Dynamic scheduling method and system, storage medium and program product

Abstract

The invention provides a dynamic scheduling method and system, a storage medium and a program product. The dynamic scheduling method comprises the following steps: acquiring road conditions and environmental parameters of a surface mine and operation data and performance parameters of a plurality of traffic participants; a graph data structure is generated based on the road condition and the environment parameters of the surface mine and the operation data and the performance parameters of the multiple traffic participants, and each node in the graph data structure comprises related information of each traffic participant; inputting the graph data structure into a neural network model to obtain probability distribution of collision between each node in the graph data structure and other nodes; and adjusting a driving strategy of at least one of the plurality of traffic participants based on the probability distribution of collision between each node in the graph data structure and other nodes.

Description

Technical Field

[0001] The present disclosure relates to the technical field of vehicle scheduling, and in particular to a dynamic scheduling method, system, storage medium and program product. Background Art

[0002] In the field of open-pit mining, with the development of science and technology, unmanned driving technology has become an important means to improve work efficiency, reduce labor costs and enhance work safety. However, the complex traffic environment of open-pit mines, including changing terrain, dense transport vehicles, and frequent cross-operations, poses a huge challenge to the stable operation of unmanned driving systems. In related technologies, transportation scheduling systems based on static planning are difficult to adapt to these dynamic changes, especially when facing emergencies or urgent tasks, which often leads to greater safety hazards. Summary of the invention

[0003] A technical problem solved by the present disclosure is that the transportation scheduling system in the related art leads to great safety hazards.

[0004] According to one aspect of the present disclosure, a dynamic scheduling method is provided, including: obtaining road conditions and environmental parameters of an open-pit mine and operating data and performance parameters of multiple traffic participants; generating a graph data structure based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the multiple traffic participants, wherein each node in the graph data structure contains relevant information of each traffic participant; inputting the graph data structure into a neural network model to obtain a probability distribution of a collision between each node in the graph data structure and other nodes; and adjusting the driving strategy of at least one of the multiple traffic participants based on the probability distribution of a collision between each node in the graph data structure and other nodes.

[0005] In some embodiments, a graph data structure is generated based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the multiple traffic participants, including: generating path points of the trajectories of the multiple traffic participants based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the multiple traffic participants; determining path relationships between different path points based on the path points of the trajectories of the multiple traffic participants; and obtaining the graph data structure based on the path points of the trajectories of the multiple traffic participants and the path relationships between different path points, wherein the path relationships serve as edges between different nodes of the graph data structure.

[0006] In some embodiments, the multiple traffic participants include multiple mining trucks; based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the multiple traffic participants, path points of the trajectories of the multiple traffic participants are generated, including: generating a task route for each mining truck based on the scheduling task of each mining truck; generating an expected speed for each mining truck based on the load information of each mining truck; and generating path points of the planned trajectory of each mining truck based on the task route, expected speed and speed limit information of different roads of each mining truck.

[0007] In some embodiments, the road conditions include: traffic conditions and speed limit information of multiple roads in the open-pit mine; the environmental parameters include: weather condition parameters of the open-pit mine; the operating data include: current position information, speed information, acceleration information and load information of each mining truck; the performance parameters include: the driving speed of each mining truck when empty and the driving speed when fully loaded.

[0008] In some embodiments, the multiple traffic participants also include one or more electric shovels; the operating data includes: the speed, direction and historical position of the one or more electric shovels; based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the multiple traffic participants, the path points of the trajectories of the multiple traffic participants are generated, and also includes: based on the speed, direction, historical position of the one or more electric shovels, the future trajectory of the one or more electric shovels is generated through a Kalman filtering algorithm.

[0009] In some embodiments, the relevant information of each traffic participant includes: the identification of each traffic participant, the coordinates of each path point of each traffic participant, the priority of the current task of each traffic participant, the current speed information of each traffic participant, the current acceleration information of each traffic participant, the current load information of each traffic participant, the estimated time of arrival of each node by each traffic participant, and the path relationship between different path points.

[0010] In some embodiments, the graph data structure is input into a neural network model to obtain a probability distribution of each node in the graph data structure colliding with other nodes, including: processing the graph data structure sequentially through a graph convolutional network, a pooling layer, and a connection layer in the neural network model to obtain each node and other nodes that are likely to collide with each node; inputting each node and other nodes that are likely to collide with each node into a feedforward neural network of the neural network model to output a probability distribution vector of each node colliding with the other nodes; and based on the probability distribution vector of each node colliding with the other nodes, calculating the probability distribution of each node colliding with the other nodes through a normalized exponential function.

[0011] In some embodiments, adjusting the driving strategy of at least one of the multiple traffic participants based on the probability distribution of each node in the graph data structure colliding with other nodes includes: adjusting the driving strategy of the at least one traffic participant based on the probability distribution of each node in the graph data structure colliding with other nodes so that the probability of the at least one traffic participant colliding with other traffic participants among the multiple traffic participants is minimized and the benefit of the at least one traffic participant is maximized.

[0012] In some embodiments, the at least one traffic participant includes at least one mining truck; the income of the at least one traffic participant is calculated based on the total cargo capacity or total unloading capacity of each of the at least one mining truck from the loading point to the unloading point, the distance of each of the at least one mining truck from the loading point to the unloading point, the energy consumption per unit distance of each of the at least one mining truck when fully loaded, the waiting time of each of the at least one mining truck, and the energy consumption per unit time of each of the at least one mining truck when waiting.

[0013] According to another aspect of the present disclosure, a dynamic scheduling system is provided, including: a data acquisition unit for acquiring road conditions and environmental parameters of an open-pit mine and operating data and performance parameters of multiple traffic participants; a generation unit for generating a graph data structure based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the multiple traffic participants, wherein each node in the graph data structure contains relevant information of each traffic participant; a probability acquisition unit for inputting the graph data structure into a neural network model to obtain a probability distribution of a collision between each node in the graph data structure and other nodes; and an adjustment unit for adjusting the driving strategy of at least one of the multiple traffic participants based on the probability distribution of a collision between each node in the graph data structure and other nodes.

[0014] According to another aspect of the present disclosure, a dynamic scheduling system is provided, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute the above-mentioned dynamic scheduling method based on instructions stored in the memory.

[0015] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which computer instructions are stored. When the computer instructions are executed by a processor, the dynamic scheduling method as described above is implemented.

[0016] According to another aspect of the present disclosure, a computer program product is provided. The computer program product includes a computer program or instructions. When the computer program or instructions are executed by a processor, the dynamic scheduling method as described above is implemented.

[0017] In the above dynamic scheduling method, the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants are obtained; a graph data structure is generated based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants, wherein each node in the graph data structure contains relevant information of each traffic participant; the graph data structure is input into a neural network model to obtain the probability distribution of collision between each node in the graph data structure and other nodes; and the driving strategy of at least one of the multiple traffic participants is adjusted based on the probability distribution of collision between each node in the graph data structure and other nodes. This can reduce safety hazards.

[0018] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0020] The present disclosure may be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:

[0021] Figure 1 is a flow chart illustrating a dynamic scheduling method according to some embodiments of the present disclosure;

[0022] Figure 2 is a schematic diagram showing a mine site map according to some embodiments of the present disclosure;

[0023] Figure 3 is a schematic diagram showing a mine site map according to other embodiments of the present disclosure;

[0024] Figure 4 is a schematic diagram showing a graph structure according to some embodiments of the present disclosure;

[0025] Figure 5 is a schematic diagram showing a graph structure according to some other embodiments of the present disclosure;

[0026] Figure 6 is a schematic diagram showing a graph convolutional network extracting a relationship according to some embodiments of the present disclosure;

[0027] Figure 7 is a block diagram schematically illustrating a structure of a dynamic scheduling system according to some embodiments of the present disclosure;

[0028] Figure 8 is a structural block diagram schematically illustrating a dynamic scheduling system according to other embodiments of the present disclosure;

[0029] Fig. 9 It is a structural block diagram schematically showing a dynamic scheduling system according to other embodiments of the present disclosure. DETAILED DESCRIPTION

[0030] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless otherwise specifically stated.

[0031] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0032] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0033] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0034] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0035] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0036] Figure 1 is a flow chart illustrating a dynamic scheduling method according to some embodiments of the present disclosure.

[0037] In step S102, the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants are obtained.

[0038] For example, the traffic participants may include mining trucks and / or electric shovels (ie, electric excavators), etc. For another example, the traffic participants may also include water trucks or pedestrians, etc.

[0039] In some embodiments, the road conditions include: traffic conditions and speed limit information of multiple roads in the open-pit mine. For example, among the roads in the open-pit mine, is there any road that is not drivable due to weather reasons, or is there any road that requires adjustment of the maximum speed of mining trucks due to weather reasons.

[0040] In some embodiments, the environmental parameters include: weather condition parameters of the open-pit mine. The weather condition parameters may include weather condition parameters such as sunny, sandstorm, foggy, rainy, snowy, etc. For example, the vehicle needs to slow down due to reduced visibility caused by weather conditions such as sandstorm, foggy, etc.

[0041] In some embodiments, the operation data includes: current position information, speed information, acceleration information and load information of each mining truck. For example, the number of available mining trucks of different models, load capacity, real-time position, speed and acceleration of mining trucks, etc. can be obtained.

[0042] In some embodiments, the performance parameters include: the driving speed of each mining truck when it is empty and the driving speed when it is fully loaded. For example, the driving speeds of mining trucks of different models when they are empty and fully loaded are obtained.

[0043] In addition, in some other embodiments, the number of transfer points and unloading points in the mine area on that day can also be obtained, as well as the working status and efficiency of the electric shovel, for example, which electric shovels are available, the time required for the electric shovel to fill different types of mining trucks, etc.

[0044] In other embodiments, the above method may further include: obtaining the mineral reserves and production plan of the day to determine the task priority of the day.

[0045] Assuming that the main product of a mining company is iron ore, the following are the relevant data for the day:

[0046] Mineral reserves: (1) High-grade iron ore reserves: 1 million tons; (2) Medium-grade iron ore reserves: 1.5 million tons; (3) Low-grade iron ore reserves: 2 million tons.

[0047] Production plan: (1) 500,000 tons of high-grade iron ore needs to be produced to meet the needs of high-value customers; (2) 700,000 tons of medium-grade iron ore needs to be produced for general market demand; (3) 300,000 tons of low-grade iron ore needs to be produced for low-value market demand or as a by-product.

[0048] Based on the above data, here are the steps to prioritize your tasks:

[0049] Step 1: Analyze market demand and profit margin. For example, high-grade iron ore has strong market demand, high price, and the highest profit margin, so it has the highest priority; medium-grade iron ore has stable market demand, moderate price, and medium profit margin, so it has the second highest priority; low-grade iron ore has low market demand, low price, and the lowest profit margin, so it has the lowest priority.

[0050] Step 2: Consider reserves and production targets. For example, high-grade iron ore reserves are relatively small, but the production target is high, so production needs to be prioritized; medium-grade iron ore reserves and production targets are relatively balanced, so production can be carried out as planned; low-grade iron ore reserves are sufficient, but the production target is low, so production can be carried out on the premise of ensuring the production of other grades of ore.

[0051] Step 3, determine the priority of tasks for the day. For example, give priority to the production of high-grade iron ore to ensure that the needs of high-value customers are met; secondly, ensure the production of medium-grade iron ore to meet general market demand; and finally arrange the production of low-grade iron ore based on remaining resources and time.

[0052] In step S104, a graph data structure is generated based on the road conditions and environmental parameters of the open-pit mine and the operation data and performance parameters of multiple traffic participants. Here, each node in the graph data structure contains relevant information of each traffic participant.

[0053] In some embodiments, step S104 includes: generating path points of the trajectories of multiple traffic participants based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants; determining the path relationship between different path points based on the path points of the trajectories of multiple traffic participants; and obtaining a graph data structure based on the path points of the trajectories of multiple traffic participants and the path relationship between different path points, wherein the path relationship serves as an edge between different nodes of the graph data structure.

[0054] For example, the following combination Figures 2 to 5 Describe in detail the process of generating graph data structures.

[0055] For example, Figure 2 As shown, it is assumed that there are L loading areas and UL unloading areas (L and UL are positive integers) in the mining area, where loading area A is equipped with m electric shovels and loading area B is equipped with n electric shovels (m and n are positive integers). There are parking spots in the mining area for charging and maintenance of mining trucks and electric shovels, with a total of T idle mining trucks parked (T is a positive integer). These mining trucks are numbered according to their load capacity. For example, T1-1 represents mining truck No. 1 with a load of 100 tons, T1-2 represents mining truck No. 2 with a load of 100 tons, T2-1 corresponds to mining truck No. 1 with a load of 200 tons, and so on. For another example, S represents an electric shovel used for shoveling materials, and each electric shovel has a unique label. As Figure 3 As shown, the arrowed lines in the figure represent roads between different locations. When mining trucks are fully charged and start a day's work, they mainly move along established routes. Among them, the routes marked with solid lines have a higher risk of collision and conflict due to frequent use, and the routes marked with dotted lines are less chosen.

[0056] Next, if Figure 4 As shown, planned trajectories are set for mining trucks T1-1, T1-2, T1-3, T1-4, T2-1, T2-2, T2-3 and T2-4 at a certain moment, and a predicted trajectory is set for the electric shovel S4.

[0057] As described above, the multiple traffic participants may include multiple mining trucks. For example, based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the multiple traffic participants, the path points of the trajectories of the multiple traffic participants are generated, including: generating the task route of each mining truck based on the scheduling task (e.g., dynamic scheduling task) of each mining truck; generating the expected speed of each mining truck (i.e., the expected actual speed) based on the load information of each mining truck; and generating the path points of the planned trajectory of each mining truck based on the task route, expected speed and speed limit information of different roads of each mining truck. In this way, by generating these path points, the planned trajectory of each mining truck can be generated.

[0058] For example, in the process of planning a trajectory, tasks and constraints can be specified, such as tasks and constraints (1) to (3) described below.

[0059] (1) Determine the dynamic scheduling task. The path information generated by the dynamic scheduling system can be obtained, which usually includes a series of waypoints or road sections, as well as related task requirements (e.g., arrival time, load limit, etc.).

[0060] (2) Determine the speed limit of the road. The speed limit information of different roads can be obtained based on electronic maps or real-time traffic information.

[0061] (3) Determine the load and estimated speed. The actual speed of the vehicle on different road sections can be estimated based on the vehicle's load and power performance. The vehicle's dynamic parameters such as acceleration and deceleration can be used as a reference.

[0062] For example, the trajectory planning process includes steps a1 to a4.

[0063] In step a1, the path is refined. The path generated by the dynamic scheduling task is refined into a series of specific road sections, each of which has a clear starting point, end point, road speed limit and expected speed.

[0064] In step a2, speed planning. On each road section, the target speed curve of the vehicle on the road section is determined according to the road speed limit and the expected speed. This can refer to the acceleration change rate of the vehicle to avoid sudden acceleration and deceleration and improve riding comfort.

[0065] In step a3, time planning, the estimated driving time of the vehicle on each road section can be calculated based on the target speed curve and the length of the road section. This helps to ensure that the vehicle can arrive on time according to the mission requirements.

[0066] In step a4, trajectory generation. The results of velocity planning and time planning are combined to generate the planned trajectory of the vehicle. This can include a series of position, velocity and time information, which can be represented as a continuous curve or a series of discrete points.

[0067] In this way, through the above steps a1 to a4, trajectory planning is performed.

[0068] Furthermore, the multiple traffic participants may also include one or more electric shovels. The operating data includes: the speed, direction and historical position of one or more electric shovels. For example, based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants, the path points of the trajectories of multiple traffic participants are generated, and it also includes: based on the speed, direction, and historical position of one or more electric shovels, the future trajectory of one or more electric shovels is generated through the Kalman filter algorithm. In other words, the predicted trajectory of the electric shovel (for example, electric shovel S4) is the future trajectory preliminarily generated by the Kalman filter based on the speed, direction, and historical position of the electric shovel.

[0069] Here, the Kalman filter algorithm is a recursive state estimation method that updates the optimal estimate of the state through the system model and the measurement value. In a linear dynamic system, the state transition and observation model of the system are linear, and the system noise and observation noise are both Gaussian distributed and independent of each other. The Kalman filter algorithm is an algorithm known to those skilled in the art and will not be described in detail here.

[0070] like Figure 4 As shown, it is assumed that the planned trajectory data of each traffic participant can be represented as a sequence of coordinate points:

[0071] P ij ={(x i1 ,y i1 ),(x i2 ,y i2 ),…,(x ij ,y ij ),…,(x im ,y im )},i=1,2,…,n, (1)

[0072] Among them, P ij represents the path point of the ith traffic participant at time point j (i.e., the jth path point of the ith traffic participant), (x ij ,y ij ) represents the coordinate position of the i-th traffic participant at time point j, where i, j and n are positive integers.

[0073] In addition, the task priority is set for the current task of each mining truck according to the task plan of the day. For example, if the current demand for ore in unloading area A is more urgent, the task priority leading to unloading area A is relatively high. The initial path points of the mining truck are extracted as a graph structure, such as Figure 5 As shown, the data of each node in the graph contains the following information:

[0074] H={ID,P,Pr,V,A,L,T}, (2)

[0075] Among them, ID is the identifier of each traffic participant (such as a mining truck or other dynamic traffic participant) (i.e., the mark is the unique identifier of each traffic participant), P is the coordinate of each path point of each traffic participant, Pr is the priority of the current task of each traffic participant, V is the current speed information of each traffic participant, A is the current acceleration information of each traffic participant, L is the current load information of each traffic participant, and T is the estimated time for each traffic participant to arrive at each node (i.e., the timestamp corresponding to each node).

[0076] The planned trajectory of the mining truck is represented as a graph data structure (i.e., graph structure) G = (H, E), where H represents the node set of the graph, that is, each coordinate point is regarded as a node in the graph, and E represents the edge set of the graph, that is, the edge between the nodes, which represents the path relationship between the nodes. For example, if the first node and the second node are adjacent on the planned trajectory of the mining truck, there is an edge between them. For example, an n×n adjacency matrix A can be used to represent the graph structure, where if there is an edge from node i to node j, then Aij = 1, and if there is no edge from node i to node j, then Aij = 0. Therefore, the adjacency matrix A is a matrix composed of 0 and 1.

[0077] Each node represents a discrete coordinate point in the planned trajectory of a mining truck or other traffic participant. Specifically, the coordinate position of each node can be expressed as (x ij ,y ij ), where x ij and ij Represent the horizontal and vertical coordinates of the node respectively.

[0078] In this way, each node in the graph data structure contains relevant information of each traffic participant. The relevant information of each traffic participant includes: the identification of each traffic participant, the coordinates of each path point of each traffic participant, the priority of the current task of each traffic participant, the current speed information of each traffic participant, the current acceleration information of each traffic participant, the current load information of each traffic participant, the estimated time of arrival of each traffic participant at each node (e.g., the node) (i.e., the timestamp corresponding to each node) and the path relationship between different path points.

[0079] In step S106, the graph data structure is input into the neural network model to obtain the probability distribution of collision between each node in the graph data structure and other nodes.

[0080] In some embodiments, step S106 includes: processing the graph data structure in sequence through the graph convolutional network, pooling layer and connection layer in the neural network model to obtain each node and other nodes that may collide with each node; inputting each node and other nodes that may collide with each node into the feedforward neural network of the neural network model to output a probability distribution vector of each node colliding with other nodes; and based on the probability distribution vector of each node colliding with other nodes, calculating the probability distribution of each node colliding with other nodes through a normalized exponential function.

[0081] In practical applications, there are a large number of traffic participants, and the nodes representing the paths contain many elements. Depending on the number of traffic participants, each scheduling task may have different graph structures at different times. Therefore, node features can be learned and propagated through GCN (Graph Convolutional Network). Through multi-layer graph convolution operations, the features of the nodes will gradually aggregate and update, thereby capturing the relationship and dynamic information between traffic participants in subsequent processing. GCN will learn the feature representation of each node, which contains the node's own information and its relationship with neighboring nodes. These features can be used to predict the probability of collision between traffic participants. Here, graph convolutional network is a technology known to those skilled in the art.

[0082] The graph convolutional network is an adapted version of the convolutional neural network for graph encoding, and is used to process graph data tasks, such as node classification, graph classification, and link prediction. In the embodiments of the present disclosure, the graph convolutional network can be used to predict link relationships, thereby predicting collision conflicts between different traffic participants through subsequent processing.

[0083] For example, given a graph with n nodes, we can use an n×n adjacency matrix A to represent the graph structure. If there is an edge from node i to node j, then A ij = 1. In a GCN with L layers (L is a positive integer), if the input vector of the i-th node in the l-th layer is represented as The output vector is represented as The graph convolution operation can be written as:

[0084]

[0085] Among them, W (l) is the linear transformation matrix, b (l) is the bias term, and σ is a nonlinear function (e.g., ReLU function). (l) and b (l) Parameters that the model automatically learns and adjusts during the training process.

[0086] In some cases, it is possible to normalize the data before passing it into each nonlinear layer of the graph convolutional network and add self-loops to each node in the graph:

[0087]

[0088] in, I is the n×n identity matrix. is the degree of node i in the graph.

[0089] By stacking L layers of this operation, we can get a deep GCN network, such as Figure 6 As shown, you can use To represent the vector of the input GCN network, and use To represent the vector output from the GCN network.

[0090] Let H = [h1,…,h n ] represents the path, where h i is the i-th path point and corresponds to two intervals in the path: H s =[h s1 ,…,h sr ] and H o =[h o1 ,…,h ou ].like Figure 6 As shown, h sent is any node among all nodes, Hs represents the node connected to node h sent The first part of the node, Ho can be represented by the node h sent The following part of the nodes, the nodes in the interval Hs and Ho are all connected to the node h sent Nodes that may collide.

[0091] In this way, Figure 6 As shown, the graph data structure is processed sequentially through the graph convolution network GCN, the pooling layer, and the concatenation layer (e.g., the fully connected layer) in the neural network model to obtain each node and other nodes that may collide with each node. In the neural network model, the goal of relationship extraction is to predict the relationship r∈R between entities (R is a predefined relationship set, i.e., conflict) or "no relationship".

[0092] exist Figure 6 In the figure, the overall architecture is shown on the left, and the detailed graph convolution calculation of the path points is shown on the right. After applying L layers of GCN on the path vector, we can get the hidden representation of each collision conflict relationship, which is directly affected by the path points of several edges in the graph structure. In order to use these representations to extract relations, we can get the following sentence representation (see Figure 6 Left side):

[0093] h sent =f(h (L) )=f(GCN(h (0) )B, (5)

[0094] Among them, h (l) represents the implicit representation of the output vector of the lth layer of GCN, and f is the maximum pooling function that maps from n output vectors to the relationship vector. The path points with close distances in the figure are usually the core of relationship extraction. Therefore, as shown below, from h (l) The corresponding representation h is obtained s , the corresponding representation h can be obtained in a similar way o .

[0095]

[0096] The representations used for classification are obtained by concatenating the waypoint representations and inputting them into a feed-forward neural network (FFNN):

[0097] h final =FFNN([h sent ;h s ;h o ]). (7)

[0098] Then put this h final The final representation is input into the linear layer, and then the Softmax function (i.e., normalized exponential function) is used for operation to obtain the probability distribution of the relationship, that is, the probability distribution of conflicts between different nodes.

[0099] That is to say, each node and other nodes that may collide with each node are input into the feedforward neural network of the neural network model to output the probability distribution vector of each node colliding with other nodes. Then, based on the probability distribution vector of each node colliding with other nodes, the probability distribution of each node colliding with other nodes is calculated through the normalized exponential function (Softmax function).

[0100] Here, the probability distribution vector is a vector whose elements are real numbers. The Softmax function is a mathematical function that is mainly used to convert a set of arbitrary real numbers into real numbers representing probability distribution. Its core function is to map a set of real values ​​into the interval (0,1) and ensure that the sum of these values ​​is 1, so that it can be interpreted as a probability distribution. Here, the Softmax function is used to calculate the probability value of each node colliding with other nodes, and these probability values ​​constitute the probability distribution.

[0101] In the GCN model, a hidden layer with multiple nodes (e.g., 200 nodes) is used as the output feed-forward layer. For example, 2 GCN layers and 2 feed-forward layers (FFNN) are used in the experiment. Relu is used as the activation function for all nonlinear layers in GCN, and standard maximum pooling operations are used in all pooling layers.

[0102] For training, stochastic gradient descent is used, the initial learning rate is 1.0, and the exponential decay learning rate Lr is set to 0.001. Based on the learned node feature representation, a regressor can be used to predict the probability of collision.

[0103] In step S108, the driving strategy of at least one of the multiple traffic participants is adjusted based on the probability distribution of collision between each node in the graph data structure and other nodes.

[0104] In some embodiments, step S108 includes: based on the probability distribution of each node in the graph data structure colliding with other nodes, adjusting the driving strategy of at least one traffic participant so that the probability of at least one traffic participant colliding with other traffic participants among multiple traffic participants is minimized and the benefit of at least one traffic participant is maximized.

[0105] For example, for graph structures with a high probability of collision, several adjustment strategies can be used to ensure that the scheduling tasks with high priority remain unchanged. For example, (1) the scheduling tasks of mining trucks with low task priorities (for example, mining trucks with task priorities lower than the priority threshold, or mining trucks with lower priorities than other mining trucks) are adjusted to change the planned trajectory of the mining truck, such as delaying its departure time. For another example, let the mining truck with low priority wait in place and calculate its waiting time. For another example, if there are other paths that can be detoured, adjust its planned path and calculate the time required for the additional paths it travels.

[0106] When the graph structure has no collision risk and is sparse, the scheduling task of mining trucks can be added to improve the overall work efficiency. Figure 4 As shown, there are 3 electric shovels in loading area A. When mining truck T1-2 arrives at loading area A, if each mining truck needs an electric shovel to load it, and there are idle mining trucks at the parking point, a scheduling task for the mining truck will be added to the path where mining truck T1-2 is located.

[0107] In some embodiments, after each dynamic adjustment of the scheduling task, the graph structure may be regenerated, and several graph structures at different times may be generated each time. For example, a total of five graph structures may be generated each time, for the current time, 10 seconds later, 20 seconds later, 30 seconds later, and 40 seconds later. The process of step S106 is repeated until a driving strategy with the lowest risk (e.g., no risk) and the highest benefit is obtained, that is, the final scheduling solution is obtained.

[0108] The at least one traffic participant includes at least one mining truck.

[0109] In some embodiments, the benefit of the at least one traffic participant is calculated based on the total cargo capacity or total unloading capacity of each of the at least one mining trucks from the loading point to the unloading point, the distance of each of the at least one mining trucks from the loading point to the unloading point, the energy consumption per unit distance of each of the at least one mining trucks when fully loaded, the waiting time of each of the at least one mining trucks, and the energy consumption per unit time of each of the at least one mining trucks when waiting.

[0110] For example, each time a new dispatch task is generated, the total revenue E of t (t is a positive integer) mining trucks is calculated. total for

[0111]

[0112] in, is the total cargo load or total unloading volume of the k-th mining truck from loading point i to unloading point j, is the distance of the kth mining truck from loading point i to unloading point j, Q f,k is the energy consumption per kilometer (i.e., energy consumption per unit distance) of the kth mining truck when fully loaded, t k is the waiting time of the kth mining truck, Q w,k is the energy consumption per unit time when the kth mining truck is waiting, where k is a positive integer. k Q w,k is the energy consumption loss of the k-th mining truck due to waiting. It should be noted that these parameters may vary depending on the type, load and loading task of the k-th mining truck and whether waiting is required.

[0113] In the specific scheduling scenario for mining truck operations, given that its scheduling tasks often have a high degree of priority clarity and execution purpose, when the collision risk introduced by the new scheduling task is at a low threshold and the expected benefits are maximized, a dynamic scheduling strategy combined with refined trajectory prediction technology can be used to achieve collaborative optimization. This process can fully take into account the unique nature of mining truck operations: compared with general commercial vehicle environments, the operation of mining trucks in mining areas enjoys a higher priority for road use rights, resulting in other traffic participants being more inclined to take active avoidance measures when facing potential path conflicts, such as planning detours in advance or waiting in place to ensure the uninterrupted and efficient operation of mining trucks. Therefore, by using the collaborative prediction mechanism of scheduling tasks to accurately predict and optimize the travel trajectories of other traffic participants, precise adjustment of detour or avoidance strategies can be achieved.

[0114] So far, a dynamic scheduling method of some embodiments of the present disclosure is provided. The dynamic scheduling method includes: obtaining the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants; generating a graph data structure based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants, wherein each node in the graph data structure contains relevant information of each traffic participant; inputting the graph data structure into a neural network model to obtain the probability distribution of collision between each node in the graph data structure and other nodes; and adjusting the driving strategy of at least one of the multiple traffic participants based on the probability distribution of collision between each node in the graph data structure and other nodes. This dynamic scheduling method can reduce safety hazards and improve transportation efficiency.

[0115] As a key technology to solve traffic congestion in open-pit mines and optimize resource allocation, dynamic task scheduling can dynamically adjust the scheduling allocation of different vehicles and different time periods according to real-time traffic conditions, operational needs and safety requirements. This mechanism can not only effectively alleviate traffic pressure and reduce vehicle waiting time, but also flexibly schedule according to task priority to ensure that key operations are executed in a timely manner. More importantly, dynamic scheduling management also fully considers the real-time location and movement status of other traffic participants (such as electric shovels, other mining trucks and other transportation equipment), and reduces conflicts and interference between traffic participants through intelligent decision-making, thereby significantly improving the flexibility and efficiency of task scheduling.

[0116] In the complex and ever-changing traffic environment of open-pit mines, accurate prediction of the movement trajectories of other traffic participants can achieve efficient and safe transportation. Trajectory prediction methods in related technologies are often based on fixed models or historical data, which makes it difficult to cope with frequently changing factors in the actual environment. The trajectory prediction model combined with dynamic scheduling management in the embodiment of the present disclosure can dynamically adjust the parameters of the prediction model according to the real-time updated road right data, so that the prediction results are closer to reality. For example, when there are many vehicles on a road due to urgent scheduling tasks, the probability of other traffic participants frequently traveling on this road will be relatively reduced, so the trajectory prediction will take this change into account, thereby improving the accuracy of the prediction.

[0117] Furthermore, the mutual influence between traffic participants is also a factor in trajectory prediction. Dynamic scheduling management achieves collaborative prediction and optimization by comprehensively considering the movement trends, speed changes and potential conflict points of each traffic participant. This collaborative mechanism can not only warn of potential collision risks in advance, but also maximize the use of road resources through intelligent scheduling, thereby improving the overall efficiency and safety of mine transportation.

[0118] Therefore, the application of dynamic scheduling management and trajectory collaborative prediction methods provides strong support for open-pit mine unmanned driving technology. Through real-time perception, intelligent decision-making and collaborative optimization, the above dynamic scheduling method can solve the problems faced by the transportation scheduling system in related technologies as much as possible, injecting new vitality into the sustainable development of open-pit mines.

[0119] The embodiments of the present disclosure can provide a method for dynamic scheduling and trajectory collaborative prediction of open-pit mines, which can realize intelligent scheduling and collision prevention of mining trucks by combining dynamic scheduling algorithms and intelligent trajectory prediction algorithms, and maximize the safety, efficiency and cost-effectiveness of automated operations in mining scenarios. In this method, the possibility of collision is predicted based on the planned trajectory of the mining truck, which improves the operating safety of the mining truck; the operation priority is set according to the daily plan of the ore material, which improves the operating efficiency of the mining truck; all mining trucks are uniformly pre-scheduled, providing a completely unmanned operation system and saving human resources; the planned trajectory is dynamically adjusted over time, further improving the rationality and timeliness of decision-making; and the scheduling task is added for the trajectory prediction of other participants, which further improves the prediction accuracy.

[0120] Different from other vehicle models, the method of the disclosed embodiment is designed for autonomous driving of mining trucks in open-pit mining scenarios. For mining trucks, ensuring the total transportation volume of all mining trucks is a hard indicator, so the tasks of mining trucks can be prioritized according to the demand for mineral materials.

[0121] In the methods of some embodiments of the present disclosure, in order to judge whether the mining truck scheduling task is reasonable and whether there is any conflict in the planned trajectory, it is necessary to consider not only the two-dimensional trajectory coordinate points but also multiple information dimensions such as time. When there are many mining trucks and the tasks are more complex, the prediction accuracy through neural networks is higher.

[0122] When dynamically adjusting scheduling tasks, the core indicator should also be "ensuring the maximization of total transportation volume". For scheduling tasks with a higher probability of collision, a refined scheduling adjustment strategy can be adopted: Under the premise of ensuring the stable execution of high-priority scheduling tasks, the operation arrangements of low-priority mining trucks can be dynamically adjusted through intelligent algorithms. Specific strategies include: First, flexibly adjust the scheduling sequence of low-priority mining trucks, such as appropriately delaying their departure time to avoid peak hours; second, implement a waiting strategy in place to accurately calculate the waiting time to minimize the impact on the overall transportation plan; third, use a path optimization algorithm to find and plan alternative, low-conflict risk driving routes for low-priority mining trucks, while taking into account additional time and safety to ensure that the adjusted plan is both efficient and safe. Through such dynamic adjustments, the transportation efficiency of mining trucks can be maximized while ensuring safety, and the total volume requirements for mineral material transportation can be met.

[0123] Furthermore, when predicting the future trajectories of other traffic participants, the interactive impact of mining truck dispatching tasks can also be considered. By real-time monitoring and analyzing mining truck dispatching tasks and future planning paths, the future trajectories of other traffic participants can be accurately predicted and optimized, which improves the accuracy of the prediction model and realizes intelligent coordination and efficient management of global traffic flow. This design significantly enhances the robustness and practicality of the system, and provides strong support for the intelligent and refined management of open-pit mine traffic.

[0124] Figure 7 FIG. 1 is a block diagram schematically showing a structure of a dynamic scheduling system according to some embodiments of the present disclosure. Figure 7 As shown, the dynamic scheduling system includes: a data acquisition unit 702 , a generation unit 704 , a probability acquisition unit 706 and an adjustment unit 708 .

[0125] The data acquisition unit 702 is used to acquire the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants.

[0126] In some embodiments, road conditions include: traffic conditions and speed limit information of multiple roads in an open-pit mine; environmental parameters include: weather condition parameters of the open-pit mine; operating data include: current position information, speed information, acceleration information and load information of each mining truck; performance parameters include: the driving speed of each mining truck when empty and the driving speed when fully loaded.

[0127] The generating unit 704 is used to generate a graph data structure based on the road conditions and environmental parameters of the open-pit mine and the operation data and performance parameters of multiple traffic participants. Each node in the graph data structure contains relevant information of each traffic participant.

[0128] In some embodiments, the generation unit 704 is used to generate path points of the trajectories of multiple traffic participants based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants, determine the path relationship between different path points based on the path points of the trajectories of multiple traffic participants, and obtain a graph data structure based on the path points of the trajectories of multiple traffic participants and the path relationship between different path points, wherein the path relationship serves as edges between different nodes of the graph data structure.

[0129] In some embodiments, the plurality of traffic participants includes a plurality of mining trucks.

[0130] In some embodiments, the generation unit 704 is used to generate a task route for each mining truck based on the scheduling task of each mining truck, generate an expected speed for each mining truck based on the load information of each mining truck, and generate path points for the planned trajectory of each mining truck based on the task route, expected speed and speed limit information of different roads of each mining truck.

[0131] In some embodiments, the plurality of traffic participants further includes one or more electric shovels.

[0132] In some embodiments, the operational data includes: speed, orientation, and historical position of one or more shovels.

[0133] In some embodiments, the generation unit 704 is configured to generate future trajectories of one or more electric shovels based on the speed, orientation, and historical position of the one or more electric shovels by using a Kalman filter algorithm.

[0134] In some embodiments, the relevant information of each traffic participant includes: the identification of each traffic participant, the coordinates of each path point of each traffic participant, the priority of the current task of each traffic participant, the current speed information of each traffic participant, the current acceleration information of each traffic participant, the current load information of each traffic participant, the estimated time of arrival of each traffic participant at each node (i.e., the timestamp corresponding to each node) and the path relationship between different path points.

[0135] The probability obtaining unit 706 is used to input the graph data structure into the neural network model to obtain the probability distribution of each node in the graph data structure colliding with other nodes.

[0136] In some embodiments, the probability acquisition unit 706 is used to process the graph data structure through the graph convolution network, pooling layer and connection layer in the neural network model in sequence to obtain each node and other nodes that may collide with each node, and input each node and other nodes that may collide with each node into the feedforward neural network of the neural network model to output the probability distribution vector of each node colliding with other nodes, and based on the probability distribution vector of each node colliding with other nodes, calculate the probability distribution of each node colliding with other nodes through a normalized exponential function.

[0137] The neural network model disclosed in the present invention includes GCN, which may include an input layer, a graph convolution layer, an activation function, etc.

[0138] (1) Input layer

[0139] The input of GCN is a graph consisting of nodes and edges between them. Nodes represent entities in the graph, and edges represent connections between nodes. Each node has a feature vector that represents the feature information of the node. The graph can be represented by an adjacency matrix, in which each element represents the connection relationship between nodes.

[0140] (2) Graph Convolutional Layer

[0141] The core of GCN is the graph convolution layer, which updates the feature representation of a node by aggregating the neighbor information of the node. In each layer of graph convolution operation, the feature vector of the node is weighted and summed with the feature vectors of the neighboring nodes to obtain the aggregated feature vector. This process can be represented in the form of matrix multiplication.

[0142] (3) Activation function

[0143] After each graph convolution layer, a nonlinear activation function can be applied to introduce nonlinear transformations. Common activation functions include ReLU, Sigmoid, and tanh functions. Activation functions can increase the network's expressive power and nonlinear fitting capabilities.

[0144] GCN usually consists of multiple graph convolutional layers, each of which updates the feature representation of the node. Through the superposition of multiple layers, GCN can gradually capture deeper graph structure information. The training process of GCN usually adopts supervised learning to optimize network parameters by minimizing the loss function. Commonly used loss functions include cross entropy loss and mean square error loss.

[0145] The neural network model disclosed in the present invention also includes a pooling layer. The pooling layer is used to aggregate and reduce the dimension of the overall features of the graph. Commonly used pooling methods include maximum pooling and average pooling of the graph.

[0146] Furthermore, the neural network model of the present disclosure also includes an output layer for generating the final prediction result. The form of the output layer depends on the specific task and can be a classifier, a regressor, or a generator. For example, a regressor can be used to predict the probability of a collision.

[0147] As mentioned earlier, GCN is a deep learning model that uses graph structures for information propagation and feature learning. Through multi-layer graph convolution operations, the neural network model can capture the relationship and dynamic information between nodes and effectively predict the possibility of collision.

[0148] The adjustment unit 708 is used to adjust the driving strategy of at least one of the multiple traffic participants based on the probability distribution of each node in the graph data structure colliding with other nodes.

[0149] In some embodiments, the adjustment unit 708 is used to adjust the driving strategy of at least one traffic participant based on the probability distribution of collision between each node in the graph data structure and other nodes, so that the probability of collision between at least one traffic participant and other traffic participants among multiple traffic participants is minimized and the benefit of at least one traffic participant is maximized.

[0150] In some embodiments, the at least one traffic participant includes at least one mining truck.

[0151] In some embodiments, the benefit of at least one traffic participant is calculated based on the total cargo capacity or total unloading capacity of each of at least one mining truck from the loading point to the unloading point, the distance of each of at least one mining truck from the loading point to the unloading point, the energy consumption per unit distance of each of at least one mining truck when fully loaded, the waiting time of each of at least one mining truck, and the energy consumption per unit time of each of at least one mining truck when waiting.

[0152] So far, a dynamic dispatching system according to some embodiments of the present disclosure is provided. In the dynamic dispatching system, a data acquisition unit is used to acquire the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants; a generation unit is used to generate a graph data structure based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of multiple traffic participants, wherein each node in the graph data structure contains relevant information of each traffic participant; a probability acquisition unit is used to input the graph data structure into a neural network model to obtain the probability distribution of collision between each node in the graph data structure and other nodes; and an adjustment unit is used to adjust the driving strategy of at least one of the multiple traffic participants based on the probability distribution of collision between each node in the graph data structure and other nodes. The dynamic dispatching system can reduce safety hazards and improve transportation efficiency.

[0153] Embodiments of the present disclosure provide a dynamic scheduling method and system, which realizes the coordinated prediction of dynamic right of way and trajectory in an open-pit mine.

[0154] Firstly, the planned trajectory of each mining truck and the predicted trajectories of other traffic participants are converted into a graph data structure, where each node represents a discrete coordinate point on the path, providing a solid data foundation for subsequent collision conflict prediction.

[0155] Secondly, the above method and system cleverly apply graph convolutional networks to deeply learn the complex relationships between mining trucks and between mining trucks and traffic participants, and then accurately predict possible path conflicts. This method makes full use of the inherent characteristics of graph data, significantly improves the accuracy and efficiency of path conflict prediction, and provides a strong guarantee for the safe operation of mining trucks.

[0156] Furthermore, the above method and system propose a series of intelligent dynamic scheduling strategies that can flexibly respond to complex and changeable mining operating environments. When allocating scheduling tasks, factors such as the number of electric shovels, waiting time, collision risk, path length, and scheduling efficiency are considered. For scheduling tasks with high collision risks, the strategy ensures the stable execution of high-priority tasks and takes a variety of measures to avoid conflicts: first, adjust the scheduling of low-priority mining trucks, such as delaying departure; second, let low-priority vehicles wait in place to optimize waiting time; third, explore alternative paths to balance extra time and safety. In a sparse environment without collision risk, scheduling tasks can be intelligently added to maximize resource utilization and improve overall operating efficiency. These strategies together constitute an efficient and safe dynamic scheduling system for mining operations.

[0157] In addition, the above method and system can integrate the interactive prediction model between the dispatch trajectory of mining trucks and other traffic participants, and accurately predict and optimize the future trajectory of other traffic participants by real-time monitoring and analysis of the driving path and speed of mining trucks. This model not only takes into account the impact of mining trucks as a dominant factor on other vehicles or equipment, but also reversely evaluates the feedback effect of the potential behavior of other traffic participants on the dispatch efficiency and safety of mining trucks, thereby realizing the intelligent coordination and efficient management of global traffic flow. This design can improve the traffic fluency and safety of mining operation areas, and provide solid technical support for the automation and intelligent transformation of mines.

[0158] Furthermore, the above method and system also take into account the impact of mineral material demand on operation priorities. By reasonably setting operation priorities, it ensures that high-demand minerals can be mined, loaded and transported first, thereby effectively meeting the immediate needs of the production line, reducing inventory backlogs, and improving the response speed and flexibility of the entire mining operation chain.

[0159] In summary, the embodiments of the present disclosure have constructed a comprehensive and intelligent mine operation scheduling and traffic management system, which deeply integrates advanced data processing, graph neural network analysis, intelligent scheduling strategies and dynamic traffic prediction models, and realizes comprehensive optimization from mining truck scheduling to global traffic management. The system can not only accurately predict and reduce potential conflicts between mining trucks and other traffic participants to ensure operation safety, but also flexibly adjust operation priorities according to mineral material demand, and improve resource utilization efficiency and overall operation efficiency.

[0160] The planned trajectory of the mining truck and the predicted trajectory of other traffic participants are used as discrete points of the graph structure. For example, the planned trajectory points are the trajectory data of the next 7 seconds, sampled at a frequency of 10Hz, that is, a total of 70 nodes, and GCN is used for supervised training. GCN updates the feature representation of the node by aggregating the information of local neighbor nodes on the graph, thereby utilizing the relationship between nodes for information propagation and feature learning.

[0161] The embodiments of the present disclosure provide a method and system for dynamic scheduling and trajectory collaborative prediction in an open-pit mine, which ensures that mining trucks can drive safely while improving operating efficiency to save resources.

[0162] For example, multi-sensor fusion can improve the accuracy and robustness of detecting other traffic participants. Combining multiple sensors such as GPS (Global Positioning System), LiDAR, millimeter-wave radar, camera, V2X (vehicle to everything), etc., can achieve high-precision positioning and real-time perception of traffic participants in the mining environment (such as mining trucks, pedestrians, electric shovels, etc.), which is especially suitable for complex environments with high dust and large changes in lighting in mining scenes.

[0163] According to the task requirements and the status of mining trucks, a dynamic optimization algorithm is used to allocate scheduling tasks, and the priorities of mining trucks are dynamically set to optimize resource allocation, thus realizing intelligent task allocation and priority management of the method and system.

[0164] By combining high-precision maps with real-time task management, the dispatching path is planned and the trajectories of other traffic participants are predicted to ensure path safety. This realizes the path planning and trajectory prediction of the method and system.

[0165] In the method and system of the embodiments of the present disclosure, all trajectory points can be generated into a graph structure, and the probability of conflict can be judged more accurately through deep learning. The conflict situation in the loading and unloading area in the future is predicted based on the future trajectory of the mining truck scheduling path and other traffic participants in the loading and unloading area. The scheduling tasks are dynamically adjusted according to the conflict situation, and the predicted trajectory is collaboratively optimized according to the task priority. The possibility of collision is taken into account while allocating scheduling tasks, which improves the safety of mining trucks. The system can be in unmanned mode and does not require human participation. The scheduling task priority is adjusted according to the task status to ensure the maximum efficiency of unmanned operations.

[0166] In summary, the method and system of the embodiments of the present disclosure can dynamically adjust the dispatching task allocation in different time periods according to the real-time traffic conditions and operation requirements to ensure the efficiency and safety of mine transportation operations. By utilizing the real-time updated dispatching path and considering the interaction between traffic participants, the predicted trajectories of other traffic participants (such as electric shovels, transportation equipment, etc.) are dynamically adjusted, so as to more accurately predict the future movement trajectories of other traffic participants, realize collaborative prediction and optimization, and improve the efficiency and safety of mine transportation.

[0167] Figure 8 8 is a block diagram schematically showing a structure of a dynamic scheduling system according to some other embodiments of the present disclosure. The dynamic scheduling system includes a memory 810 and a processor 820. Among them:

[0168] The memory 810 can be a disk, a flash memory, or any other non-volatile storage medium. The memory is used to store Figure 1 The instructions in the corresponding embodiment.

[0169] The processor 820 is coupled to the memory 810 and can be implemented as one or more integrated circuits, such as a microprocessor or a microcontroller. The processor 820 is used to execute instructions stored in the memory, which can reduce safety risks and improve transportation efficiency.

[0170] In some embodiments, it is also possible to Fig. 9 As shown, the dynamic scheduling system 900 includes a memory 910 and a processor 920. The processor 920 is coupled to the memory 910 via a BUS 930. The dynamic scheduling system 900 can also be connected to an external storage device 950 via a storage interface 940 to call external data, and can also be connected to a network or another computer system (not shown) via a network interface 960, which will not be described in detail here.

[0171] In this embodiment, by storing data instructions in a memory and then processing the instructions through a processor, safety risks can be reduced and transportation efficiency can be improved.

[0172] In another embodiment, the present disclosure also provides a computer-readable storage medium (eg, a non-transitory computer-readable storage medium) having computer program instructions stored thereon, which are executed by a processor to implement Figure 1 The steps of the method in the corresponding embodiment. Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, devices, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0173] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0174] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0175] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0176] In some embodiments of the present disclosure, a computer program product is provided. The computer program product includes a computer program or instructions. When the computer program or instructions are executed by a processor, the dynamic scheduling method as described above is implemented.

[0177] In some embodiments of the present disclosure, a computer program is further provided, comprising: instructions, which, when executed by a processor, cause the processor to execute the dynamic scheduling method as described above.

[0178] So far, the present disclosure has been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0179] Although some specific embodiments of the present disclosure have been described in detail by way of example, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. It should be understood by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A dynamic scheduling method, comprising: Obtain road conditions and environmental parameters in open-pit mines as well as operating data and performance parameters of multiple traffic participants; Generate a graph data structure based on the road conditions and environmental parameters of the open-pit mine and the operation data and performance parameters of the plurality of traffic participants, wherein each node in the graph data structure contains relevant information of each traffic participant; Inputting the graph data structure into a neural network model to obtain a probability distribution of collision between each node in the graph data structure and other nodes; and The driving strategy of at least one of the multiple traffic participants is adjusted based on the probability distribution of each node in the graph data structure colliding with other nodes.

2. The dynamic scheduling method according to claim 1, wherein: A graph data structure is generated based on the road conditions and environmental parameters of the open-pit mine and the operation data and performance parameters of the plurality of traffic participants, including: generating waypoints of trajectories of the plurality of traffic participants based on road conditions and environmental parameters of the open-pit mine and operating data and performance parameters of the plurality of traffic participants; Based on the path points of the trajectories of the plurality of traffic participants, determining path relationships between different path points; and The graph data structure is obtained based on the path points of the trajectories of the multiple traffic participants and the path relationships between different path points, wherein the path relationships serve as edges between different nodes of the graph data structure.

3. The dynamic scheduling method according to claim 2, wherein: The plurality of traffic participants include a plurality of mining trucks; Based on the road conditions and environmental parameters of the open-pit mine and the operation data and performance parameters of the multiple traffic participants, path points of the trajectories of the multiple traffic participants are generated, including: Generate a task route for each mining truck based on the dispatching task of each mining truck; generating an estimated speed for each mining truck based on the load information for each mining truck; and Based on the mission route, expected speed and speed limit information of different roads of each mining truck, the waypoints of the planned trajectory of each mining truck are generated.

4. The dynamic scheduling method according to claim 2, wherein: The road conditions include: traffic conditions and speed limit information of multiple roads in the open-pit mine; The environmental parameters include: weather condition parameters of the open-pit mine; The operation data includes: current position information, speed information, acceleration information and load information of each mining truck; The performance parameters include: the driving speed of each mining truck when it is empty and the driving speed when it is fully loaded.

5. The dynamic scheduling method according to claim 3, wherein: The plurality of traffic participants also includes one or more electric shovels; The operating data includes: the speed, orientation and historical position of the one or more electric shovels; Based on the road conditions and environmental parameters of the open-pit mine and the operating data and performance parameters of the plurality of traffic participants, path points of the trajectories of the plurality of traffic participants are generated, further comprising: Based on the speed, direction, and historical position of the one or more electric shovels, a future trajectory of the one or more electric shovels is generated through a Kalman filter algorithm.

6. The dynamic scheduling method according to claim 2, wherein: The relevant information of each traffic participant includes: the identification of each traffic participant, the coordinates of each path point of each traffic participant, the priority of the current task of each traffic participant, the current speed information of each traffic participant, the current acceleration information of each traffic participant, the current load information of each traffic participant, the estimated time of arrival of each node by each traffic participant, and the path relationship between different path points.

7. The dynamic scheduling method according to claim 1, wherein: Inputting the graph data structure into a neural network model to obtain a probability distribution of collision between each node in the graph data structure and other nodes, including: Processing the graph data structure sequentially through a graph convolutional network, a pooling layer, and a connection layer in the neural network model to obtain each node and other nodes that may collide with each node; Input each node and other nodes that may collide with each node into the feedforward neural network of the neural network model to output a probability distribution vector of collision between each node and the other nodes; and Based on the probability distribution vector of the collision between each node and the other nodes, the probability distribution of the collision between each node and the other nodes is calculated by a normalized exponential function.

8. The dynamic scheduling method according to claim 1, wherein: Adjusting the driving strategy of at least one of the plurality of traffic participants based on the probability distribution of collision between each node in the graph data structure and other nodes includes: Based on the probability distribution of each node in the graph data structure colliding with other nodes, the driving strategy of the at least one traffic participant is adjusted so that the probability of the at least one traffic participant colliding with other traffic participants among the multiple traffic participants is minimized and the benefit of the at least one traffic participant is maximized.

9. The dynamic scheduling method according to claim 8, wherein: The at least one traffic participant comprises at least one mining truck; The benefit of the at least one traffic participant is calculated based on the total cargo capacity or total unloading capacity of each of the at least one mining trucks from the loading point to the unloading point, the distance of each of the at least one mining trucks from the loading point to the unloading point, the energy consumption per unit distance of each of the at least one mining trucks when fully loaded, the waiting time of each of the at least one mining trucks, and the energy consumption per unit time of each of the at least one mining trucks when waiting.

10. A dynamic scheduling system, comprising: A data acquisition unit for acquiring road conditions and environmental parameters of an open-pit mine as well as operating data and performance parameters of multiple traffic participants; A generating unit, configured to generate a graph data structure based on the road conditions and environmental parameters of the open-pit mine and the operation data and performance parameters of the plurality of traffic participants, wherein each node in the graph data structure contains relevant information of each traffic participant; A probability obtaining unit, used for inputting the graph data structure into a neural network model to obtain a probability distribution of collision between each node in the graph data structure and other nodes; and An adjustment unit is used to adjust the driving strategy of at least one of the multiple traffic participants based on the probability distribution of each node in the graph data structure colliding with other nodes.

11. A dynamic scheduling system, comprising: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the dynamic scheduling method according to any one of claims 1 to 9 based on instructions stored in the memory.

12. A computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions, when executed by a processor, implement the dynamic scheduling method according to any one of claims 1 to 9.

13. A computer program product, comprising a computer program or instructions, which implements the dynamic scheduling method according to any one of claims 1 to 9 when executed by a processor.

Images