Moving strike system based on dynamic positioning algorithm

Abstract

The present invention relates to the technical field of strike analysis on the move, and specifically to a strike on the move system based on a dynamic positioning algorithm, comprising a time environment data collection module, a time-space feature extraction and fusion module, and a path planning and replanning module, wherein: the time environment data collection module collects time environment data, wherein the time environment data includes environment data at a current time point and environment data at a past time point; the time-space feature extraction and fusion module uses a 3D convolutional neural network to extract time-space features from the time environment data, and fuses the time-space features with static terrain features to form environment features at a current time point and environment features at a past time point; the path planning and replanning module calculates an initial path according to the environment features at a past time point using an A* algorithm; and calculates a difference threshold according to the difference between different time points in the environment features at the past time point; and determines whether to trigger path replanning according to the difference threshold.

Description

Technical Field

[0001] The present invention relates to the technical field of strike-on-the-go analysis, and in particular to a strike-on-the-go system based on a dynamic positioning algorithm. Background Art

[0002] In a dynamic battlefield environment, environmental interference factors (such as weather changes, terrain changes, etc.) will change over time. These changes may cause the initial path of the strike unit to no longer be the optimal path, or even encounter unexpected obstacles. Traditional path planning systems usually use static or semi-static methods for path planning, that is, at the beginning of the mission, an optimal path is calculated based on known environmental data (such as terrain maps, weather forecasts, etc.), and it is expected that this path will remain optimal throughout the mission execution. However, this assumption is often difficult to hold in an actual dynamic battlefield environment.

[0003] For example, in a dynamic strike mission, the strike unit encountered a rain area during the initial path planning, but the rain area moved to the strike unit's path over time. When the traditional system first planned the path, although it was able to use the existing static terrain and meteorological data to generate an initial optimal path, once the environment changed, such as the movement of the rain area, the system could not adjust the path in time. In this case, the strike unit may enter the rain area, resulting in a significant decrease in strike accuracy or even failure to complete the mission. In addition, the traditional system usually needs to recalculate the entire path from the starting point, which is not only time-consuming, but may not guarantee the real-time and optimality of the path in a complex and changing environment.

[0004] To deal with these problems, traditional systems usually require external human intervention to manually adjust the path or re-plan the path. This manual-dependent method not only increases the burden on operators, but may also lead to delayed response time, making it impossible for strike units to quickly make the best decision at critical moments. Summary of the invention

[0005] The purpose of the present invention is to provide a moving strike system based on a dynamic positioning algorithm to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a strike system on the move based on a dynamic positioning algorithm, comprising a time environment data collection module, a spatiotemporal feature extraction and fusion module, and a path planning and re-planning module, wherein:

[0007] The time environment data collection module collects time environment data, wherein the time environment data includes environment data at a current time point and environment data at a past time point;

[0008] The spatiotemporal feature extraction and fusion module uses a 3D convolutional neural network to extract spatiotemporal features from temporal environmental data, and fuses them with static terrain features to form comprehensive environmental features, wherein the comprehensive environmental features include environmental features at the current time point and environmental features at past time points;

[0009] The path planning and replanning module calculates the initial path using the A* algorithm according to the environmental characteristics of the past time point; and calculates the difference threshold according to the difference between the environmental characteristics of the past time point at different time points; calculates the difference between the environmental characteristics of the current time point and the environmental characteristics of the previous time point, and determines whether to trigger path replanning according to the difference threshold; wherein:

[0010] If path replanning is triggered, the D*Lite algorithm is used to recalculate the optimal path based on the environmental characteristics at the current time point.

[0011] As a further improvement of the present technical solution, the spatiotemporal feature extraction and fusion module includes a feature extraction unit, and the process of extracting spatiotemporal features by the feature extraction unit specifically includes:

[0012] The temporal environment data is organized into a 3D tensor in chronological order. The data at each time point is represented as a 2D feature map, and the channels represent different types of environment data.

[0013] The organized 3D tensor is input into the 3D convolutional neural network model, which can process data in both time and space dimensions and extract the temporal and spatial features of environmental changes.

[0014] The 3D convolutional neural network model gradually extracts and compresses the spatiotemporal features in the 3D tensor through multiple layers of convolution and pooling operations, where the spatiotemporal features include time, height and width; each layer of convolution operation captures spatiotemporal information of different scales through convolution kernels of different sizes and numbers; pooling operations are used to reduce the dimension of feature maps, improve computational efficiency and reduce the risk of overfitting; and finally outputs a high-dimensional spatiotemporal feature map.

[0015] As a further improvement of the technical solution, the spatiotemporal feature extraction and fusion module includes a feature fusion unit, and the process of feature fusion performed by the feature fusion unit specifically includes:

[0016] Acquire static terrain features using terrain data collection equipment or an existing terrain data source; and represent the static terrain features as a 2D feature map as a static terrain feature map;

[0017] The spatiotemporal feature map extracted by the 3D neural network model is divided according to time points to generate the spatiotemporal feature map of the current time point and the spatiotemporal feature map of the past time points;

[0018] The spatiotemporal feature map of the current time point and the spatiotemporal feature map of the past time point are weighted fused with the static terrain feature map to form a comprehensive environmental feature map , the formula is as follows:

[0019] ,in is the number of past time points; A spatiotemporal feature map representing the current time point; Indicates The spatiotemporal feature map of past time points; Represents a static terrain feature map; , and Respectively represent the weight of the spatiotemporal feature map at the current time point, the weight of the spatiotemporal feature map at the past time point, and the weight of the static terrain feature map;

[0020] According to the comprehensive environmental characteristics map , determine the environmental characteristics at the current time point and environmental characteristics at past time points , where the environmental characteristics at the current time point are , environmental characteristics map at past time points .

[0021] As a further improvement of the technical solution, the path planning and re-planning module includes a path planning unit, and the process of calculating the initial path by the path planning unit specifically includes:

[0022] Environmental characteristics map based on past time points The terrain and environment information in the program is used to set the heuristic function and the cost function. Starting from the starting point, the path is gradually expanded by evaluating and selecting the node with the lowest cost function value. The open list and the closed list are continuously updated until the end point is found or the node in the open list is empty. The optimal path from the starting point to the end point is generated as the initial path by tracing back the path through the recorded parent node information.

[0023] As a further improvement of the technical solution, the path planning and re-planning module includes a path re-planning unit, and the process of calculating the difference threshold by the path re-planning unit specifically includes:

[0024] Environmental characteristics map for past time points The environmental characteristics at each time point in the past are calculated, and the absolute value of the difference between the environmental characteristics at the previous time point is calculated as the difference between the environmental characteristics at different time points in the past time point. The difference between all time points is statistically analyzed and its mean is calculated. and standard deviation ; Calculate the difference threshold based on its mean and standard deviation , the calculation formula is as follows:

[0025] ,in is an adjustment factor.

[0026] As a further improvement of the technical solution, the process of the path replanning unit triggering the path replanning and performing the path replanning specifically includes:

[0027] If the difference between the current time point and the previous time point is greater than the difference threshold , it is determined that the environment has changed significantly, triggering path replanning. According to the environmental characteristics at the current time point, the D*Lite algorithm is used to recalculate the optimal path; the algorithm is initialized according to the environmental characteristic map at the current time point; starting from the last node of the currently known initial path, the cost function value and heuristic function value of each node are gradually evaluated and updated, and the node with the lowest cost function value is selected for expansion through the priority queue; when the environment changes significantly, the affected nodes are recalculated and the path is updated; by backtracking the parent node information, a new optimal path from the start point to the end point is generated.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The on-the-go strike system based on the dynamic positioning algorithm collects and processes environmental data at current and past time points in real time, uses a 3D convolutional neural network to extract spatiotemporal features, extracts features from the time and space dimensions, and weightedly fuses the features with the static terrain feature map to form a comprehensive environmental feature map. This enables the system to fully understand environmental changes, such as the movement of rain areas;

[0030] In addition, the system determines the significance of environmental changes by calculating a difference threshold. If the difference between the environmental characteristics at the current time point and the environmental characteristics at the previous time point exceeds the threshold, the system will use the D*Lite algorithm to gradually evaluate and update the cost function value and heuristic function value of each node starting from the last node of the current path, and select the node with the lowest comprehensive cost for expansion. Through this dynamic adjustment, the strike unit can avoid moving rain areas to ensure strike accuracy and smooth completion of the mission. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic diagram of the overall module of the present invention;

[0032] Figure 2 It is a schematic diagram of the spatiotemporal feature extraction and fusion module unit of the present invention;

[0033] Figure 3 Schematic diagram of the path planning and re-planning module unit of the present invention.

[0034] In the figure: 100, time environment data collection module; 200, spatiotemporal feature extraction and fusion module; 201, feature extraction unit; 202, feature fusion unit; 300, path planning and re-planning module; 301, path planning unit; 302, path re-planning unit. DETAILED DESCRIPTION

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0036] Next, see Figure 1-Figure 3 The present invention provides a technical solution: a moving strike system based on a dynamic positioning algorithm, including a time environment data collection module 100, a spatiotemporal feature extraction and fusion module 200 and a path planning and re-planning module 300.

[0037] The time environment data collection module 100 collects time environment data, wherein the time environment data includes environment data at the current time point and environment data at the past time points, specifically including:

[0038] Environmental data at various time points during the operation of the strike system are collected as temporal environmental data, which includes not only environmental data at the current time point, such as images, radar, temperature, humidity, wind speed and lighting conditions, but also environmental data at past time points; the temporal environmental data collection module 100 obtains the information in real time through sensors and data acquisition equipment deployed in the war zone, and stores and marks it as the corresponding time point.

[0039] The temporal environment data collection module 100 ensures that the system can fully and timely grasp the changes in the battlefield environment by collecting environmental data of the current and past time points in real time, such as images, radar, temperature, humidity, wind speed and lighting conditions; this continuous data collection not only improves the system's perception ability, but also provides rich information support for subsequent modules, allowing the strike unit to make more reasonable decisions when facing a dynamically changing environment, avoiding path failure or mission failure due to environmental changes.

[0040] The feature extraction unit 201 in the spatiotemporal feature extraction and fusion module 200 uses a 3D convolutional neural network to extract spatiotemporal features from the temporal environment data, specifically including:

[0041] The temporal environmental data is organized into a 3D tensor (time × height × width × channel) in chronological order. The data at each time point is represented as a 2D feature map, and the channels represent different types of environmental data (such as image, radar, temperature, humidity, wind speed, and lighting conditions). For example, suppose we have data at 5 time points, and the data at each time point includes 200×200 pixel image data and 1×1 temperature, humidity, wind speed, and lighting condition data, then the shape of the 3D tensor is (5×200×200×6);

[0042] The organized 3D tensor is input into the 3D convolutional neural network model, which can process data in both time and space dimensions and extract the temporal and spatial features of environmental changes.

[0043] The 3D convolutional neural network model gradually extracts and compresses the spatiotemporal features in the 3D tensor through multiple layers of convolution and pooling operations, where the spatiotemporal features include time, height and width; each layer of convolution operation captures spatiotemporal information of different scales through convolution kernels of different sizes and numbers; pooling operations are used to reduce the dimension of feature maps, improve computational efficiency and reduce the risk of overfitting; and finally outputs a high-dimensional spatiotemporal feature map.

[0044] The feature extraction unit 201 uses a 3D convolutional neural network to extract spatiotemporal features from temporal environmental data, and gradually captures and compresses spatiotemporal information of different scales through multi-layer convolution and pooling operations; this enables the system to effectively identify and analyze the temporal and spatial features of environmental changes, and form a high-dimensional spatiotemporal feature map; in a dynamic battlefield environment, this multi-dimensional feature extraction method can quickly capture subtle differences in environmental changes, providing a more accurate and detailed data basis for path planning and re-planning, ensuring that the strike unit can respond to environmental changes in a timely manner and maintain strike accuracy and mission success rate.

[0045] The feature fusion unit 202 in the spatiotemporal feature extraction and fusion module 200 fuses the spatiotemporal features with the static terrain features to form a comprehensive environmental feature, wherein the comprehensive environmental feature includes the environmental feature at the current time point and the environmental feature at the past time point. The specific process is as follows:

[0046] Using terrain data collection equipment or existing terrain data sources to obtain static terrain features, the static terrain features include terrain height, slope, vegetation coverage and other information; and representing the static terrain features as a 2D feature map as a static terrain feature map;

[0047] The spatiotemporal feature map extracted by the 3D neural network model is divided according to time points to generate the spatiotemporal feature map of the current time point and the spatiotemporal feature map of the past time points;

[0048] The spatiotemporal feature map of the current time point and the spatiotemporal feature map of the past time point are weighted fused with the static terrain feature map to form a comprehensive environmental feature map , the formula is as follows:

[0049] ,in is the number of past time points; A spatiotemporal feature map representing the current time point; Indicates The spatiotemporal feature map of past time points; Represents a static terrain feature map; , and Respectively represent the weight of the spatiotemporal feature map at the current time point, the weight of the spatiotemporal feature map at the past time point, and the weight of the static terrain feature map;

[0050] According to the comprehensive environmental characteristics map , determine the environmental characteristics at the current time point and environmental characteristics at past time points , where the environmental characteristics at the current time point are , environmental characteristics map at past time points .

[0051] The feature fusion unit 202 generates a comprehensive environmental feature map by weighted fusion of the spatiotemporal feature maps of the current time point and the past time points and the static terrain feature map; this fusion method not only retains the information of dynamic environmental changes, but also combines the stability characteristics of the static terrain, so that the system can more comprehensively understand and evaluate the battlefield environment; when the subsequent environmental changes are significant, the comprehensive environmental feature map provides rich input data for path replanning, improves the robustness and reliability of the path, and ensures that the strike unit can select the optimal path in a constantly changing environment to avoid mission failure due to environmental interference.

[0052] The path planning unit 301 in the path planning and re-planning module 300 calculates the initial path using the A* algorithm according to the environmental characteristics at the past time point, specifically including:

[0053] The A* algorithm is a heuristic search algorithm that can efficiently find the optimal path from the starting point to the end point on a weighted map; according to the environmental feature map at the past time point The terrain and environment information in the algorithm is used to set the heuristic function (such as Manhattan distance or Euclidean distance) and the cost function (such as path length, obstacle density, etc.); starting from the starting point, the path is gradually expanded by evaluating and selecting the node with the lowest cost function value; the open list and the closed list are continuously updated until the end point is found or the node in the open list is empty; the path is traced back through the recorded parent node information to generate the optimal path from the starting point to the end point as the initial path.

[0054] The path planning unit 301 uses the A* algorithm to calculate the initial path according to the environmental characteristics of the past time point; by setting the heuristic function and the cost function, the A* algorithm can efficiently find the optimal path from the starting point to the end point; in a dynamic battlefield environment, this initial path optimization provides a reliable basis for subsequent path replanning; when the environment changes, the efficiency and accuracy of the initial path enable the strike unit to respond to environmental changes more quickly and ensure the smooth progress of the mission.

[0055] The path re-planning unit 302 in the path planning and re-planning module 300 calculates the difference threshold according to the difference between the environmental characteristics at different time points in the past time points, specifically including:

[0056] Environmental characteristics map for past time points The environmental characteristics at each time point in the past are calculated, and the absolute value of the difference between the environmental characteristics at the previous time point is calculated as the difference between the environmental characteristics at different time points in the past time point. The difference between all time points is statistically analyzed and its mean is calculated. and standard deviation ; Calculate the difference threshold based on its mean and standard deviation , the calculation formula is as follows:

[0057] ,in is an adjustment factor.

[0058] The path replanning unit 302 in the path planning and replanning module 300 calculates the difference between the environmental characteristics at the current time point and the environmental characteristics at the previous time point, and determines whether to trigger path replanning based on the difference threshold, specifically including:

[0059] If the difference between the current time point and the previous time point is greater than the difference threshold , it is determined that the environment has changed significantly, triggering path replanning, and using the D*Lite algorithm to recalculate the optimal path based on the environmental characteristics at the current time point; D*Lite is a dynamic path planning algorithm that can efficiently update the path when the environment changes dynamically without recalculating the entire path; first, the algorithm is initialized based on the environmental characteristic map at the current time point; starting from the last node of the currently known initial path, the cost function value and heuristic function value of each node are gradually evaluated and updated, and the node with the lowest cost function value is selected for expansion through the priority queue; when the environment changes significantly, the affected nodes are recalculated and the path is updated; by backtracking the parent node information, a new optimal path from the start point to the end point is generated to ensure that the path always adapts to the latest environmental characteristics.

[0060] The path replanning unit 302 realizes dynamic path replanning when the environment changes significantly by calculating the difference threshold and using the D*Lite algorithm; the D*Lite algorithm can efficiently update the path when the environment changes dynamically without recalculating the entire path, which not only improves the response speed of the system, but also ensures that the strike unit can quickly adjust the path when facing sudden environmental changes (such as the movement of rain areas) to maintain strike accuracy and smooth completion of the task; in addition, this automated method reduces the need for external manual intervention, reduces the burden on operators, and improves the autonomy and real-time performance of the system.

[0061] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A moving strike system based on a dynamic positioning algorithm, characterized in that: It comprises a temporal environment data collection module (100), a temporal and spatial feature extraction and fusion module (200) and a path planning and re-planning module (300), wherein: The time environment data collection module (100) collects time environment data, wherein the time environment data includes environment data at a current time point and environment data at a past time point; The spatiotemporal feature extraction and fusion module (200) uses a 3D convolutional neural network to extract spatiotemporal features from temporal environmental data, and fuses the spatiotemporal features with static terrain features to form comprehensive environmental features, wherein the comprehensive environmental features include environmental features at a current time point and environmental features at a past time point; the spatiotemporal feature extraction and fusion module (200) includes a feature fusion unit (202), and the process of feature fusion performed by the feature fusion unit (202) specifically includes: Acquire static terrain features using terrain data collection equipment or an existing terrain data source; and represent the static terrain features as a 2D feature map as a static terrain feature map; The spatiotemporal feature map extracted by the 3D neural network model is divided according to time points to generate the spatiotemporal feature map of the current time point and the spatiotemporal feature map of the past time points; The spatiotemporal feature map of the current time point and the spatiotemporal feature map of the past time point are weightedly fused with the static terrain feature map to form a comprehensive environmental feature map; According to the comprehensive environmental characteristic map, the environmental characteristics at the current time point and the environmental characteristic map at the past time point are determined; the calculation formula of the environmental characteristic map at the past time point is as follows: ;in A map of environmental features representing a past point in time; The weights of the spatiotemporal feature maps representing past time points; Indicates The spatiotemporal feature map of past time points; Represents the weight of the static terrain feature map; Represents a static terrain feature map; The path planning and re-planning module (300) calculates an initial path using an A* algorithm based on environmental characteristics at past time points; and calculates a difference threshold based on the difference between environmental characteristics at different time points in the past time points; calculates the difference between environmental characteristics at the current time point and environmental characteristics at the previous time point, and determines whether to trigger path re-planning based on the difference threshold; wherein: If path replanning is triggered, the D*Lite algorithm is used to recalculate the optimal path based on the environmental characteristics at the current time point.

2. The on-the-go strike system based on dynamic positioning algorithm according to claim 1, characterized in that: The spatiotemporal feature extraction and fusion module (200) comprises a feature extraction unit (201), and the process of extracting spatiotemporal features by the feature extraction unit (201) specifically comprises: The temporal environment data is organized into a 3D tensor in chronological order. The data at each time point is represented as a 2D feature map, and the channels represent different types of environment data. The organized 3D tensor is input into the 3D convolutional neural network model, which can process data in both time and space dimensions and extract the temporal and spatial features of environmental changes. The 3D convolutional neural network model gradually extracts and compresses the spatiotemporal features in the 3D tensor through multiple layers of convolution and pooling operations, where the spatiotemporal features include time, height and width; each layer of convolution operation captures spatiotemporal information of different scales through convolution kernels of different sizes and numbers; pooling operations are used to reduce the dimension of feature maps, improve computational efficiency and reduce the risk of overfitting; and finally outputs a high-dimensional spatiotemporal feature map.

3. The on-the-go strike system based on dynamic positioning algorithm according to claim 1, characterized in that: The comprehensive environmental characteristics map The calculation formula is as follows: ,in is the number of past time points; A spatiotemporal feature map representing the current time point; Indicates The spatiotemporal feature map of past time points; Represents a static terrain feature map; , and Respectively represent the weight of the spatiotemporal feature map at the current time point, the weight of the spatiotemporal feature map at the past time point, and the weight of the static terrain feature map; According to the comprehensive environmental characteristics map , determine the environmental characteristics at the current time point and environmental characteristics at past time points , where the environmental characteristics at the current time point are .

4. The on-the-go strike system based on dynamic positioning algorithm according to claim 3 is characterized in that: The path planning and re-planning module (300) comprises a path planning unit (301), and the process of calculating the initial path by the path planning unit (301) specifically comprises: Environmental characteristics map based on past time points The terrain and environment information in the program is used to set the heuristic function and the cost function. Starting from the starting point, the path is gradually expanded by evaluating and selecting the node with the lowest cost function value. The open list and the closed list are continuously updated until the end point is found or the node in the open list is empty. The optimal path from the starting point to the end point is generated as the initial path by tracing back the path through the recorded parent node information.

5. The on-the-go strike system based on dynamic positioning algorithm according to claim 4, characterized in that: The path planning and re-planning module (300) comprises a path re-planning unit (302), and the process of calculating the difference threshold by the path re-planning unit (302) specifically comprises: Environmental characteristics map for past time points The environmental characteristics at each time point in the past are calculated, and the absolute value of the difference between the environmental characteristics at the previous time point is calculated as the difference between the environmental characteristics at different time points in the past time point. The difference between all time points is statistically analyzed and its mean is calculated. and standard deviation ; Calculate the difference threshold based on its mean and standard deviation , the calculation formula is as follows: ,in is an adjustment factor.

6. The on-the-go strike system based on dynamic positioning algorithm according to claim 5, characterized in that: The process of the path re-planning unit (302) triggering the path re-planning and performing the path re-planning specifically includes: If the difference between the current time point and the previous time point is greater than the difference threshold , it is determined that the environment has changed significantly, triggering path replanning. According to the environmental characteristics at the current time point, the D*Lite algorithm is used to recalculate the optimal path; the algorithm is initialized according to the environmental characteristic map at the current time point; starting from the last node of the currently known initial path, the cost function value and heuristic function value of each node are gradually evaluated and updated, and the node with the lowest cost function value is selected for expansion through the priority queue; when the environment changes significantly, the affected nodes are recalculated and the path is updated; by backtracking the parent node information, a new optimal path from the start point to the end point is generated.

Images