an electronic device, a machine-readable storage medium, and a computer program product. Background Art

[0002] Currently, some intelligent operating machines, such as construction machinery, agricultural machinery, and robots, are equipped with autonomous driving decision-making systems to enable intelligent driving and improve operational efficiency. For example, in closed environments like mines, installing autonomous driving decision-making systems on each truck allows for real-time driving strategy development, improving both safety and efficiency in mining operations.

[0003] Existing autonomous driving strategies for intelligent working machines often rely on sensor input to determine the distance between them and make real-time decisions. However, when working in harsh environments, the input data from a single sensor is unreliable and prone to recognition errors. This results in low accuracy in identifying the distance between intelligent working machines, posing a risk to autonomous driving.

[0004] Application Contents

[0005] The purpose of the embodiments of the present application is to provide a method, device and electronic device for determining the distance between intelligent working machines, so as to solve the problem of low accuracy in distance recognition between intelligent working machines.

[0006] To achieve the above objectives, an embodiment of the present application provides a method for determining the distance between intelligent working machines, comprising:

[0007] Acquire a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine;

[0008] Inputting the first distance and the second distance into a target distance determination model, and obtaining a distance prediction result between the first working machine and the second working machine output by the target distance determination model;

[0009] Among them, the first distance is obtained based on the positions of the first working machine and the second working machine in the road network coordinate system; the second distance is obtained based on the mutual perception between the first working machine and the second working machine; the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result according to at least one of the first distance and the second distance.

[0010] On the other hand, an embodiment of the present application further provides a device for determining the distance between intelligent working machines, comprising:

[0011] An acquisition module, configured to acquire a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine;

[0012] a prediction module, configured to input the first distance and the second distance into a target distance determination model, and obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model;

[0013] Among them, the first distance is obtained based on the positions of the first working machine and the second working machine in the road network coordinate system; the second distance is obtained based on the mutual perception between the first working machine and the second working machine; the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result according to at least one of the first distance and the second distance.

[0014] On the other hand, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for determining the distance between intelligent working machines when executing the program.

[0015] On the other hand, the present application also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for determining the distance between intelligent working machines.

[0016] On the other hand, the present application also provides a computer program product, including a computer program, which implements the above-mentioned method for determining the distance between intelligent working machines when executed by a processor.

[0017] Through the above technical solution, the embodiment of the present application determines the distance between the intelligent working machines by using a target distance determination model to perform a fusion calculation of the multi-path distance of the perception distance and the relative distance in the road network coordinate system based on the first distance between the first working machine and the second working machine obtained in the road network coordinate system, and the second distance obtained based on the mutual perception between the first working machine and the second working machine. The embodiment of the present application solves the problem that the input data of a single perception sensor or satellite positioning signal has low credibility, is prone to recognition errors and the risk of satellite positioning signal disconnection, and leads to low accuracy in distance recognition between intelligent working machines. The embodiment of the present application can also adjust the influence of the first distance and the second distance on the distance prediction result according to the first distance and / or the second distance, which is conducive to improving the accuracy of distance determination between intelligent working machines.

[0018] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present application but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:

[0020] Figure 1 This is one of the flow charts of the method for determining the distance between intelligent working machines provided in this application;

[0021] Figure 2 This is the second flow chart of the method for determining the distance between intelligent operating machines provided by this application;

[0022] Figure 3 is a schematic diagram of the road network coordinate system provided by this application;

[0023] Figure 4 is a schematic diagram of determining the first distance provided by this application;

[0024] Figure 5 This is a schematic diagram of 5G-based V2N multi-vehicle cooperative communication provided by this application;

[0025] Figure 6 This is the third flow chart of the method for determining the distance between intelligent operating machines provided in this application;

[0026] Figure 7 This is a schematic diagram of the structure of the device for determining the distance between intelligent working machines provided in this application;

[0027] Figure 8 It is a structural diagram of the electronic device provided in this application. DETAILED DESCRIPTION

[0028] The following describes the specific implementation of the embodiment of the present application in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiment of the present application and is not intended to limit the embodiment of the present application.

[0029] Method Example

[0030] Please refer to Figure 1 , an embodiment of the present application further provides a method for determining the distance between working machines, comprising:

[0031] Step 100: Acquire a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine.

[0032] The electronic device on the first working machine obtains a first distance between the first working machine and the second working machine, and a second distance between the first working machine and the second working machine. The working machine can be a machine that performs working tasks in various closed environments. For example, the working machine can be a mining truck that performs transportation tasks in a mining area, or a vehicle that performs transportation tasks in a closed environment at a material collection area, or an agricultural machine that performs combined operations in the field, or an operating robot. The working machine in the embodiment of the present application is described using a mining truck that performs transportation tasks in a mining area as an example.

[0033] The first distance may be the distance between the first working machine and the second working machine in a road network coordinate system. The second distance may be obtained based on mutual perception between the first working machine and the second working machine. The second distance may be obtained based on mutual perception between the first working machine and the second working machine using various distance-capable sensing sensors. For example, the sensing sensor may utilize any one of a camera, a lidar, an ultrasonic radar, and a millimeter-wave radar. The camera may be any one of a wide-angle camera, a visible light camera, and an infrared camera. The lidar may be any one of a multi-line lidar and a single-line lidar.

[0034] Step 200: Input the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model.

[0035] The electronic device inputs the first distance and the second distance into a target distance determination model, and obtains a distance prediction result between the first and second work machines output by the target distance determination model. The target distance determination model adjusts the influence of the first and second distances on the distance prediction result based on at least one of the first and second distances.

[0036] In one example, the target distance determination model can be a regression model. The regression model can perform a weighted calculation on the first distance and the second distance to obtain a distance prediction result. The weight used in the weighted calculation is affected by the first distance and / or the second distance. For example, the first distance can be considered an initially determined distance. As the first distance decreases, the weight of the first distance decreases, while the weight of the second distance increases. Thus, as the initially determined distance between the first working machine and the second working machine decreases, the credibility of the second distance obtained based on perception can be considered higher, thereby increasing the weight of the second distance and improving the accuracy of the distance prediction result.

[0037] In other examples, the second distance may also be used as the initially determined distance.

[0038] In some other examples, the initially determined distance can also be determined in combination with the first distance and the second distance. For example, the first distance and the second distance can be weighted using the default weight to obtain the initially determined distance. When the initially determined distance is less than the preset distance threshold, the weight used in the weighted calculation is adjusted to obtain the distance prediction result.

[0039] Of course, in some feasible implementations, the target distance determination model may also be a neural network model, and the first distance and the second distance are input into the neural network model to obtain a distance prediction result. During the training phase of the neural network model, by setting training samples, the trained neural network model may be capable of adjusting the impact of the first distance and the second distance on the distance prediction result based on the first distance and / or the second distance. For example, when the first distance is smaller, the output distance prediction result is significantly closer to the second distance than when the first distance is larger.

[0040] The embodiment of the present application determines the distance between the operating machines by performing a multi-path distance fusion calculation of the perceived distance and the relative distance in the road network coordinate system based on a first distance between the first operating machine and the second operating machine obtained in a road network coordinate system, and a second distance obtained based on mutual perception between the first operating machine and the second operating machine using a target distance determination model. The embodiment of the present application solves the problem that the input data of a single perception sensor or satellite positioning signal has low credibility, is prone to recognition errors and the risk of satellite positioning signal disconnection, and leads to low accuracy in distance recognition between intelligent operating machines. The embodiment of the present application is also capable of adjusting the influence of the first distance and the second distance on the distance prediction result based on the first distance and / or the second distance, which is conducive to improving the accuracy of distance determination between intelligent operating machines.

[0041] In one embodiment, the target distance determination model estimates the relative distance x to the road network coordinate system of the working machine. position And the perception output distance x perception , mathematical modeling is performed, based on the principle that long-distance road network coordinate estimation accounts for a large proportion, while short-distance perception accounts for a large proportion. The output is the distance prediction result z between the operating machines after data fusion. For example, the target distance determination model can be mathematically modeled based on linear regression modeling.

[0042] In the embodiment of this application, the input and output variables are defined as follows: x represents the relative distance between the working machines estimated based on the road network coordinate system (first distance); y represents the relative distance between the working machines based on the mutual perception output of the perception sensors (second distance); and z represents the predicted distance between the working machines. It is assumed that the predicted distance z between the working machines is composed of a linear combination of the above input variables:

[0043] z=F(β0,β1,β2,x,y);

[0044] Where β0 is the bias term, representing the intercept of the model; β1 and β2 are weight coefficients learned from the training data, representing the impact of each input variable on the actual relative distance between the two vehicles. The weight coefficients are estimated using the least squares method, and a loss function is constructed to minimize the error. To minimize the loss function, the optimal weight coefficients β0, β1, and β2 are determined. An analytical method is used to calculate the optimal weight coefficients β (substituting β0, β1, and β2), thereby obtaining the prediction model.

[0045] According to the optimal weight coefficient β, for the new input values x and y, we can get the predicted value of the distance between the operating machines as:

[0046] z pred =F(β, x, y).

[0047] The embodiment of the present application uses multiple information sources as input, upgrading from narrow multi-sensor fusion to broad multi-path distance data input, fully covering the relative positions of vehicles beyond visual range and within the perception range, providing data support for the decision-making system, and improving system robustness.

[0048] For other aspects of the embodiments of this application, please refer to Figure 2 The first distance between the first operating machine and the second operating machine can be obtained by the following steps:

[0049] Step 10: Acquire a road network coordinate system, first path planning data of a first operating machine in the road network coordinate system, and second path planning data of a second operating machine in the road network coordinate system.

[0050] The electronic device obtains a road network coordinate system, first path planning data for a first operating machine in the road network coordinate system, and second path planning data for a second operating machine in the road network coordinate system. The operating machine can be a machine that performs operating tasks in various closed environments. For example, the operating machine can be a mining truck that performs transportation operations in a mining area, or a vehicle that performs transportation operations in a closed environment at a material collection area, or an agricultural machine that performs combined operations in the field, or an operating robot. The operating machine in the embodiments of the present application is described using a mining truck that performs transportation operations in a mining area as an example.

[0051] The road network coordinate system includes multiple target areas and road network segments connecting adjacent target areas. For example, please refer to Figure 3The multiple target areas include parking areas, loading areas, and unloading areas. The road network coordinate system also includes road network segments connecting adjacent parking areas and loading areas, adjacent parking areas and unloading areas, and adjacent loading areas and unloading areas. The first path planning data includes the road network segments that the first working machine is planned to reach at each travel time and the road network points on the road network segments. The second path planning data includes the road network segments that the second working machine is planned to reach at each travel time and the road network points on the road network segments.

[0052] Step 20: Determine a first position of the first working machine in the road network coordinate system based on the first path planning data.

[0053] Step 30: Determine a second position of the second working machine in the road network coordinate system based on the second path planning data.

[0054] The first position includes the first road network segment where the first operating machine is located and the first road network point on the first road network segment. The second position includes the second road network segment where the second operating machine is located and the second road network point on the second road network segment. The first path planning data includes the road network segment that the first operating machine is planned to reach at each driving time and the road network points on the road network segment. The second path planning data includes the road network segment that the second operating machine is planned to reach at each driving time and the road network points on the road network segment. Therefore, when the first driving time of the first operating machine is obtained, the road network segment reached at the first driving time and the first road network point on the road network segment can be obtained. When the second driving time of the second operating machine is obtained, the road network segment reached at the second driving time and the second road network point on the road network segment can be obtained.

[0055] Step 40: Determine the distance between the first working machine and the second working machine based on the relative positions of the first road network segment and the second road network segment in the road network coordinate system, the first road network point, and the second road network point.

[0056] For example, see Figure 4 Based on step 20, the distance between the first network point of vehicle A in the +n network segment and the starting point of the +n network segment is Based on step 30, the distance between the second network point of vehicle B in the +(n+1) network segment and the starting point of the +(n+1) network segment is The distance between car B and car A is Among them, L n is the length of the +n network segment.

[0057] In an embodiment of the present application, a first position of the first working machine in the road network coordinate system and a second position of the second working machine in the road network coordinate system are obtained based on first path planning data of the first working machine in the road network coordinate system and second path planning data of the second working machine in the road network coordinate system, respectively, thereby determining the distance between the first position and the second position in the road network coordinate system as the distance between the working machines. In an embodiment of the present application, the distance between the working machines obtained based on the path planning data of the working machines in the road network coordinate system can effectively overcome the problem of low reliability of sensor data caused by complex environments and improve the accuracy of determining the distance between the working machines.

[0058] In other aspects of the embodiments of the present application, the plurality of target areas include a parking area, a loading area, and an unloading area; and the road network coordinate system is constructed by the following steps:

[0059] Step 11: Divide the working path between the parking area and the loading area, and the working path between the parking area and the unloading area into road network segments based on bifurcation points to obtain a plurality of road network segments.

[0060] Step 12: Identify the road network segments between the parking area and the loading area and the road network segments between the parking area and the unloading area one by one, and use different symbols to identify the same road network segment according to the different driving directions of the target working machinery, so as to obtain the identification information of all road network segments in the road network coordinate system, so as to distinguish different road network segments in the road network coordinate system and distinguish the driving direction of the target working machinery on the road network segment.

[0061] Please refer to Figure 3 Taking the unmanned driving of a mining truck in a closed mining area as an example, the principles for establishing the road network coordinate system in the embodiment of the present application are as follows:

[0062] Numbering: In the unmanned mining truck scenario, various functional areas are set up, such as parking areas, loading areas, and unloading areas. The parking area is the starting point and assigned the number 1. The remaining target points are numbered, such as Loading Area 1 (Target Area 1), Loading Area 2 (Target Area 2), Unloading Area 1 (Target Area 3), and Unloading Area 2 (Target Area 4). The road network segments from the parking area to Target Area 1 are assigned ascending numbers. The next target area, Target Area 2, is assigned ascending numbers after the fork in the road, and so on.

[0063] Segmentation: Extract bifurcation points to segment the road network, and connect any two adjacent bifurcation points in the road network to obtain a road network segment. Among them, for a multi-branch road network scenario, first divide the road network into different road network segments, and label them to establish a road network coordinate system for the mining area. The same road network segment can be divided into two sections with the same number but different labels according to different directions. For example, "+1" represents the forward direction of road network segment No. 1, and "-1" represents the reverse direction of road network segment No. 1. The specific direction can be defined by yourself and it can be kept consistent. Please refer to Figure 4 , then the forward network segment set Forward = {"+1","+2","+3",...,"+N"}, with a total of n segments. The reverse network segment set Backward = {"-1","-2","-3",...,"-N"}, with a total of n segments. They correspond one to one. In other embodiments, the network segments can also be identified by characters, for example, "+A" represents the forward direction of network segment A, and "-A" represents the reverse direction of network segment A.

[0064] It should be noted that after adding a new target area to the road network coordinate system, the assigned values can continue to be based on the numbering principle. For example, if there is loading area 2 in the original road network coordinate system, a new loading area can be assigned the value of loading area 3. If the original target area changes, such as the cancellation of loading area 1, the road network coordinate system is re-established, and the identifiers (numbers) of the target area and the road network are re-assigned.

[0065] Based on the road network coordinate system, in the embodiment of the present application, the relative distance between the working machines is calculated based on the planned path of the working machine. First, a heuristic search algorithm (such as the A* algorithm) is used to plan the overall road network segment path, and for each road network segment, smoothed adjacent road network points with an equal distance d are generated. The distance between two adjacent road network points on the road network segment is set to a set distance value d, and the road network points on the road network segment are obtained by the following steps:

[0066] For each network segment in the network coordinate system, perform the following steps:

[0067] Step 31: Convert the road network segment labels into a grid, wherein the grid includes a plurality of grid nodes.

[0068] Step 32: Initialize the algorithm parameters of the A-star search algorithm. The algorithm parameters include at least an initial identifier, a target identifier, an open list, and a closed list. The open list is used to store the grid nodes to be explored, and the closed list is used to store the grid nodes that have been explored. The initial identifier and the target identifier represent the starting point and target grid node of the road network segment, respectively.

[0069] Repeat the following steps until the target identifier is found or the open list is empty:

[0070] Step 33: Find the grid node with the smallest total evaluation value from the open list as the current node, remove the current node from the open list, and add it to the closed list.

[0071] Step 34: When the distance between the adjacent nodes around the current node and the current node is the set distance value, calculate the actual cost, heuristic cost and total evaluation of the adjacent nodes around the current node, and update the open list and the closed list.

[0072] Step 35: trace back from the target identifier according to the parent node pointer until tracing back to the initial identifier, implement path planning for the road network segment, and obtain the path planning points of the road network segment.

[0073] Step 36: Use the B-spline curve to interpolate the path planning points of the road network segment, so as to obtain smoothed adjacent road network points with equal distances equal to the set distance value.

[0074] The A* algorithm is a heuristic search algorithm used to find the shortest path from a starting point to an end point in a graph or grid. Its core idea is to select nodes by evaluating the function f(n) = g(n) + h(n), where:

[0075] g(n) is the actual cost from the starting point to the current node, and h(n) is the heuristic estimated cost from the current node to the end point. For a road network segment in a road network coordinate system, the road network segment mark is converted into a grid, and the grid includes multiple grid nodes. Each grid node represents a possible location. Use Euclidean distance as a heuristic function. Initialize the open list and the closed list. Add the starting point of the adjacent road network segment to the open list, set its g value to 0, and its h value to the heuristic estimated distance from the starting point to the end point. Select the node with the smallest f value from the open list as the current node. Add the current node to the closed list. For each adjacent node of the current node, if the distance between the adjacent node and the current node is d, calculate its g value and h value, update the f value, and add it to the open list. Repeat the above steps until the end point is found or the open list is empty.

[0076] Repeat the above steps for all road network points to form the path planning points for the entire road network segment. Use B-spline curves to interpolate the path points and smooth the path, thus obtaining adjacent road network points with equal distance d after smoothing.

[0077] Determining the first position of the first operating machine in the road network coordinate system based on the first path planning data includes: determining the first position according to the first driving time of the first operating machine and the first path planning data. The second path planning data includes the road network segments that the second operating machine plans to reach at each driving time and the road network points on the road network segments; determining the second position of the second operating machine in the road network coordinate system based on the second path planning data includes: determining the second position according to the second driving time of the second operating machine and the second path planning data. The electronic device outputs the road network segments that are planned to be reached at each driving time and the road network points on the road network segments in real time. Taking the first operating machine as an example, the first driving time t of the first operating machine is obtained, and combined with the road network segments that are planned to be reached at the first driving time and the adjacent road network points with equal distances d, the distance l from the current first operating machine to the starting point of the road network segment can be obtained. t =t×d. After the design of each road network section is completed, the total distance of the single section is L n is a fixed value, which should be L n =t nmax ×d. By building a table for the tree-like road network, the relative position of the current operating machine in the self-built road network coordinate system can be obtained, thereby obtaining the distance between each operating machine and the self-vehicle.

[0078] Please refer to Figure 4 , taking car A as the ego car, the position of car A in the +n road network segment is The position of vehicle B in the +(n+1) network segment The position of vehicle C in the -(n+1) network segment

[0079] When obtaining the position of car A in the +n road network segment, the car A is used as the self-car. The position of vehicle B in the +(n+1) network segment The position of vehicle C in the -(n+1) network segment In the case of, the distance between car B and car A is The distance between car C and car A

[0080] Please note that, please refer to Figure 5In the embodiment of the present application, each operating machine can upload its own path planning data and theoretical satellite positioning data in real time through 5G networking and MQTT transmission protocol, and obtain the path planning data and theoretical satellite positioning data of other operating machines in the road network coordinate system. Among them, the ego vehicle is egoA, and the set of other operating machines is Object = {objectB, objectC, objectD, ..., objectM}. Through this information, the relative path distance between the ego vehicle and other operating machines can be obtained. Thus, the distance set X = {x B , x C , x D ,...,x M 5G-based V2N multi-vehicle collaborative communication enables multi-vehicle data exchange based on a standard dispatching system. Through the exchange of core data, critical data from multiple vehicles can be obtained under low-load conditions, achieving a system design that is light on dispatching and focused on individual units. V2N-based multi-vehicle collaborative communication collects real-time vehicle information within the road network coordinate system, eliminating the need for prediction and enabling accurate and efficient spatiotemporal autonomous driving behavior decisions for the vehicle itself.

[0081] Therefore, the embodiment of the present application establishes a road network coordinate system for the mining area based on a closed known route, and estimates the position of the vehicle and other vehicles and calculates the relative distance based on the known path planning data. The method is simple and efficient, and the data results are accurate.

[0082] In other aspects of the embodiments of the present application, during the mathematical modeling and testing process, it was observed that when the predicted distance is the set distance boundary value D, the difference trend between the first distance x and the second distance y of the input source will jump, and the contribution value of the two to the predicted value will also change greatly. Therefore, the embodiment of the present application introduces a set distance boundary value D to dynamically adjust the contribution of the first distance x and the second distance y of the input source. When the predicted distance is less than the set distance boundary value D, it relies more on the relative distance (second distance) y between the operating machines based on the mutual perception output of the perception sensor. When the distance is greater than or equal to the set distance boundary value D, it relies more on the relative distance (first distance) x between the operating machines estimated based on the road network coordinate system.

[0083] Therefore, the embodiment of the present application sets the target distance determination model to include a first model and a second model.

[0084] Step 200: Inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model, including:

[0085] Step 210: Input the first distance and the second distance into the first model to obtain an initial predicted distance;

[0086] Step 220: When the initial predicted distance is less than the set distance boundary value, the first distance and the second distance are input into the second model to obtain the distance prediction result; when the initial predicted distance is greater than or equal to the set distance boundary value, the distance prediction result is determined to be the initial predicted distance.

[0087] Specifically, the first model can be constructed based on a conventional linear regression prediction model. The second model can be constructed based on a weighted linear regression prediction model. The electronic device inputs the first distance and the second distance into the linear regression prediction model to obtain an initial predicted distance. If the initial predicted distance is less than a set distance boundary value, the first distance and the second distance are input into the weighted linear regression prediction model to obtain a distance prediction result output by the weighted linear regression prediction model. If the initial predicted distance is greater than or equal to the set distance boundary value, the distance prediction result is determined to be the initial predicted distance output by the linear regression prediction model. The influence of the second distance on the distance prediction result in the linear regression prediction model is less than the influence of the second distance on the distance prediction result in the weighted linear regression prediction model. That is, the initial predicted distance of the linear regression prediction model is more dependent on the relative distance between the working machines (the first distance) estimated based on the road network coordinate system; the distance prediction result of the weighted linear regression prediction model is more dependent on the relative distance between the working machines (the second distance) based on the mutual perception output of the perception sensors. By setting the distance boundary value D, the bias of the target distance determination model within different distance ranges can be flexibly adjusted, thereby improving the accuracy of distance determination between working machines.

[0088] In other aspects of the present application, to further improve the accuracy of determining the distance between working machines, the confidence level of the first distance and the confidence level of the second distance can be combined and input into a target distance determination model to obtain a predicted distance between the first and second working machines output by the target distance determination model. By introducing data confidence, a value judgment is made on the distance between working machines in the mine's custom road network coordinate system and the distance between working machines obtained by the perception sensors. Through mathematical modeling, the fused data is output.

[0089] Specifically, step 200, inputting the first distance and the second distance into the target distance determination model, and obtaining the distance prediction result between the first working machine and the second working machine output by the target distance determination model, includes: inputting the first distance, the confidence level of the first distance, the second distance and the confidence level of the second distance into the target distance determination model, and obtaining the distance prediction result between the first working machine and the second working machine output by the target distance determination model.

[0090] The confidence level of the first distance is obtained by the following steps:

[0091] Step 41: Acquire the road network planning data of the target operating machine, wherein the road network planning data includes a plurality of road network points and theoretical satellite positioning data corresponding to each of the road network points. The target operating machine includes at least one of the first operating machine and the second operating machine.

[0092] The satellite positioning data can be GPS data or BeiDou data, and the embodiment of the present application is explained by taking GPS data as an example. The road network points in the road network planning data can be any point on the road network segment. The target operating machinery includes the first operating machinery and / or the second operating machinery. Taking the first operating machinery as an example, in the process of constructing the road network coordinate system in the embodiment of the present application, the road network points and the GPS geodetic coordinates have a one-to-one correspondence, and the electronic equipment can perform real-time query by building a table. According to the current road network segment and road network point, the theoretical GPS coordinate point (x plan ,y plan , z plan ).

[0093] Step 42: Acquire the real satellite positioning data of the target operating machine when it arrives at at least one of the road network points.

[0094] The electronic device obtains the real GPS coordinate point (x current ,y current , z current ).

[0095] Step 43: Determine the confidence level of the first distance based on the positioning errors between the theoretical satellite positioning data and the actual satellite positioning data, wherein the positioning error is negatively correlated with the confidence level of the first distance.

[0096] The electronic device is based on the theoretical GPS coordinate point of the current road network point (x plan ,y plan , z plan ) and the real GPS coordinate point (x current ,y current , z current) is compared, and according to the size of the positioning error, the confidence of the first distance can be calculated to perform real-time verification of the periodic vehicle position estimation.

[0097] In one embodiment, the GPS network positioning quality is defined as Q gps , the positioning error of a single network point is l err .

[0098] Q gps ∈[0, 1];

[0099]

[0100] According to the status of the integrated navigation system, the GPS network positioning quality Q is determined gps Numeric value.

[0101] Among them, the system status is 1-satellite navigation mode, 2-integrated navigation mode, 3-pure inertial navigation mode;

[0102] The satellite status is 0 - no positioning and no orientation, 1 - single-point positioning and orientation, 2 - pseudo-range differential positioning and orientation, 3 - combined dead reckoning, 4 - RTK stable positioning and orientation, 5 - RTK floating-point positioning and orientation, 6 - single-point positioning and no orientation, 7 - pseudo-range differential positioning and no orientation, 8 - RTK stable positioning and no orientation, 9 - RTK floating-point positioning and no orientation.

[0103] When the system status is 2 and the satellite status is 4, GPS is the most stable and the GPS network positioning quality is Q gps 1. Under other combinations, the GPS network positioning quality Q is determined based on supplier information and actual application tests. gps Reasonable value of .

[0104] The mean square error is calculated by the theoretical satellite positioning data and the actual satellite positioning data of multiple road network points to obtain a reliable positioning error L err ′.

[0105]

[0106] Where n is the number of road network points, is the positioning error of the i-th road network point.

[0107] In one embodiment, it is necessary to determine the confidence level of the first distance in combination with the positioning error and a set error tolerance threshold. Thus, step 43, determining the confidence level of the first distance based on the positioning error between the theoretical satellite positioning data and the actual satellite positioning data, includes:

[0108] Step 431: Obtain an error tolerance threshold.

[0109] When the GPS network positioning quality Q gps When high, the positioning error L err ′ is closer to the real result. gps When low, the positioning error L err 'The result is unreliable. Therefore, according to the GPS network positioning quality Q gps The value of the error tolerance threshold L permit Adjust. Set the error tolerance threshold L permit It is obtained by calculating with the following formula:

[0110] L permit =L standard ×(1+k(1-Q gps );

[0111] Among them, L standard is the standard tolerance threshold, and k is the adjustment coefficient. Thus, the error tolerance threshold is related to the GPS network positioning quality Q in the target working machine. gps Negative correlation, that is, the error tolerance threshold is negatively correlated with the reliability of the satellite positioning system in the target working machine.

[0112] Step 432: Determine the confidence level of the first distance based on the difference between the positioning error and the error tolerance threshold.

[0113] In the embodiment of the present application, the difference between the positioning error and the error tolerance threshold may be defined as negatively correlated with the confidence level of the first distance. Specifically, in the embodiment of the present application, the confidence level of the first distance may be defined as Q position , Q position ∈[0,1]. The confidence of the first distance expresses the system’s confidence in the detection result. err ′ and error tolerance threshold L permit The difference between the first distance and the confidence level Q is calculated using an exponential function. position The specific formula is as follows:

[0114]

[0115] Among them, k>0 is the parameter that controls the rate of change and determines the confidence Q of the first distance position With positioning error L err 'The steepness of the change. err ′>L permit When the confidence level of the first distance is Q position The value drops steeply.

[0116] In dust-free, rain-free, and snow-free weather, sensors such as lidar, cameras, ultrasonic radar, and millimeter-wave radar can obtain the second distance between stable operating machines. In complex working conditions, the present embodiment performs post-processing on the second distance by setting a confidence level for the second distance.

[0117] In addition, the present application may set the confidence level of the second distance to Q perception Specifically, the confidence level of the second distance may be calculated based on the accuracy and recall rate of the distance between the operating machines sensed by the perception sensor.

[0118] Precision: This indicates the proportion of samples that are actually positive among those predicted to be positive based on the perception sensor. The calculation formula is:

[0119]

[0120] Among them, TP is True Positives and FP is False Positives.

[0121] Recall: It indicates the ratio of correctly detected positive samples to all positive samples based on the perception of the sensor. The calculation formula is:

[0122]

[0123] Among them, FN is False Negatives.

[0124] Confidence is usually related to the prediction score based on the perception of the sensor. The values of precision and recall can be changed by adjusting the confidence threshold. The specific steps are as follows:

[0125] 1. Determine the confidence threshold: According to the prediction score based on the perception output of the perception sensor (such as the confidence score in target detection), set different confidence thresholds, such as 0.1, 0.3, 0.5, 0.7, etc.

[0126] 2. Calculate precision and recall: For each confidence threshold, calculate the corresponding precision and recall values.

[0127] 3. Draw the precision-recall curve: Draw the curve with recall as the horizontal axis and precision as the vertical axis.

[0128] The confidence Q of the second distance perceptionIt can be defined by the relationship between precision and recall. Through manual calibration during the early testing process, the system recall rate of the forward mining truck and the target mining truck within the visual range is calculated under the current algorithm limit. In this embodiment of the application, the confidence score can be used as the harmonic mean of the perception accuracy and recall rate to comprehensively evaluate the performance of the perception system. In this embodiment of the application, the harmonic mean of the precision and recall rate can be used as a proxy indicator of the confidence level of the second distance:

[0129]

[0130] Select Q perception The threshold with the highest score determines the best confidence of the system, that is, the confidence of the second distance. perception A higher score indicates a higher confidence level based on the perception of the perception sensor.

[0131] The specific process is as follows:

[0132] 1. Data preparation: Collect labeled data sets, including positive and negative samples.

[0133] 2. Model prediction: Use sensor-based perception to predict the data set and output a confidence score for each sample.

[0134] 3. Calculate metrics: Calculate precision and recall based on different confidence thresholds.

[0135] 4. Draw a curve: Draw a precision-recall curve and select an appropriate threshold.

[0136] 5. Evaluate confidence: through Q perception The score or other relevant indicators evaluate the confidence level of the perception based on the perception sensor.

[0137] Calculating the confidence of the second distance through the relationship between precision and recall can help the autonomous driving system dynamically adjust the confidence threshold in different scenarios, thereby optimizing perception performance. perception The score is an effective metric for balancing precision and recall, and thus assessing the confidence level of the second distance. This embodiment of the present application balances perception accuracy and recall by setting a confidence level for the second distance, reducing the misidentification of obstacles due to perception errors and the resulting misapplication of brakes. This application improves both form efficiency and the lifespan of the vehicle's mechanical structure.

[0138] In addition, during the data collection and optimization process, information on specific road network sections (such as dust-prone areas) and severe weather (such as rain, snow, and fog) is added and associated with the confidence level of the second distance, and adaptive calculations are performed based on the actual vehicle location and weather data.

[0139] The embodiment of the present application introduces the confidence level of the first distance and the confidence level of the second distance to make value judgments on the distance between the operating machines in the road network coordinate system defined by the mining area and the distance between the operating machines obtained by the perception sensors. Through mathematical modeling, the fused data is output to improve the robustness of the distance prediction results.

[0140] When obtaining the confidence level of the first distance and the confidence level of the second distance, the electronic device inputs the first distance, the confidence level of the first distance, the second distance and the confidence level of the second distance into the target distance determination model, and obtains the distance prediction result between the first working machine and the second working machine output by the target distance determination model.

[0141] In one embodiment, the target distance determination model estimates the relative distance x to the road network coordinate system of the working machine. position And the perception output distance x perception , mathematical modeling is performed, and the distance prediction result z between the operating machines after data fusion is output based on the principle that the long-distance road network coordinate system estimation accounts for a large proportion and the close-range perception accounts for a large proportion. For example, the target distance determination model can be mathematically modeled based on the linear regression modeling method. In the embodiment of the present application, the input and output variables are defined as follows: x represents the relative distance between the operating machines estimated based on the road network coordinate system (first distance); y represents the relative distance between the operating machines based on the mutual perception output of the perception sensor (second distance); z represents the distance prediction result between the operating machines; Q represents the confidence of the first distance; W represents the confidence of the second distance. Assume that the distance prediction result z between the operating machines is composed of a linear combination of the above-mentioned input variables:

[0142] z=F(β0,β1,β2,β3,β4,x,Q,y,W);

[0143] Where β0 is the bias term, representing the intercept of the model; β1, β2, β3, and β4 are weight coefficients learned from the training data, representing the impact of each input variable on the true relative distance between the two vehicles. The weight coefficients are estimated using the least squares method, and a loss function is constructed to minimize the error. To minimize the loss function, the optimal weight coefficients β0, β1, β2, β3, and β4 are determined. An analytical method is used to calculate the optimal weight coefficients β (substituting β0, β1, β2, β3, and β4), thereby obtaining the target distance determination model.

[0144] According to the optimal weight coefficient β, for the new input values x and y, we can get the predicted value of the distance between the operating machines as:

[0145] z pred =F(β,x,Q,y,W).

[0146] In other aspects of the embodiments of the present application, during the mathematical modeling and testing process, it was observed that when the predicted distance is the set distance boundary value D, the difference trend between the first distance x and the second distance y of the input source will jump, and the contribution value of the two to the predicted value will also change greatly. Therefore, the embodiment of the present application introduces a set distance boundary value D to dynamically adjust the contribution of the first distance x and the second distance y of the input source. When the predicted distance is less than the set distance boundary value D, it relies more on the relative distance (second distance) y between the operating machines based on the mutual perception output of the perception sensor. When the distance is greater than or equal to the set distance boundary value D, it relies more on the relative distance (first distance) x between the operating machines estimated based on the road network coordinate system.

[0147] Therefore, the embodiment of the present application sets the target distance determination model to include a first model and a second model.

[0148] The electronic device inputs the first distance, the confidence of the first distance, the second distance, and the confidence of the second distance into the first model to obtain an initial predicted distance; when the initial predicted distance is less than a set distance boundary value, the first distance, the confidence of the first distance, the second distance, and the confidence of the second distance are input into the second model to obtain a distance prediction result; when the initial predicted distance is greater than or equal to the set distance boundary value, the distance prediction result is determined to be the initial predicted distance.

[0149] Specifically, the first model can be constructed based on a conventional linear regression prediction model. The second model can be constructed based on a weighted linear regression prediction model. The electronic device inputs the first distance, the confidence level of the first distance, the second distance, and the confidence level of the second distance into the linear regression prediction model to obtain an initial predicted distance. If the initial predicted distance is less than a set distance boundary value, the first distance, the confidence level of the first distance, the second distance, and the confidence level of the second distance are input into the weighted linear regression prediction model to obtain a distance prediction result output by the weighted linear regression prediction model. If the initial predicted distance is greater than or equal to the set distance boundary value, the distance prediction result is determined to be the initial predicted distance output by the linear regression prediction model. The influence of the second distance on the distance prediction result in the linear regression prediction model is less than the influence of the second distance on the distance prediction result in the weighted linear regression prediction model. That is, the initial predicted distance of the linear regression prediction model is more dependent on the relative distance between the working machines (the first distance) estimated based on the road network coordinate system; the distance prediction result of the weighted linear regression prediction model is more dependent on the relative distance between the working machines (the second distance) based on the mutual perception output of the perception sensors. By setting the distance boundary value D, the bias of the target distance determination model in different distance ranges can be flexibly adjusted, thereby improving the accuracy of distance determination between operating machines.

[0150] In other aspects of the embodiments of the present application, the first model is constructed by using a linear regression prediction model, and the linear regression prediction model is trained by the following steps:

[0151] Step 51: Acquire multiple data sets, including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance.

[0152] Repeat the following steps until the set stop condition is reached:

[0153] Step 52: Input the first data set into the linear regression prediction model respectively to obtain the distance prediction value output by the linear regression prediction model.

[0154] Step 53: Calculate a loss function based on the distance prediction value and the distance label corresponding to the first data set.

[0155] Step 54: Adjust the weight coefficient of the linear regression prediction model based on the loss function, so as to adjust the influence of the first sample distance and the second sample distance on the distance prediction value according to the first sample distance.

[0156] The data set can be obtained based on the historical data of the operating machinery in the road network coordinate system. The electronic device inputs the first data set into the linear regression prediction model respectively to obtain the distance prediction value output by the linear regression prediction model. The first data set is any one of the multiple data sets. The electronic device calculates the loss function based on the distance prediction value and the distance label corresponding to the first data set. When the predicted distance is greater than or equal to the set distance value D, the loss function L1(β) is obtained. In one embodiment, the loss function L1(β) is calculated by the following formula:

[0157] Among them, y i is the distance prediction value of the i-th data set, y i ' is the distance label corresponding to the i-th dataset, and n represents the number of datasets. The loss function L1(β) is analytically optimized to obtain the optimal model parameter β. Based on the first sample distance, the effects of the first and second sample distances on the distance prediction value are adjusted. This makes the initial predicted distance of the linear regression prediction model more dependent on the relative distance between working machines (the first distance) estimated based on the road network coordinate system, thereby maximizing the predicted value.

[0158] In other aspects of the embodiments of the present application, the second model is constructed by using a weighted linear regression prediction model, and the weighted linear regression prediction model is trained by the following steps:

[0159] Step 61: Acquire multiple data sets, including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance.

[0160] Repeat the following steps until the set stop condition is reached:

[0161] Step 62: Input the first data set into the weighted linear regression prediction model respectively to obtain the distance prediction value output by the weighted linear regression prediction model; the first data set is any one of the multiple data sets.

[0162] Step 63: Calculate a loss function based on the distance prediction value, the distance label corresponding to the first data set, and a weighting factor.

[0163] Step 64: Adjust the weight coefficient of the weighted linear regression prediction model based on the loss function, so as to adjust the influence of the first sample distance and the second sample distance on the distance prediction value according to the second sample distance.

[0164] The data set can be obtained based on the historical data of the operating machinery in the road network coordinate system. The electronic device inputs the first data set into the linear regression prediction model respectively to obtain the distance prediction value output by the linear regression prediction model. The first data set is any one of the multiple data sets. The electronic device calculates the loss function based on the distance prediction value, the distance label corresponding to the first data set and the weighting factor. When the predicted distance is less than the set distance value D, the loss function L2(β, δ) is obtained. In one embodiment, the loss function L2(β, δ) is calculated by the following formula:

[0165] Among them, y i is the distance prediction value of the i-th data set, y i ' is the distance label corresponding to the i-th data set, n represents the number of data sets, δ i The weighting factor for the i-th dataset is calculated by analytically optimizing the loss function L2(β, δ) to obtain the optimal model parameters β and δ. Based on the second sample distance, the influence of the first and second sample distances on the distance prediction value is adjusted, so that the distance prediction result of the weighted linear regression prediction model is more dependent on the relative distance between the working machines (the second distance) based on the mutual perception output of the perception sensors, resulting in a prediction result that is closest to the true value. The weighting factor δ represents δ1 to δn.

[0166] By setting the boundary value D and the weighting factor δ, the bias of the model within different distance ranges can be flexibly controlled. The loss function can be optimized through continuous data training to obtain the final mathematical model, and the relative positions of the two workshops can be obtained through the mathematical model.

[0167] In other aspects of the embodiments of the present application, the target distance determination model is constructed using a machine learning model, and the machine learning model is trained by the following steps:

[0168] Step 71: Acquire multiple data sets, including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance.

[0169] Repeat the following steps until the set stop condition is reached:

[0170] Step 72: Input the first data set into the machine learning model respectively to obtain the distance prediction value output by the machine learning model; the first data set is any one of the multiple data sets.

[0171] Step 73: Calculate a loss function based on the distance prediction value and the distance label corresponding to the first data set.

[0172] Step 74: Adjust the network parameters of the machine learning model based on the loss function so that when the distance prediction value is greater than or equal to the set distance boundary value, the first sample distance and the second sample distance are adjusted according to the first sample distance to affect the distance prediction value; and when the distance prediction value is less than the set distance boundary value, the first sample distance and the second sample distance are adjusted according to the second sample distance to affect the distance prediction value.

[0173] The machine learning model in the embodiments of this application can employ machine learning models such as decision trees, support vector machines, and neural networks. This embodiment employs a convolutional neural network for mathematical modeling. First, multiple data sets are cleaned, normalized, and standardized, converted into a standard neural network input format, and then divided into a training set, a validation set, and a test set. The convolutional neural network is then used to determine network maturity, the number of neurons, and the activation function. The training set data is trained using a backpropagation algorithm, and network parameters are continuously adjusted to minimize the error in the calculated distance between the two vehicles. The validation set is used to monitor the model's generalization ability and prevent overfitting. The trained model is evaluated using the test set, and its performance is measured using mean squared error, precision, and recall, ultimately resulting in a suitable mathematical model. The trained convolutional neural network adjusts the influence of the first and second sample distances on the predicted distance value based on the first sample distance when the predicted distance value is greater than or equal to the set distance threshold. Furthermore, when the predicted distance value is less than the set distance threshold, the influence of the first and second sample distances on the predicted distance value is adjusted based on the second sample distance.

[0174] In other aspects of the embodiments of the present application, after step 200, inputting the first distance and the second distance into a target distance determination model and obtaining a distance prediction result between the first working machine and the second working machine output by the target distance determination model, the following steps are further included:

[0175] Step 300: Determine an autonomous driving strategy based on the distance prediction result.

[0176] Based on the path planning data set obtained from other working machines, the distance prediction results between the working machines in step 200 and the speed and acceleration of each vehicle are integrated as input parameters to formulate detailed logic and rules, and output behavioral decisions, speed limits and acceleration limits.

[0177] In one embodiment, the electronic device determines the autonomous driving strategy based on the distance prediction result as follows:

[0178] When it is detected that the distance between the vehicle in front and the ego vehicle is greater than dis1, the ego vehicle is judged to enter the cruise mode and automatically drive at the speed specified in the interval.

[0179] When the distance between the vehicle in front and the ego vehicle is detected to be between dis2 and dis1, the ego vehicle is determined to enter the following mode and the ego vehicle is controlled according to the speed and acceleration of the vehicle in front to follow the vehicle at the same speed. Both dis1 and dis2 are distance thresholds, and dis1>dis2.

[0180] When it is detected that the distance between the vehicle in front and the vehicle is less than dis2, the vehicle is judged to enter parking mode and, based on the speed and acceleration of the vehicle in front, slow braking or emergency braking commands are issued to stop the vehicle and avoid collision.

[0181] When it is detected that the distance between the oncoming vehicle and the ego vehicle is less than dis2, the ego vehicle is judged to enter the meeting mode and the speed and acceleration of the ego vehicle are reduced to ensure safe meeting and avoid collision.

[0182] In closed scenarios, the driving trajectory of vehicles can be finitely classified, that is, the orderly arrangement of different adjacent road network segments. Based on the analysis of various data of other vehicles on the driving path of the vehicle itself, the corresponding optimal behavioral decisions under different road scenarios can be selected. The method of the embodiment of the present application can be widely used in closed routes in various open-pit mining areas. It can realize low-cost and efficient automatic driving decisions of mining trucks based on the established road network information, reduce labor requirements, and improve production efficiency. In a closed environment, the embodiment of the present application makes flexible single-machine autonomous intelligent decisions based on the position set of each vehicle relative to the vehicle obtained by multi-data fusion, combined with their respective speed, acceleration, task and other information, to improve production efficiency.

[0183] In summary, please refer to Figure 6 , the following describes a method for determining the distance between intelligent working machines through an embodiment.

[0184] First, a road network coordinate system is proposed based on the working scenario in the mining area. On the one hand, the electronic device obtains the first travel time of the vehicle and the first path planning data to determine the first position and estimate the vehicle's position. Theoretical satellite positioning data is then obtained to cross-validate the vehicle's position and determine the confidence level of the first distance. The electronic device then implements multi-vehicle communication based on 5G, obtains estimated position information of other vehicles, and calculates the first distance between the vehicle and the other vehicles. On the other hand, the electronic device's perception sensor perceives the position information of other vehicles and obtains the second distance between the vehicle and the other vehicles; the confidence level of the second distance is then obtained. Finally, the electronic device performs data fusion calculations based on the first distance, the confidence level of the first distance, the second distance, and the second distance confidence level, and outputs a relative distance. Based on physical information such as relative distance, vehicle speed, and acceleration, autonomous driving decisions are made. This embodiment of the present application selectively performs data fusion and judgment based on multiple information sources, including basic sensor recognition input, vehicle planning information, satellite positioning inertial navigation information, and roadside unit information. This allows for real-time autonomous driving decision-making technology for mining trucks in closed and complex mining environments, without being dispatched by a dispatching system. The embodiments of the present application can implement real-time driving strategy formulation for mining trucks in a known closed mining environment to improve the safety and efficiency of mining operations.

[0185] The method for determining the distance between working machines in the embodiment of the present application has the following advantages:

[0186] Low resource consumption: By combining path planning data with known mining routes, the road network coordinate system is established and the position is estimated. While accurately obtaining the mining truck's position, no additional computing resources are consumed, saving processor power.

[0187] Low development costs: During the development process, the last 20% of performance gap often requires significant development effort. For example, addressing complex environmental conditions such as rain, snow, fog, and dust during perception, and network fluctuations at the intersection of base station coverage areas during GPS data acquisition, are time-consuming and labor-intensive. This application utilizes multidimensional data fusion to cleverly avoid these complex scenarios and maximize resource utilization.

[0188] High Security: Confidence-based modeling, comparing and integrating input data, significantly improves the safety of autonomous driving decisions compared to a single data source. Independent stand-alone decision modules also increase the system's safety redundancy.

[0189] High efficiency: Independent of the dispatching system, stand-alone decision-making optimizes vehicle driving and operation routes and choices, greatly improving operational efficiency.

[0190] The embodiment of the present application creates a new autonomous driving decision-making mode for mining trucks. In addition to the scheduling system, it increases the autonomous decision-making ability of a single machine, provides safety redundancy for key units, and enhances the safety of the system. Focusing on single machines and neglecting scheduling is the future development trend. Direct decision-making by a single machine eliminates the impact of network delays, improves response speed, avoids safety hazards caused by network fluctuations, and can greatly reduce the rate of dangerous accidents. By using the fusion of multiple information sources as input, it achieves full coverage of behavioral decisions beyond visual range and within the perception range, improving operation and production efficiency. At the same time, it saves development costs and hardware costs in a single direction, and makes full use of resources by corresponding different distances to different information sources. Multi-machine communication based on 5G networks eliminates the development cost and computing resources of the prediction module, and is simple and efficient. Data analysis is performed on different data sources to obtain confidence. According to the confidence, linear regression mathematical modeling based on weighted segmented optimization is performed to improve the robustness of the system.

[0191] Device embodiment

[0192] Please refer to Figure 7 On the other hand, an embodiment of the present application further provides a device for determining the distance between intelligent working machines, comprising:

[0193] An acquisition module 701 is configured to acquire a first distance between a first operating machine and a second operating machine, and a second distance between the first operating machine and the second operating machine;

[0194] A prediction module 702 is configured to input the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model;

[0195] Among them, the first distance is obtained based on the positions of the first working machine and the second working machine in the road network coordinate system; the second distance is obtained based on the mutual perception between the first working machine and the second working machine; the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result according to at least one of the first distance and the second distance.

[0196] Optionally, the target distance determination model includes a first model and a second model;

[0197] Inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model includes:

[0198] Inputting the first distance and the second distance into the first model to obtain an initial predicted distance;

[0199] When the initial predicted distance is less than a set distance boundary value, inputting the first distance and the second distance into the second model to obtain the distance prediction result;

[0200] The influence of the second distance on the distance prediction result in the first model is smaller than the influence of the second distance on the distance prediction result in the second model.

[0201] Optionally, the device further includes:

[0202] A determination module is used to determine that the distance prediction result is the initial prediction distance when the initial prediction distance is greater than or equal to a set distance boundary value.

[0203] Optionally, inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model includes:

[0204] The first distance, the confidence level of the first distance, the second distance, and the confidence level of the second distance are input into the target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model.

[0205] Optionally, the confidence level of the first distance is obtained by the following steps:

[0206] Acquire road network planning data of a target operating machine, the road network planning data including a plurality of road network points and theoretical satellite positioning data corresponding to each of the road network points, the target operating machine including at least one of the first operating machine and the second operating machine;

[0207] Acquiring real satellite positioning data of the target operating machine when it arrives at at least one of the road network points;

[0208] The confidence level of the first distance is determined based on a positioning error between the theoretical satellite positioning data and the actual satellite positioning data, where the positioning error is negatively correlated with the confidence level of the first distance.

[0209] Optionally, the confidence level of the first distance is determined based on a positioning error between the theoretical satellite positioning data and the actual satellite positioning data;

[0210] obtaining an error tolerance threshold, wherein the error tolerance threshold is negatively correlated with the reliability of the satellite positioning system in the target working machine;

[0211] The confidence level of the first distance is determined based on a difference between the positioning error and the error tolerance threshold, wherein the difference between the positioning error and the error tolerance threshold is negatively correlated with the confidence level of the first distance.

[0212] Optionally, the first model is constructed by using a linear regression prediction model, and the linear regression prediction model is trained by the following steps:

[0213] Acquire multiple data sets, the data sets including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance;

[0214] Repeat the following steps until the set stop condition is reached:

[0215] Inputting a first data set into the linear regression prediction model respectively to obtain a distance prediction value output by the linear regression prediction model; the first data set is any one of the multiple data sets;

[0216] Calculating a loss function based on the distance prediction value and the distance label corresponding to the first data set;

[0217] The weight coefficient of the linear regression prediction model is adjusted based on the loss function, so that the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the first sample distance.

[0218] Optionally, the second model is constructed by using a weighted linear regression prediction model, and the weighted linear regression prediction model is trained by the following steps:

[0219] Acquire multiple data sets, the data sets including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance;

[0220] Repeat the following steps until the set stop condition is reached:

[0221] Inputting a first data set into the weighted linear regression prediction model respectively to obtain a distance prediction value output by the weighted linear regression prediction model; the first data set is any one of the multiple data sets;

[0222] Calculating a loss function based on the distance prediction value, the distance label corresponding to the first data set, and a weighting factor;

[0223] The weight coefficient of the weighted linear regression prediction model is adjusted based on the loss function, so that the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the second sample distance.

[0224] Optionally, the target distance determination model is constructed using a machine learning model, and the machine learning model is trained through the following steps:

[0225] Acquire multiple data sets, the data sets including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance;

[0226] Repeat the following steps until the set stop condition is reached:

[0227] Inputting a first data set into the machine learning model respectively to obtain a distance prediction value output by the machine learning model; the first data set is any one of the multiple data sets;

[0228] Calculating a loss function based on the distance prediction value and the distance label corresponding to the first data set;

[0229] The network parameters of the machine learning model are adjusted based on the loss function so that when the distance prediction value is greater than or equal to the set distance boundary value, the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the first sample distance; and when the distance prediction value is less than the set distance boundary value, the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the second sample distance.

[0230] The distance determination device between the intelligent working machines includes a processor and a memory. The above-mentioned acquisition module 701, prediction module 702, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0231] The processor includes a kernel, which calls the corresponding program unit from the memory. There can be one or more kernels.

[0232] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0233] Figure 8 An example of a physical structure diagram of an electronic device is shown below. Figure 8 As shown, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communications bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other via the communications bus 840. The processor 810 may invoke logic instructions in the memory 830 to execute a method for determining the distance between intelligent working machines. The method includes: obtaining a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine; inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model; wherein the first distance is obtained based on the positions of the first working machine and the second working machine in a road network coordinate system; the second distance is obtained based on mutual perception between the first working machine and the second working machine; and the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result based on at least one of the first distance and the second distance.

[0234] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0235] On the other hand, the present application also provides a computer program product, which includes a computer program, which can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a method for determining the distance between intelligent working machines, the method including: obtaining a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine; inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model; wherein, the first distance is obtained based on the positions of the first working machine and the second working machine in the road network coordinate system; the second distance is obtained based on the mutual perception between the first working machine and the second working machine; the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result according to at least one of the first distance and the second distance.

[0236] On the other hand, the present application also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for determining the distance between intelligent working machines, the method comprising: obtaining a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine; inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model; wherein, the first distance is obtained based on the positions of the first working machine and the second working machine in a road network coordinate system; the second distance is obtained based on the mutual perception between the first working machine and the second working machine; the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result according to at least one of the first distance and the second distance.

[0237] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0238] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0239] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for determining the distance between intelligent working machines, characterized in that: include: Acquire a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine; Inputting the first distance and the second distance into a target distance determination model, and obtaining a distance prediction result between the first working machine and the second working machine output by the target distance determination model; Wherein, the first distance is obtained based on the positions of the first working machine and the second working machine in the road network coordinate system; The second distance is obtained based on mutual perception between the first working machine and the second working machine; the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result according to at least one of the first distance and the second distance.

2. The method for determining the distance between intelligent working machines according to claim 1, characterized in that: The target distance determination model includes a first model and a second model; Inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model includes: Inputting the first distance and the second distance into the first model to obtain an initial predicted distance; When the initial predicted distance is less than a set distance boundary value, inputting the first distance and the second distance into the second model to obtain the distance prediction result; The influence of the second distance on the distance prediction result in the first model is smaller than the influence of the second distance on the distance prediction result in the second model.

3. The method for determining the distance between intelligent working machines according to claim 2, characterized in that: The method further comprises: When the initial predicted distance is greater than or equal to the set distance boundary value, the distance prediction result is determined to be the initial predicted distance.

4. The method for determining the distance between intelligent working machines according to any one of claims 1 to 3, characterized in that: Inputting the first distance and the second distance into a target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model includes: The first distance, the confidence level of the first distance, the second distance, and the confidence level of the second distance are input into the target distance determination model to obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model.

5. The method for determining the distance between intelligent working machines according to claim 4, characterized in that: The confidence level of the first distance is obtained by the following steps: Acquire road network planning data of a target operating machine, the road network planning data including a plurality of road network points and theoretical satellite positioning data corresponding to each of the road network points, the target operating machine including at least one of the first operating machine and the second operating machine; Acquiring real satellite positioning data of the target operating machine when it arrives at at least one of the road network points; The confidence level of the first distance is determined based on a positioning error between the theoretical satellite positioning data and the actual satellite positioning data, where the positioning error is negatively correlated with the confidence level of the first distance.

6. The method for determining the distance between intelligent working machines according to claim 5, characterized in that: determining a confidence level of the first distance based on a positioning error between the theoretical satellite positioning data and the actual satellite positioning data; obtaining an error tolerance threshold, wherein the error tolerance threshold is negatively correlated with the reliability of the satellite positioning system in the target working machine; The confidence level of the first distance is determined based on a difference between the positioning error and the error tolerance threshold, wherein the difference between the positioning error and the error tolerance threshold is negatively correlated with the confidence level of the first distance.

7. The method for determining the distance between intelligent working machines according to claim 2, characterized in that: The first model is constructed by using a linear regression prediction model, and the linear regression prediction model is trained by the following steps: Acquire multiple data sets, the data sets including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance; Repeat the following steps until the set stop condition is reached: Inputting a first data set into the linear regression prediction model respectively to obtain a distance prediction value output by the linear regression prediction model; the first data set is any one of the multiple data sets; Calculating a loss function based on the distance prediction value and the distance label corresponding to the first data set; The weight coefficient of the linear regression prediction model is adjusted based on the loss function, so that the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the first sample distance.

8. The method for determining the distance between intelligent working machines according to claim 2, characterized in that: The second model is constructed by using a weighted linear regression prediction model, and the weighted linear regression prediction model is trained by the following steps: Acquire multiple data sets, the data sets including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance; Repeat the following steps until the set stop condition is reached: Inputting a first data set into the weighted linear regression prediction model respectively to obtain a distance prediction value output by the weighted linear regression prediction model; the first data set is any one of the multiple data sets; Calculating a loss function based on the distance prediction value, the distance label corresponding to the first data set, and a weighting factor; The weight coefficient of the weighted linear regression prediction model is adjusted based on the loss function, so that the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the second sample distance.

9. The method for determining the distance between intelligent working machines according to claim 1, characterized in that: The target distance determination model is constructed by using a machine learning model, and the machine learning model is trained by the following steps: Acquire multiple data sets, the data sets including a first sample distance between a first working machine and a second working machine, a confidence level of the first sample distance, a second sample distance between the first working machine and the second working machine, and a confidence level of the second sample distance; Repeat the following steps until the set stop condition is reached: Inputting a first data set into the machine learning model respectively to obtain a distance prediction value output by the machine learning model; the first data set is any one of the multiple data sets; Calculating a loss function based on the distance prediction value and the distance label corresponding to the first data set; The network parameters of the machine learning model are adjusted based on the loss function so that when the distance prediction value is greater than or equal to the set distance boundary value, the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the first sample distance; and when the distance prediction value is less than the set distance boundary value, the influence of the first sample distance and the second sample distance on the distance prediction value is adjusted according to the second sample distance.

10. A device for determining the distance between intelligent working machines, characterized in that: include: An acquisition module, configured to acquire a first distance between a first working machine and a second working machine, and a second distance between the first working machine and the second working machine; a prediction module, configured to input the first distance and the second distance into a target distance determination model, and obtain a distance prediction result between the first working machine and the second working machine output by the target distance determination model; Wherein, the first distance is obtained based on the positions of the first working machine and the second working machine in the road network coordinate system; The second distance is obtained based on mutual perception between the first working machine and the second working machine; the target distance determination model adjusts the influence of the first distance and the second distance on the distance prediction result according to at least one of the first distance and the second distance.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for determining the distance between intelligent working machines according to any one of claims 1 to 9 is implemented.

12. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for determining the distance between intelligent working machines according to any one of claims 1 to 9 is implemented.

Images