Coal mine underground anti-collision method, equipment and medium

Abstract

The invention discloses a coal mine underground anti-collision method, equipment and a medium. The method comprises the following steps: acquiring a positioning data set of each target object; constructing a space-time frame according to the constructed time bucket and the positioning data set; wherein the space-time frame comprises a positioning data set of each target object corresponding to each time bucket; inputting the space-time frame into a pre-constructed trajectory prediction model to obtain a predicted trajectory segment; and determining whether the predicted trajectory segments corresponding to the target objects respectively intersect in the same time bucket or not so as to perform early warning and collision prevention in advance. According to the method, whether collision exists or not can be predicted through the space-time frame and the trajectory, early warning is carried out, and collision is prevented.

Description

Technical Field

[0001] This invention relates to the field of coal mine production, and in particular to a method, equipment and medium for preventing collisions in underground coal mines. Background Technology

[0002] In the underground coal mine production environment, workers and vehicles such as transport vehicles and continuous mining vehicles pass through various narrow roadways to carry out production activities. However, when vehicles pass through roadway intersections, the limited visibility caused by the intersections prevents the vehicle's onboard sensors from detecting the situation on the side of the intersection, significantly increasing the risk of collision. Therefore, existing collision avoidance systems have obvious limitations in the underground coal mine production environment. Summary of the Invention

[0003] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, this invention proposes a method for preventing collisions in underground coal mines, which can predict the existence of collisions through spatiotemporal frames and trajectory prediction, providing early warnings and preventing impending collisions.

[0004] The present invention also proposes equipment and media having the above-mentioned method for preventing collisions in underground coal mines.

[0005] A method for preventing collisions in underground coal mines according to a first aspect of the present invention includes: Obtain the location dataset for each target object; Based on the constructed time buckets and the location dataset, a spatiotemporal frame is constructed; wherein, the spatiotemporal frame includes: the location dataset of each target object corresponding to each time bucket; The spatiotemporal frame is input into a pre-built trajectory prediction model to obtain the predicted trajectory segment; Determine whether the predicted trajectory segments corresponding to each target object intersect within the same time bucket, so as to provide early warning and avoid collisions.

[0006] A method for preventing collisions in underground coal mines according to an embodiment of the present invention has at least the following beneficial effects: In underground coal mine production scenarios, the narrowness of underground roadways limits the visibility of vehicles or workers within the roadways. Therefore, trajectory prediction is performed on the running trajectories (i.e., spatiotemporal frames) of various target objects to predict their trajectories. If it is determined that the predicted trajectory segments corresponding to the target objects intersect within the same time bucket, it indicates that a collision is imminent, thus requiring early warning. Furthermore, the periods for uploading positioning data from various target objects in underground coal mine production are not uniform. Therefore, by using time buckets and constructing time frames, the time axes of the target objects' running trajectories are aligned to determine the position information of different target objects within each time bucket, ultimately determining whether an intersection or collision occurs. This invention predicts the existence of collisions through spatiotemporal frames and trajectory prediction, providing early warnings to prevent impending collisions.

[0007] According to some embodiments of the present invention, constructing a spatiotemporal frame based on the constructed time bucket and the location dataset includes: Build a time bucket; Filtering the location dataset according to the time bucket specifically includes: using the location data of at least two target objects in the same time bucket as the filtered location dataset. The location dataset after filtering each target object in the same time bucket and the corresponding time bucket are used as time frames.

[0008] According to some embodiments of the present invention, the location dataset includes: The location card number, timestamp, and coordinates of the target object; The time frame includes: time bucket number, the positioning card number corresponding to the target object, and the coordinates of the target object.

[0009] According to some embodiments of the present invention, the training method of the trajectory prediction model includes: Based on the acquired historical location dataset, historical spatiotemporal frames are obtained; A social attention mechanism is applied to each target object in the historical spatiotemporal frame to obtain additional features, and temporal features are extracted from the additional features to obtain temporal features; Based on the additional features and the temporal features, the feature vectors corresponding to each target object are obtained.

[0010] According to some embodiments of the present invention, the additional features obtained by applying a social attention mechanism to each target object in the historical spatiotemporal frame include: For each target object in each historical spatiotemporal frame, nearby objects of each target object are filtered according to a preset distance threshold; the relative distance and velocity difference between each target object and its corresponding nearby object are determined, and the influence factor of each nearby object relative to its corresponding target object is obtained. Based on the influence factors of each neighboring object relative to the corresponding target object, the neighboring influence factor of the target object, i.e., the additional feature, is obtained.

[0011] According to some embodiments of the present invention, the training method of the trajectory prediction model further includes: The loss function is calculated based on the historical spatiotemporal frames and the predicted trajectory segments output by the trajectory prediction model. The parameters of the trajectory prediction model are adjusted according to the loss function. The loss function includes: position prediction error and KL divergence loss of the latent variables of the variational autoencoder in the trajectory prediction model.

[0012] According to some embodiments of the present invention, it further includes: Based on the preset underground road network of the coal mine, taking the starting point of the predicted trajectory segment as the starting point, and using the predicted trajectory segment, the predicted position points in the predicted trajectory segment are sequentially corrected by a real-time hidden Markov map matching algorithm.

[0013] According to some embodiments of the present invention, the step of sequentially correcting each predicted position point in the predicted trajectory segment using a real-time hidden Markov map matching algorithm includes: Using the current predicted location as the center, a set of candidate points is obtained based on the underground road network of the coal mine; Based on the underground road network of the coal mine, determine the transition probability from the current predicted location point to each candidate point in the candidate point set; Based on the underground road network of the coal mine, determine the emission probability of each candidate point in the candidate point set to the next predicted position point of the predicted trajectory segment; Based on the transition probability and the emission probability, the sequence of candidate point combinations with the highest probability within the time period corresponding to the predicted trajectory segment is obtained, and the predicted trajectory segment is corrected accordingly.

[0014] An electronic device according to a second aspect of the present invention includes: Memory, used to store programs; A processor for executing a program stored in the memory, wherein when the processor executes the program stored in the memory, the processor is configured to perform the method as described in any one of the first aspects.

[0015] According to a third aspect of the present invention, a storage medium stores computer-executable instructions for performing the method as described in any one of the first aspects.

[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description

[0017] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0018] Figure 1 This is a schematic diagram of a scenario illustrating a method for preventing collisions in underground coal mines according to an embodiment of the present invention; Figure 2 This is a flowchart of a method for preventing collisions in underground coal mines, provided by another embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] It should be understood that in the description of the embodiments of the present invention, "multiple" (or "amounts") means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. If "first," "second," etc., are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0021] like Figure 1 As shown, taking the example of a worker and a vehicle about to collide at the intersection of roadways, this invention addresses the main technical problem it solves. The vehicle and worker are about to reach the intersection, but due to the narrowness of the roadway and limited visibility, the vehicle's sensors cannot detect the side of the vehicle's direction of travel (i.e., the situation between the worker and the intersection), and the worker cannot perceive that the vehicle is about to enter the roadway. Furthermore, the side of the vehicle is often obstructed by roadway obstacles, so installing sensors on the side of the vehicle cannot effectively detect vehicles or workers approaching the intersection. Therefore, there is a risk of collision between the vehicle and the worker at the intersection, which is a high-incidence area for collision accidents in continuous mining faces of underground coal mines.

[0022] like Figure 2 As shown, this embodiment of the invention provides a method for preventing collisions in underground coal mines, including: Step S100: Obtain the location dataset for each target object; Step S200: Construct a spatiotemporal frame based on the constructed time buckets and location datasets; wherein, the spatiotemporal frame includes: the location datasets of each target object corresponding to each time bucket; Step S300: Input the spatiotemporal frame into the pre-built trajectory prediction model to obtain the predicted trajectory segment; Step S400: Determine whether the predicted trajectory segments corresponding to each target object intersect within the same time bucket, so as to provide early warning and avoid collisions.

[0023] In underground coal mine production scenarios, the narrowness of underground roadways limits the visibility of vehicles and workers. Therefore, trajectory prediction is performed on the running trajectories (i.e., spatiotemporal frames) of various target objects to predict their trajectories. If the predicted trajectory segments corresponding to different target objects intersect within the same time bucket, it indicates that a collision is imminent, requiring early warning. Furthermore, the periods for uploading location data from different target objects in underground coal mine production are not uniform. Therefore, time buckets and constructed time frames are used to align the timelines of the target objects' running trajectories, thereby determining the positional information of different target objects within each time bucket, and ultimately determining whether an intersection or collision has occurred. This invention uses spatiotemporal frames and trajectory prediction to predict the possibility of collisions, providing early warnings to prevent impending collisions.

[0024] It's easy to understand that the target objects include vehicles and workers.

[0025] In one embodiment, in step S200, constructing a spatiotemporal frame based on the constructed time bucket and location dataset includes: Constructing a time bucket specifically includes: taking the earliest and latest time points in the location data of each target object as the total time period; dividing the total time period evenly according to a preset time step to form a time bucket; in one embodiment, the time step is the shortest period for uploading location data among several target objects; Based on the time bucket, the location dataset is filtered, specifically including: the location data in which at least two target objects exist in the same time bucket are used as the filtered location dataset. The location dataset after filtering each target object in the same time bucket and the corresponding time bucket are used as time frames.

[0026] It's easy to understand that a time bucket is actually a period of time.

[0027] In one embodiment, the location dataset includes: The location card number, timestamp, and coordinates of the target object; The time frame includes: time bucket number, the positioning card number corresponding to the target object, and the coordinates of the target object.

[0028] It should be noted that a time frame is actually a dataset of the location of several target objects within the corresponding time bucket, which includes the location information of each target object to align with the timeline of the target object's trajectory.

[0029] In one embodiment, the training method for the trajectory prediction model includes: Based on the acquired historical location dataset, historical spatiotemporal frames are obtained; A social attention mechanism is applied to each target object in the historical spatiotemporal frame to obtain additional features, and temporal features are extracted from the additional features to obtain temporal features; Based on additional features and temporal features, feature vectors corresponding to each target object are obtained.

[0030] It is easy to understand that obtaining historical spatiotemporal frames is similar to obtaining spatiotemporal frames. The difference is that historical spatiotemporal frames are based on a large amount of historical data, namely the complete trajectory of each target object, while spatiotemporal frames are based on real-time data.

[0031] In one embodiment, the trajectory prediction model uses the SocialVAE model, which employs a time-dimensional variational autoencoder (VAE) and recurrent neural network architecture, uses a social attention mechanism, uses the Adam optimizer to optimize and adjust the learning rate, and updates the model parameters through backpropagation algorithm + stochastic gradient descent.

[0032] The feature processing layer in the SocialVAE model originally extracted temporal features to predict features for future time. This embodiment adds additional features to capture the interaction between different target objects, and enhances trajectory prediction by focusing on the state of adjacent target objects (including relative distance and speed difference).

[0033] In one embodiment, the first 8 spatiotemporal frames of the target object are used as input to the trajectory prediction model, and the predicted trajectory for the next 12 frames is output.

[0034] In one embodiment, a social attention mechanism is applied to each target object in the historical spatiotemporal frame to obtain additional features, including: For each target object in each historical spatiotemporal frame, nearby objects of each target object are filtered according to a preset distance threshold; the relative distance and velocity difference between each target object and its corresponding nearby object are determined, and the influence factor of each nearby object relative to its corresponding target object is obtained. Based on the influence factors of each neighboring object relative to the corresponding target object, the neighboring influence factor, i.e., the additional feature, of the target object is obtained.

[0035] It's easy to understand that neighboring objects are actually other target objects that are near a target object.

[0036] In one embodiment, the neighboring influence factor of the target object is obtained by weighted averaging of the influence factors of each neighboring object relative to the corresponding target object.

[0037] In one embodiment, filtering the neighboring objects of each target object according to a preset distance threshold includes: drawing a circle with a target object as the center and the distance threshold as the radius, and the other target objects included within the range of the circle are the neighboring objects of the target object; The relative distance between the target object and its corresponding neighboring objects refers to the total distance traveled from the target object to the neighboring object via various paths, taking into account the underground road network of the coal mine.

[0038] In one embodiment, the feature vectors corresponding to each target object are obtained based on additional features and temporal features, including: The additional features and temporal features are concatenated and fused to obtain the feature vector; The feature vector is input into the variational autoencoder (VAE) module in the trajectory prediction model to participate in the calculation of latent variables. The latent variables directly determine the prediction results output by the trajectory prediction model, thereby realizing the integration of neighbor influence into the trajectory prediction of the target object.

[0039] For example, if the neighboring influencing factor indicates "there is a rapidly approaching vehicle near the target object", the latent variable will guide the model to predict "the trajectory of the target object to avoid to the side", rather than traveling in a straight line.

[0040] In one embodiment, the training method for the trajectory prediction model further includes: Calculate the loss function based on historical spatiotemporal frames and the predicted trajectory segments output by the trajectory prediction model; Adjust the parameters of the trajectory prediction model based on the loss function; The loss function includes: location prediction error and KL divergence loss of latent variables of variational autoencoder in trajectory prediction model.

[0041] In one embodiment, the expression for the loss function is:

[0042] Where L is the loss function, These are the coordinates of the target object within a historical time frame. It is the model's predicted location. It is a latent variable, where T is the starting frame number of the prediction dataset in the historical spatiotemporal frames. Here, H is the KL divergence loss calculation function, and H is the total number of historical spatiotemporal frames.

[0043] It is easy to understand that during the model training process, the previous T frames of the historical spatiotemporal frame are used as the training dataset and input into the model to obtain the prediction dataset. The output result is compared with the spatiotemporal frames from T+1 to T+H to calculate the loss function, thereby adjusting the model so that the predicted trajectory at each time step is as close as possible to the actual trajectory; the predicted positions from T+1 to T+H constitute the predicted trajectory segment.

[0044] In one embodiment, the method further includes: Based on the pre-set underground road network of the coal mine, starting from the beginning of the predicted trajectory segment, the predicted position points in the predicted trajectory segment are sequentially corrected using a real-time hidden Markov map matching algorithm.

[0045] The error of the predicted trajectory segment output by the trajectory prediction model increases with the increase of the prediction time step. Therefore, the output result of the trajectory prediction model in the medium and long term (8-12s) is inaccurate. Therefore, it is necessary to further correct and adjust the predicted trajectory segment. The post-processing logic of this embodiment treats the model's prediction as the measurement value of the inaccurate position measurement device. Referring to the prior underground road network data, the post-processing correction algorithm is used to achieve accurate prediction of the medium and long term trajectory.

[0046] It should be noted that since the predicted trajectory segment includes several predicted position points, the starting point of the predicted trajectory segment is used as the starting point, and the predicted position points are corrected one by one in this way to correct the complete trajectory segment.

[0047] In one embodiment, the process of sequentially correcting each predicted location point in the predicted trajectory segment using a real-time hidden Markov map matching algorithm includes: Using the current predicted location as the center, a set of candidate points is obtained based on the underground road network of the coal mine; Based on the underground road network of the coal mine, determine the transition probability from the current predicted location point to each candidate point in the candidate point set; Based on the underground road network of the coal mine, determine the emission probability of each candidate point in the candidate point set to the next predicted position point of the predicted trajectory segment; Based on the transition probability and the launch probability, the sequence of candidate point combinations with the highest probability within the time period corresponding to the predicted trajectory segment is obtained, and the predicted trajectory segment is corrected accordingly.

[0048] In one embodiment, the candidate point set obtained based on the underground road network of the coal mine, centered on the current predicted location point, includes: Candidate road segments are obtained based on the underground road network of the coal mine, centered on the predicted location point and within a preset step distance range. For each candidate road segment, the specific location of the predicted location point within the candidate road segment is obtained, thus forming a set of candidate points.

[0049] In one embodiment, based on the transition probability and the emission probability, a sequence of candidate point combinations with the highest probability within the time period corresponding to the predicted trajectory segment is obtained, and the predicted trajectory segment is corrected accordingly, including: The total probability is obtained by adding the transition probability and the emission probability. The candidate point combination sequence with the highest total probability is used as the predicted trajectory segment.

[0050] It is easy to understand that the candidate point and the predicted location point may be offset. Correction refers to slightly offsetting the predicted location point to obtain a more accurate predicted location point.

[0051] Since the prediction results of the SocialVAE model inevitably contain errors, more accurate trajectory prediction results can be achieved by using post-processing of the Hidden Markov Model and performing probabilistic correction based on prior underground road network data.

[0052] This invention also provides an electronic device, which includes, but is not limited to: Memory, used to store programs; The processor is used to execute programs stored in memory. When the processor executes the programs stored in memory, it is used to execute the aforementioned method for preventing collisions in underground coal mines.

[0053] The processor and memory can be connected via a bus or other means.

[0054] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs, such as the method described in the embodiments of the present invention. The processor implements the above method by running the non-transitory software program and instructions stored in the memory.

[0055] The memory may include a program storage area and a data storage area, wherein the program storage area may store the operating system and application programs required for at least one function; the data storage area may store data for executing the methods described above. Furthermore, the memory may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0056] The non-transitory software program and instructions required to implement the above terminal selection method are stored in memory and are executed by one or more processors.

[0057] This invention also provides a storage medium storing computer-executable instructions for performing the above-described methods.

[0058] In one embodiment, the storage medium stores computer-executable instructions that are executed by one or more control processors.

[0059] The embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0060] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically include computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0061] This document describes embodiments of the invention, including preferred embodiments known to the inventors for carrying out the invention. Variations of these embodiments will become apparent to those skilled in the art upon reading the foregoing description. The inventors encourage those skilled in the art to adopt such variations as appropriate, and the inventors intend to practice embodiments of the invention in ways other than those specifically described herein. Therefore, the scope of the invention includes all modifications and equivalents of the subject matter set forth in the appended claims, as permitted by applicable law. Furthermore, the scope of the invention covers any combination of the foregoing elements in all possible variations thereof, unless otherwise indicated herein or otherwise clearly contradicted by the context.

Claims

1. A method of collision avoidance in a coal mine, characterised by, The method comprises: acquiring positioning data sets of each target object; constructing a space-time frame according to the constructed time bucket and the positioning data sets; wherein the space-time frame comprises positioning data sets of each target object corresponding to each time bucket respectively; inputting the space-time frame into a pre-constructed trajectory prediction model to obtain a predicted trajectory segment; determining whether the predicted trajectory segments corresponding to each target object respectively exist in the same time bucket to provide early warning against collision.

2. A method of collision avoidance in a coal mine according to claim 1 wherein, The method of constructing a space-time frame according to the constructed time bucket and the positioning data sets comprises: constructing a time bucket; filtering the positioning data sets according to the time bucket, specifically comprising: taking the positioning data of at least two target objects existing in the same time bucket as the filtered positioning data sets; taking the filtered positioning data sets of each target object in the same time bucket and each time bucket as a time frame respectively.

3. A method of collision avoidance in a coal mine according to claim 2, characterised in that, The positioning data sets comprise: a positioning card number, a timestamp, and coordinates of a target object corresponding to the target object; The time frame comprises: a time bucket number, a positioning card number corresponding to a target object, and coordinates of the target object.

4. The method of claim 1, wherein, The training method of the trajectory prediction model comprises: obtaining a historical space-time frame according to the acquired historical positioning data sets; adopting a social attention mechanism for each target object in the historical space-time frame to obtain an additional feature, and performing time sequence feature extraction on the additional feature to obtain a time sequence feature; obtaining a feature vector corresponding to each target object according to the additional feature and the time sequence feature.

5. A method of collision avoidance in a coal mine according to claim 4 wherein, The method of adopting a social attention mechanism for each target object in the historical space-time frame to obtain an additional feature comprises: for each target object in each historical space-time frame, screening the nearby objects of each target object according to a preset distance threshold; determining the relative distance and speed difference between each target object and the corresponding nearby object to obtain an influence factor of each nearby object relative to the corresponding target object; obtaining the additional feature of the target object according to the influence factor of each nearby object relative to the corresponding target object.

6. A method of collision avoidance in a coal mine according to claim 4 wherein, The training method of the trajectory prediction model further comprises: calculating a loss function according to the historical space-time frame and the predicted trajectory segment output by the trajectory prediction model; adjusting the parameters of the trajectory prediction model according to the loss function; wherein the loss function comprises: a position prediction error, and a KL divergence loss of a latent variable of a variational autoencoder in the trajectory prediction model.

7. The method of claim 1, wherein, The method further comprises: according to a preset coal mine underground road network, taking the starting point of the predicted trajectory segment as the starting point, and correcting each predicted position point in the predicted trajectory segment in sequence by a real-time hidden Markov map matching algorithm according to the predicted trajectory segment.

8. A method of collision avoidance in a coal mine according to claim 7 wherein, The method of correcting each predicted position point in the predicted trajectory segment in sequence by a real-time hidden Markov map matching algorithm comprises: taking the current predicted position point as the center, obtaining a candidate point set according to the coal mine underground road network; determining the transition probability of the current predicted position point to each candidate point in the candidate point set according to the coal mine underground road network; determining the emission probability of each candidate point in the candidate point set to the next predicted position point of the predicted trajectory segment according to the coal mine underground road network; According to the transition probability and the emission probability, a candidate point combination sequence with the maximum probability in a time period corresponding to the predicted trajectory segment is obtained, so as to correct the predicted trajectory segment.

9. An electronic device, comprising: Comprise: a memory for storing a program; a processor for executing the program stored in the memory, and when the processor executes the program stored in the memory, the processor is configured to perform the method in any one of claims 1 to 8.

10. A storage medium, characterized by Computer executable instructions are stored, and the computer executable instructions are used to perform the method in any one of claims 1 to 8.

Images