Edge computing device layout method for roadside spatio-temporal data collection

Abstract

The invention discloses an edge computing device layout method for roadside spatio-temporal data collection, and belongs to the field of roads. The invention aims to solve the problems of complex calculation and low matching rate with the positions of sensor nodes in the existing mode of deploying edge computing equipment nodes, all sensor nodes to be covered on a road are acquired, and all sensor nodes are layered by using a KD-Tree space; drawing a circle by taking each sensor node as a circle center according to a preset radius; the coverage potentials of all the circles in each layer are sequentially ranked from high to low, searching is started from the first circle in each layer, the positions of the edge computing devices are set according to whether the first circle intersects with other circles or not, the coverage potentials and the coverage scores, and arrangement of all the edge computing devices is completed according to the query sequence of the circles and the arrangement mode. The method is used for arranging the edge computing device.

Description

Technical Field

[0001] The present invention belongs to the field of roads and relates to the layout of edge computing devices. Background Art

[0002] With the continuous advancement of the construction of intelligent transportation and smart cities, the technology of collecting spatio-temporal data on the roadside has become a key link in supporting traffic management and urban planning.

[0003] In actual projects, the deployment of edge computing device nodes has gradually become an important means to improve data processing capabilities. Taking the Nansha Pearl Bay project in Shenzhen as an example, this project uses high-performance Ascend 310 chip edge computing devices to cover sensor nodes distributed on the roadside, thereby realizing real-time processing and efficient collection of data, and ensuring the accuracy and precision of data collection.

[0004] However, the existing deployment methods of edge computing device nodes still have obvious deficiencies. In actual projects, extensive grid or random distribution strategies are often adopted, and the coverage range is determined by simple distance calculation. This method cannot be optimized according to the specific positions and distributions of sensor nodes, and thus exposes many problems in actual applications, including but not limited to increasing economic costs, affecting the timeliness of data processing, and increasing traffic risks. For example, roadside units on German highways are arranged according to unified standards, and the utilization rate of 80% of some nodes is lower than 10% during non-peak hours.

[0005] The root cause of the above problems is that the extensive layout method fails to accurately plan according to the positions of sensor nodes when deploying edge computing device nodes, and it is difficult to meet the actual needs of complex roadside environments. Therefore, there is an urgent need for a more adaptable edge computing device layout method for collecting spatio-temporal data on the roadside.

[0006] Existing methods for deploying edge computing device nodes using unit circle coverage, such as the greedy algorithm and linear programming method, can theoretically provide coverage solutions, but there are limitations in actual applications. When the traditional greedy algorithm processes 200 sensor nodes, the matching rate between edge computing devices and sensor nodes is low, among which the missed matching rate is 12.3% and the false matching rate is 8.7%. Therefore, there is an urgent need for an efficient and optimized edge computing device deployment method that can reduce the computational complexity while ensuring the coverage range and meeting the requirements of large-scale computing node scenarios. Summary of the Invention

[0007] The purpose of the present invention is to solve the problems of complex calculation and low matching rate with the positions of sensor nodes in the existing method of deploying edge computing device nodes, and to propose an edge computing device layout method for collecting spatio-temporal data on the roadside.

[0008] Method for deploying edge computing devices for roadside spatio-temporal data collection, the method comprising the following:

[0009] Step 1: Obtain all sensor nodes to be covered on the road, and layer all sensor nodes using a KD-Tree space;

[0010] Step 2: Draw circles with each sensor node as the center according to a preset radius;

[0011] Step 3: Calculate the total coverage score and the dispersion of each circle in the i-th layer. Among them, the total coverage score of each circle is obtained by summing the coverage scores of multiple arcs divided by other circles, or using the coverage score of the entire circle not divided by other circles as the total coverage score. According to the total coverage of each circle and the dispersion of the corresponding circle, obtain the coverage potential of each circle, and sort the coverage potentials of all circles in the i-th layer from high to low. The initial value of i is 1;

[0012] Step 4: Search for the j-th circle in the i-th layer, and the initial value of j is 1.

[0013] If the j-th circle intersects with other circles, select 1 circle with the largest coverage potential from the intersecting circles, and deploy 1 edge computing device at any point on the arc with the maximum coverage score on this circle. After deployment, draw 1 circle with the deployment position of this edge computing device as the center and the preset radius, and delete the sensor nodes on the current layer and other layers covered by this circle.

[0014] If the j-th circle does not intersect with other circles, deploy 1 edge computing device at any point on the arc of this circle. After deployment, draw 1 circle with the deployment position of this edge computing device as the center and the preset radius, and delete the sensor nodes on the current layer and other layers covered by this circle;

[0015] Step 5: Determine whether j is equal to the total number of all sensor nodes in the i-th layer. If not, make j = j + 1 and execute Step 6. If so, execute Step 7;

[0016] Step 6: Determine whether the j-th circle is deleted. If so, make j = j + 1 and return to Step 6. If not, execute Step 4;

[0017] Step 7: Determine whether i is equal to the total number of layers. If not, make i = i + 1 and execute Step 3. If so, all sensor nodes are deleted and the deployment of all edge computing devices is completed.

[0018] Preferably, in Step 3, the process of obtaining the total coverage score of each circle is as follows:

[0019] If one circle intersects with other circles, the multiple arcs formed by this circle being segmented by other circles are used to calculate the coverage score for each arc according to the angle of each arc and the number of times the arc is covered by circles. The coverage scores of all arcs on a circle are added together to obtain the total coverage score of a circle;

[0020] If one circle does not intersect with other circles, the total coverage score of this circle is 0.

[0021] Preferably, in step 3, the dispersion is expressed as:

[0022]

[0023] In the formula, R i is the dispersion of the i-th circle, w k is the coverage score of the k-th arc, θ end,k is the end angle of the k-th arc, θ start,k is the start angle of the k-th arc.

[0024] Preferably, the method further includes step 8: using the node positions of all deployed edge computing devices to draw a distribution map.

[0025] Preferably, a python visualization tool is used to draw the distribution map.

[0026] Preferably, the preset radii in step 2 and step 4 are equal.

[0027] Preferably, the specific process of obtaining the coverage potential of each circle according to the total coverage of each circle and the dispersion of the corresponding circle is as follows:

[0028] Multiply the total coverage of each circle by the dispersion of the corresponding circle to obtain the coverage potential of each circle.

[0029] The beneficial effects of the present invention are:

[0030] First, the sensor node set is stratified, and then the coverage potential of the circle formed by each sensor node is evaluated through the intersection relationship between sensor nodes and based on the arc coverage index. Finally, the hierarchical greedy algorithm is used to select the optimal position for placing edge computing devices and generate a coverage visualization result to verify the optimization effect; the present invention uses each edge computing device to process the acquisition data of the sensor nodes covered by this edge computing device, and the matching rate between the positions of the edge computing devices deployed in the present invention and the covered sensor nodes is high.

[0031] The core idea of the present invention lies in integrating spatial indexing technology, geometric coverage analysis, and statistical methods to construct a hierarchical and adaptive optimization framework for edge computing device deployment. Specifically, through the KD-Tree hierarchical spatial index, node query and region division are accelerated. The covering potential analysis based on arcs is used to locate the optimal candidate covering direction, and the circular integral index in circular statistics is introduced to quantify the covering dispersion, guiding the greedy algorithm to select the global optimal solution. This comprehensive strategy not only significantly reduces the computational complexity but also ensures the uniformity of the coverage distribution through quantitative indicators, breaking through the local optimization limitations of traditional methods.

[0032] According to actual requirements, the present invention formulates this problem as finding the minimum number of computing nodes under the conditions of existing edge computing devices with a coverage radius, several sensor nodes generating data, and their location information, placing them at selected positions, and ultimately achieving the goal of covering all sensor nodes. This problem can be reduced to the unit circle covering problem.

[0033] The present invention uses the recursive partitioning and hierarchical management method of the KD-Tree hierarchical index to stratify sensor nodes, improving the spatial query efficiency and being more suitable for scenarios with a large scale of sensor nodes and a large number of computing nodes to be arranged.

[0034] Traditional methods mostly use brute-force direct calculation of the number of coverable nodes. However, through the arc covering potential analysis, the present invention can not only accurately quantify the covering direction of nodes based on polar angle segmentation and vector synthesis but also effectively avoid the problem of traditional methods falling into local optimal solutions, reduce the operation cost, and improve the layout efficiency.

[0035] The present invention introduces statistical methods and creatively proposes using the covering vector norm as the dispersion index to guide global optimization and ensure the uniformity of the coverage distribution and the convergence of the algorithm.

[0036] Through the deep integration of the KD-Tree spatial index, arc covering analysis, and circular integral index, the present invention provides an efficient and low-cost method for optimizing sensor deployment for roadside spatio-temporal data collection. Brief Description of the Drawings

[0037] Figure 1 It is a flowchart of the method for deploying edge computing devices for roadside spatio-temporal data collection;

[0038] Figure 2 It is a comparison diagram of the method for deploying edge computing devices in this application and the existing method for deploying edge computing devices. In the figure, Figure 2 (a) is the layout diagram of edge computing devices by the conventional greedy algorithm, Figure 2 (b) is the layout diagram of edge computing devices of the present invention, Figure 2(c) Layout diagram of edge computing devices for conventional genetic algorithms;

[0039] Figure 3 is Figure 2 The enlarged view of the part in (b). In the figure, Figure 3 (a) is an intersection diagram of multiple circles, Figure 3 (b) is a diagram of a circle divided into multiple arcs by other circles;

[0040] Figure 4 is an intersection diagram of two circles formed by two sensor nodes;

[0041] Figure 5 is a comparison diagram of coverage methods. In the figure, Figure 5 (a) is a layout diagram of edge computing devices for the conventional greedy algorithm, Figure 5 (b) is a layout diagram of edge computing devices of the present invention, Figure 5 (c) is a layout diagram of edge computing devices for conventional genetic algorithms;

[0042] Figure 6 is a comparison diagram of the number of edge computing devices deployed by the existing method and the present application. Detailed implementation manners

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0044] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0045] Next, the present invention will be further described in conjunction with the accompanying drawings and specific embodiments, but it is not a limitation of the present invention.

[0046] Embodiment:

[0047] An edge computing device layout method for roadside spatio-temporal data collection, the method includes the following contents:

[0048] Step 1: Obtain all sensor nodes to be covered on the road, and layer all sensor nodes using the KD-Tree space;

[0049] Step 2: Draw circles with each sensor node as the center according to a preset radius;

[0050] Step 3: Calculate the total coverage score and the dispersion of each circle in the i-th layer. Among them, the total coverage score of each circle is obtained by summing the coverage scores of multiple arcs divided by other circles, or using the coverage score of the entire circle not divided by other circles as the total coverage score. According to the total coverage of each circle and the dispersion of the corresponding circle, obtain the coverage potential of each circle, and sort the coverage potentials of all circles in the i-th layer from high to low. The initial value of i is 1;

[0051] Step 4: Locate the j-th circle in the i-th layer. The initial value of j is 1,

[0052] If the j-th circle intersects with other circles, select 1 circle with the largest coverage potential from the multiple intersecting circles, select any point on the arc with the maximum coverage score on this circle to deploy 1 edge computing device. After deployment, draw 1 circle with the deployment position of this edge computing device as the center and a preset radius, and delete the sensor nodes on the current layer and other layers covered by this circle.

[0053] If the j-th circle does not intersect with other circles, select any point on the arc of this circle to deploy 1 edge computing device. After deployment, draw 1 circle with the deployment position of this edge computing device as the center and a preset radius, and delete the sensor nodes on the current layer and other layers covered by this circle;

[0054] Step 5: Determine whether j is equal to the total number of all sensor nodes in the i-th layer. If not, set j = j + 1 and execute Step 6. If so, execute Step 7;

[0055] Step 6: Determine whether the j-th circle is deleted. If so, set j = j + 1 and return to Step 6. If not, execute Step 4;

[0056] Step 7: Determine whether i is equal to the total number of layers. If not, set i = i + 1 and execute Step 3. If so, all sensor nodes are deleted and the deployment of all edge computing devices is completed.

[0057] Specifically, in Step 1, the KD-Tree hierarchical spatial index is used to accelerate node query and area division, the coverage potential analysis based on arcs is used to locate the optimal candidate coverage direction, and the circular integral index in circular statistics is introduced to quantify the coverage dispersion, guiding the greedy algorithm to select the global optimal solution. This comprehensive strategy not only significantly reduces the computational complexity but also ensures the uniformity of the coverage distribution through the quantification index, breaking through the local optimization limitations of traditional methods.

[0058] After all sensor nodes are stratified by the KD-Tree, all stratified sensor nodes will appear on a coordinate graph. Figure 2The red stars represent each sensor node. With each sensor node as the center and a preset radius, a blue circle is drawn. First, an edge computing device is deployed for the sensor node with the greatest coverage potential in the first layer. If Figure 3 The circle in the lower left corner in (b) has the greatest coverage potential value among the three intersecting circles. Then, find the arc with the highest coverage score from this circle. From Figure 3 (b), it can be seen that this circle is divided into 4 arcs by other intersecting circles, which are represented by yellow, brown, blue, and red respectively. Calculate the coverage score of each arc. Through calculation, since the brown one is within three circles, the brown arc has the highest coverage score. Therefore, randomly select a point on this colored arc to place a green edge computing device. After placement, draw a green circle. If the green circle covers all three red stars, then remove these three stars and the corresponding blue circles. Follow this method to then query the second sensor node (the sensor node with the second greatest coverage potential) in this layer, and continue in this way until all sensor nodes have edge computing devices deployed. After deployment, each edge computing device is used to process the data of the covered sensors.

[0059] Through Figure 2 and Figure 5 it can be seen that the coverage quality of this embodiment has a significant improvement compared with the existing algorithm. In addition Figure 6 the number of computing nodes required by the method proposed in this embodiment and the existing method are compared in a data quantization manner. It can be seen that this method is significantly better than the existing method in most cases, and the number of computing nodes in this embodiment is significantly less than the number of existing deployed computing nodes. Therefore, this embodiment has a fast deployment speed, a small number of deployed computing nodes, and better effects.

[0060] Further defined, in step 3, the process of obtaining the total coverage score of each circle is as follows:

[0061] If one circle intersects with other circles, the multiple arcs into which this circle is divided by other circles, according to the angle of each arc and the number of circles covering this arc, calculate the coverage score of each arc, and add up the coverage scores of all arcs on a circle to obtain the total coverage score of a circle;

[0062] If one circle does not intersect with other circles, the total coverage score of this circle is 0.

[0063] Further defined, in step 3, the dispersion is expressed as:

[0064]

[0065] In the formula, R i is the dispersion of the i-th circle, w k is the coverage score of the k-th arc, θ end,k is the ending angle of the k-th arc, θstart,k is the starting angle of the kth arc.

[0066] It is further defined that the method also includes step 8: drawing a distribution map using the node positions of all deployed edge computing devices.

[0067] Further qualification, the distribution map was drawn using python visualization tools.

[0068] It is further defined that the preset radii in step 2 and step 4 are equal.

[0069] Further defined, according to the total coverage of each circle and the discreteness of the corresponding circle, the specific process of obtaining the coverage potential of each circle is:

[0070] The total coverage of each circle is multiplied by the discreteness of the corresponding circle to obtain the coverage potential of each circle.

[0071] Specifically, it is defined as follows: Given a fixed set of sensor nodes’ geographic coordinates P = {p1, p2, …, p n}, the coverage radius of the edge computing device is r, and each edge computing device is called a computing node. The computing node set C = {c1, c2, …, c k}, so that: each sensor node p i At least one computing node c i Coverage, that is, ||p i -c j ||≤r; minimize the number of sensors used k; cover the distribution evenly and avoid local dense or sparse areas.

[0072] First, the KD-Tree structure is used to recursively partition the sensor node coordinates. The sensor node set is divided alternately along the x-axis and y-axis by the median to generate a multi-layer tree structure. By recording the index information of each layer of sensor nodes, a hierarchical dictionary is formed, which actually reflects the mapping from depth to sensor node list. The algorithm can prioritize the leaf node area of the KD-Tree to reduce redundant calculations. Compared with traditional brute force search, KD-Tree reduces the complexity of neighbor query from 0(n 2 ) drops to 0(nlogn).

[0073] Secondly, when determining the coverage potential of each node, this embodiment abandons the simple distance measurement and instead accurately locates the coverage direction through polar angle segmentation and arc analysis. For the jth sensor node, KD-Tree is used to quickly query the neighboring nodes with a distance less than 2r from it and calculate the relative polar angle θ ij And generate the covering arc segment. For example, node p i With neighbors jWhen the spacing is 1.5r, the covered arc span is 2arccos(1.5r / 2r) = 120°, and the midpoint angle θ opt is the optimal coverage direction.

[0074] This embodiment also creatively introduces the result vector in circular statistics. By synthesizing the arc segments of all neighbors, a coverage vector is generated, and its modulus length R i The calculation formula is:

[0075]

[0076] This index R i is very suitable for quantifying the concentration degree of the coverage direction: R i →1 indicates that the coverage directions are highly concentrated and are suitable as candidate center points; R i →0 indicates that the coverage is dispersed, and redundant deployment needs to be avoided.

[0077] Finally, this embodiment also proposes hierarchical greedy coverage for the problem that traditional coverage methods are prone to difficulties in searching when the nearest neighbor points are dense, and creatively introduces the concept of circular integral in circular statistics, with the modulus length R of the coverage vector i as the core decision index. The specific process is as follows: Starting from the deepest layer of the KD-Tree, un-covered nodes are screened layer by layer and their R i values are calculated. The sensor node p i with the largest R j is preferentially selected as the candidate. According to the covered arc segment of p j , the optimal center position c j = p j + r(cosθ opt , sinθ opt ) is determined to ensure the maximum coverage range. Each time a sensor is deployed, its covered nodes are immediately queried through the KD-Tree and the uncovered set is updated. Compared with the traditional greedy algorithm, the circular integral index effectively balances the coverage density and uniformity and avoids the local optimal trap.

[0078] Although the present invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed, as long as they do not deviate from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the different dependent claims and the features described herein can be combined in a manner different from that described in the original claims. It should also be understood that the features described in connection with a single embodiment can be used in other described embodiments.

Claims

1. Edge computing device deployment method for roadside spatio-temporal data collection, characterized in that, The method includes the following steps: Step 1: Obtain all sensor nodes to be covered on the road and layer all sensor nodes using the KD-Tree space; Step 2: Draw circles with each sensor node as the center according to a preset radius; Step 3: Calculate the total coverage score and the dispersion of each circle in the i-th layer. Among them, the total coverage score of each circle is obtained by summing the coverage scores of multiple arcs divided by other circles, or taking the coverage score of the entire circle not divided by other circles as the total coverage score. According to the total coverage of each circle and the dispersion of the corresponding circle, obtain the coverage potential of each circle, and sort the coverage potentials of all circles in the i-th layer from high to low. The initial value of i is 1; Step 4: Locate the j-th circle in the i-th layer, and the initial value of j is 1. If the j-th circle intersects with other circles, select the circle with the largest coverage potential from the multiple intersecting circles, and deploy 1 edge computing device at any point on the arc with the maximum coverage score on this circle. After deployment, draw a circle with the deployment position of this edge computing device as the center and the preset radius, and delete the sensor nodes on the current layer and other layers covered by this circle. If the j-th circle does not intersect with other circles, deploy 1 edge computing device at any point on the arc of this circle. After deployment, draw a circle with the deployment position of this edge computing device as the center and the preset radius, and delete the sensor nodes on the current layer and other layers covered by this circle; Step 5: Determine whether j is equal to the total number of all sensor nodes in the i-th layer. If not, set j = j + 1 and execute Step 6. If so, execute Step 7; Step 6: Determine whether the j-th circle is deleted. If so, set j = j + 1 and return to Step 6. If not, execute Step 4; Step 7: Determine whether i is equal to the total number of layers. If not, set i = i + 1 and execute Step 3. If so, all sensor nodes are deleted and the deployment of all edge computing devices is completed.

2. The method for deploying edge computing devices for roadside spatio-temporal data collection according to claim 1, wherein In Step 3, the process of obtaining the total coverage score of each circle is as follows: If a circle intersects with other circles, then the circle is divided into multiple arcs by other circles. According to the angle of each arc and the number of times the arc is covered by the circle, calculate the coverage score of each arc, and sum the coverage scores of all arcs on a circle to obtain the total coverage score of a circle; If a circle does not intersect with other circles, then the total coverage score of this circle is 0.

3. The method for deploying edge computing devices for roadside spatio-temporal data collection according to claim 1, wherein In Step 3, the dispersion is expressed as: In the formula, R i is the discreteness of the ith circle, w k is the coverage score of the kth arc, θ end,k is the ending angle of the kth arc, θ start,k is the starting angle of the kth arc.

4. The method for deploying edge computing devices for roadside spatio-temporal data collection according to claim 1, wherein, The method further includes Step 8: Use the node positions of all deployed edge computing devices to draw a distribution map.

5. The method for deploying edge computing devices for roadside spatio-temporal data collection according to claim 4, wherein Python visualization tools are used to draw the distribution map.

6. The method for deploying edge computing devices for roadside spatio-temporal data collection according to claim 1, wherein The preset radii in Step 2 and Step 4 are equal.

7. The method for deploying edge computing devices for roadside spatio-temporal data collection according to claim 1, wherein The specific process of obtaining the coverage potential of each circle according to the total coverage of each circle and the dispersion of the corresponding circle is as follows: Multiply the total coverage of each circle by the dispersion of the corresponding circle to obtain the coverage potential of each circle.

Images