Air unmanned aerial vehicle and ground detector traffic perception network layout method and system

Abstract

The embodiment of the invention discloses an aerial unmanned aerial vehicle and ground detector traffic perception network layout method and system, belongs to the technical field of intelligent traffic, and solves the problem that in the prior art, due to the lack of effective cooperation of an unmanned aerial vehicle and a ground detector, all-directional and all-time traffic flow perception of a traffic road section is difficult to carry out. Comprising the steps of obtaining an unmanned aerial vehicle take-off and landing facility site selection scheme through mean shift clustering based on a road segment importance index; based on the unmanned aerial vehicle take-off and landing facility site selection scheme, constructing a multi-target optimization model; solving the multi-objective optimization model to obtain an air unmanned aerial vehicle and ground detector traffic sensing network layout scheme; according to the historical traffic data and the current traffic data, constructing a space-time traffic data sequence so as to predict a traffic flow change trend in a period of time in the future; and when the prediction result does not meet the preset condition, dynamically adjusting the weight of the target function so as to output the adjusted layout scheme again.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation technology, and in particular to a method and system for deploying a traffic perception network of aerial drones and ground detectors. Background Art

[0002] With the rapid development of the low-altitude economy, using drones and ground detectors to build a three-dimensional traffic perception network has become an important means to improve traffic management efficiency.

[0003] Traditional traffic sensing network deployments rely on a single, single-site deployment of ground-based sensors, which can provide real-time monitoring of traffic conditions within their coverage area. However, due to cost constraints, these sensors are typically deployed sparsely, typically only at locations with high accident rates or at key traffic nodes. This approach significantly limits detection coverage, particularly because the sensors' fixed locations make them difficult to adapt to random changes in traffic flow, resulting in numerous blind spots.

[0004] When using a single drone deployment, due to their high maneuverability and wide monitoring range, a single drone can monitor multiple roads, thereby improving detection coverage across all sections of a road. However, drones are limited by battery life and automatically return to their home roost after reaching their range, making it difficult to achieve continuous monitoring of road traffic at all times.

[0005] Existing traffic sensing network deployment plans generally lack effective coordination between drones and ground detectors. The above-mentioned independent detection status will lead to blind spots in the spatial dimension of the traffic sensing network, and continuous monitoring cannot be achieved in the temporal dimension. It is difficult to build an all-round, all-time three-dimensional traffic sensing system, and it is unable to meet the growing demand for refined management of intelligent transportation systems. Summary of the Invention

[0006] The embodiments of the present application provide a method and system for deploying a traffic sensing network using aerial drones and ground detectors, which are used to solve the following technical problems: existing traffic sensing network deployment solutions generally lack effective coordination between drones and ground detectors. Independent detection states will cause traffic sensing networks to have coverage blind spots in the spatial dimension, and continuous monitoring cannot be achieved in the time dimension. It is difficult to build a full-dimensional, all-time three-dimensional traffic sensing system, which makes it impossible to meet the growing demand for refined management of intelligent transportation systems.

[0007] The embodiments of this application adopt the following technical solutions:

[0008] The present invention provides a method for deploying a traffic sensing network for aerial drones and ground detectors. The method comprises: selecting a location for drone take-off and landing facilities in a set of preset road section nodes using mean shift clustering based on a road section importance index to obtain a drone take-off and landing facility location plan; constructing a multi-objective optimization model based on the drone take-off and landing facility location plan; solving the multi-objective optimization model to obtain a traffic sensing network deployment plan for aerial drones and ground detectors; constructing a spatiotemporal traffic data sequence based on historical and current traffic data to predict traffic flow trends over a period of time in the future; and, if the predicted traffic flow trends do not meet preset conditions, dynamically adjusting the weights of each objective function in the multi-objective optimization model to re-output an adjusted traffic sensing network deployment plan for aerial drones and ground detectors.

[0009] The embodiment of the present application adopts the mean shift algorithm to propose a method for selecting the location of take-off and landing facilities, which can effectively allocate the take-off and landing facilities to important sections in priority, which is conducive to improving the monitoring coverage of important sections. Secondly, by constructing a multi-objective optimization model and weighing the cost and rescue efficiency, it avoids waste of resources caused by excessive deployment, makes the limited resources optimally configured, and effectively improves the overall performance and operational efficiency of the traffic perception network. The embodiment of the present application can realize the monitoring of multiple sections of roads by one drone, improves the utilization rate of drones, and adopts the same-point take-off and landing method to optimize the layout of drones, which can effectively respond to emergencies, improve the efficiency of event monitoring and the emergency rescue efficiency of the perception network. In addition, the embodiment of the present application accurately predicts the future trend of traffic flow changes, dynamically adjusts the weights of each objective function in the multi-objective optimization model, and then re-optimizes the layout plan of the traffic perception network, realizes the adaptive adjustment of the traffic perception network, and ensures that it always maintains efficient operation in a complex and changing traffic environment.

[0010] In one implementation of the present application, based on the road section importance index, the site selection of drone take-off and landing facilities is carried out in a preset road section node set through mean shift clustering to obtain a drone take-off and landing facility site selection plan, which specifically includes: randomly selecting a road section node in the road section node set as the initial cluster center point; within the preset bandwidth range, searching for reference road section nodes contained in the initial cluster center point area, and determining the kernel density estimate of the reference road section node at the initial cluster center point to find the direction of density increase; and, based on the importance of the road section where each reference road section node is located, determining the mean shift of each reference road section node relative to the initial cluster center point; updating the position of the initial cluster center point based on the mean shift along the direction of density increase; until all cluster center points converge to the corresponding density peak, determining the distance from all reference road section nodes to all cluster centers, and classifying the reference road section nodes that converge to the same position into the same cluster; outputting the cluster set and cluster center position, and using the cluster center site selection position as the drone take-off and landing facility site selection plan.

[0011] In one implementation of the present application, a multi-objective optimization model is constructed based on the site selection plan for drone take-off and landing facilities, specifically including: establishing a cost optimization model based on the site selection plan for drone take-off and landing facilities, and taking minimizing the deployment cost of traffic sensing equipment as the optimization goal; establishing an emergency rescue efficiency optimization model based on the site selection plan for drone take-off and landing facilities, and taking minimizing the total emergency rescue time as the optimization goal; and establishing constraints based on drone-related data and ground detector-related data; and constructing a multi-objective optimization model based on the cost optimization model, the emergency rescue efficiency optimization model and the constraints.

[0012] In one implementation of the present application, a cost optimization model is established based on the site selection plan for drone take-off and landing facilities and minimizing the deployment cost of traffic sensing equipment as the optimization goal. Specifically, the cost optimization model is:

[0013]

[0014] Among them, F1 is the cost optimization model; Ω U Assemble for drones;Ω N is a set of road section nodes; i is any road section node or a candidate point for a UAV landing and take-off facility; j is another road section node or a candidate point for a UAV landing and take-off facility; is the depreciation cost of the drone per unit distance; is the maintenance cost per unit distance of the drone; The working time of the drone; n u is the number of inspections per day by the drone u; d ij is the distance between node i and node j; x uij Is UAV u flying from node i to node j?G A collection of candidate points for drone take-off and landing facilities; The purchase cost of the UAV take-off and landing facilities g; The installation cost of the UAV take-off and landing facilities g; is the maintenance cost of the UAV landing facility g; l g is the service life of the UAV take-off and landing facility g; i Whether the drone has established a drone take-off and landing facility at node i; Ω D is the ground detector set; is the purchase cost of the ground detector d; is the installation cost of ground detector d; is the maintenance cost of ground detector d; x dij Is the ground detector d covering the road section (i, j); r d is the coverage of the ground detector d.

[0015] In one implementation of the present application, based on the site selection plan for drone take-off and landing facilities and minimizing the total emergency rescue time as the optimization goal, an emergency rescue efficiency optimization model is established, specifically including: the emergency rescue efficiency optimization model is:

[0016]

[0017] Among them, F2 is the emergency rescue efficiency optimization model; Ω J is the set of all candidate points of UAV take-off and landing facilities and road section nodes; d gi is the distance between the UAV take-off and landing facility g and node i; d gj is the distance between the UAV take-off and landing facility g and node j; Ω D is a set of ground detectors; d is any ground detector; d ij is the distance between node i and node j; d ci is the distance between road administration vehicle c and node i, d cj is the distance between the road administration vehicle and node j, x dij Is the ground detector covering the road section (i, j); v u is the flight speed of the UAV; v c is the speed of the road maintenance vehicle.

[0018] In one implementation of the present application, constraints are established based on drone-related data and ground detector-related data, specifically including: based on integer constraints, the total flight time of the drone does not exceed the maximum endurance time of the drone, the total flight mileage of the drone does not exceed the maximum endurance mileage, each road section is covered at least once by the drone or ground detector, any ground detector covers at most one road section, the number of drones entering the node is equal to the number of leaving the node, sub-loops of the drone flight path are eliminated, drone take-off and landing facilities are constructed at the drone take-off point, and drone take-off and landing facilities are constructed at the drone landing point.

[0019] In one implementation of the present application, the constraints are established based on integer constraints, the total flight time of the drone does not exceed the maximum endurance of the drone, the total flight mileage of the drone does not exceed the maximum endurance, each road section is covered at least once by the drone or ground detector, any ground detector covers at most one road section, the number of drone entry nodes is equal to the number of exit nodes, sub-loops of the drone flight path are eliminated, drone take-off and landing facilities are built at the drone take-off point, and drone take-off and landing facilities are built at the drone landing point. Specifically, the constraints include:

[0020] Based on function: x uij ∈{0,1};x dij ∈{0,1};z i ∈{0,1};y u ∈{0,1};

[0021] Construct integer constraints;

[0022] Function-based:

[0023] Construct a constraint that the total flight time of the drone should not exceed the maximum endurance of the drone;

[0024] Function-based:

[0025] Construct the constraint that the total flight mileage of the drone is no greater than the maximum cruising range;

[0026] Function-based:

[0027] Construct a constraint that each road segment (i, j) must be covered at least once by a drone or ground detector;

[0028] Function-based:

[0029] Construct the constraint that any ground detector can only cover one road section;

[0030] Function-based:

[0031] Construct a constraint that the number of drones entering a node is equal to the number of drones leaving the node;

[0032] Function-based:

[0033] Construct sub-loop constraints that eliminate the UAV’s flight path;

[0034] Function-based:

[0035] Construct constraints for the construction of drone take-off and landing facilities at the drone take-off point;

[0036] Function-based:

[0037] Construct constraints for the construction of drone take-off and landing facilities at the drone landing site;

[0038] Among them, y u Whether the drone u is selected and covers the road section; t u The maximum flight time of the drone; r u is the cruising range of UAV u; x uij For the drone to fly from point i to point j; x uji The drone flies from point j to point i; t i is the order in which the drone visits node i; t j is the order in which the drone visits node j; N node is the number of nodes in the set N.

[0039] In one implementation of the present application, a spatiotemporal traffic data sequence is constructed based on historical traffic data and current traffic data to predict the trend of traffic flow changes in the future, specifically including: taking road sections as graph nodes and the connection relationships corresponding to the road sections as edges to construct a traffic network topology graph; performing convolution operations on the traffic network topology graph through a spatiotemporal graph convolution network to capture the spatial features between road sections; using a time convolution network to extract traffic data sequences based on historical traffic data and current traffic data; inputting the traffic data sequence into a preset long-short term memory network to predict the traffic flow corresponding to the road section to obtain a reference prediction result; performing weighted fusion on the obtained spatial features and reference prediction results to obtain the spatiotemporal feature vectors corresponding to each road section; arranging the spatiotemporal feature vectors corresponding to each road section at different times in sequence to construct a spatiotemporal traffic data sequence to obtain a prediction result of the traffic flow change trend in the future.

[0040] In one implementation of the present application, when the predicted traffic flow change trend does not meet the preset conditions, the weight of each objective function in the multi-objective optimization model is dynamically adjusted, specifically including: when the predicted traffic flow change trend does not meet the preset conditions, obtaining the key traffic indicators corresponding to the road section; wherein the key traffic indicators include at least the congestion index, the frequency of accidents and the emergency rescue response delay; determining the correlation between each objective function and the key traffic indicators; determining the first indicator weight corresponding to each objective function based on the correlation; and, based on the numerical size of the key traffic indicators, determining the second indicator weight corresponding to each objective function; based on the first indicator weight and the second indicator weight, dynamically adjusting the weight of each objective function in the multi-objective optimization model.

[0041] An embodiment of the present application provides a traffic perception network deployment system for aerial drones and ground detectors, including: a drone take-off and landing facility site selection unit, a multi-objective optimization model construction unit, a deployment plan solving unit, a prediction unit and a dynamic adjustment unit; the drone take-off and landing facility site selection unit, based on a road section importance index, performs drone take-off and landing facility site selection in a preset road section node set through mean shift clustering to obtain a drone take-off and landing facility site selection plan; the multi-objective optimization model construction unit, based on the drone take-off and landing facility site selection plan, constructs a multi-objective optimization model; the deployment plan solving unit, by solving the multi-objective optimization model, obtains a traffic perception network deployment plan for aerial drones and ground detectors; the prediction unit, based on historical traffic data and current traffic data, constructs a spatiotemporal traffic data sequence to predict the traffic flow change trend in a future period of time; the dynamic adjustment unit, when the predicted traffic flow change trend does not meet the preset conditions, dynamically adjusts the weights of each objective function in the multi-objective optimization model to re-output the adjusted traffic perception network deployment plan for aerial drones and ground detectors.

[0042] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: The embodiments of the present application adopt the mean shift algorithm to propose a method for selecting the location of take-off and landing facilities, which can effectively prioritize the allocation of take-off and landing facilities to important sections of road, which is conducive to improving the monitoring coverage of important sections of road. Secondly, by constructing a multi-objective optimization model and balancing the cost and rescue efficiency, it avoids waste of resources caused by excessive deployment, so that limited resources are optimally configured, and effectively improves the overall performance and operational efficiency of the traffic perception network. The embodiments of the present application can realize the monitoring of multiple sections of road by one drone, thereby improving the utilization rate of drones, and adopting the same-point take-off and landing method to optimize the layout of drones, which can effectively respond to emergencies, improve the efficiency of event monitoring and the emergency rescue efficiency of the perception network. In addition, the embodiments of the present application accurately predict the future trend of traffic flow changes, dynamically adjust the weights of each objective function in the multi-objective optimization model, and then re-optimize the layout plan of the traffic perception network, thereby realizing adaptive adjustment of the traffic perception network and ensuring that it always maintains efficient operation in a complex and changing traffic environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present application. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0044] Figure 1 A flow chart of a method for deploying a traffic sensing network of aerial drones and ground detectors provided in an embodiment of the present application;

[0045] Figure 2 A structural diagram of a traffic sensing network deployment system for aerial drones and ground detectors provided in an embodiment of the present application.

[0046] Reference numerals:

[0047] 200 Aerial UAV and ground detector traffic perception network deployment system, 201 UAV take-off and landing facility site selection unit, 202 Multi-objective optimization model construction unit, 203 Deployment plan solution unit, 204 Prediction unit, 205 Dynamic adjustment unit. DETAILED DESCRIPTION

[0048] The embodiments of the present application provide a method and system for deploying a traffic sensing network of aerial drones and ground detectors.

[0049] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0050] The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0051] It should be noted that the ground detector in the embodiment of the present application is specifically a fixed detector, which refers to a traffic detection device that is placed in a traffic network at a pre-planned location and is placed for a long time and stably. Its location will not change easily, and it is used to continuously monitor the traffic conditions in a specific area.

[0052] Figure 1 This is a flow chart of a method for deploying a traffic sensing network of an aerial drone and ground detectors provided in an embodiment of the present application. Figure 1 As shown, the method for deploying a traffic sensing network of aerial drones and ground detectors includes the following steps:

[0053] S101. Based on the road section importance index, the location of UAV take-off and landing facilities is selected in the preset road section node set through mean shift clustering to obtain a UAV take-off and landing facility location selection plan; wherein the UAVs are set to the same take-off and landing point.

[0054] In one implementation of the present application, a road section accident rate measurement index is constructed.

[0055] (1) Obtain basic data of road section:

[0056] Collect information such as traffic data, accident data, and road section length, and statistically analyze the basic data of the road section based on latitude and longitude coordinates.

[0057] (2) Establishing road section accident rate indicators:

[0058] The road section accident rate represents the accident rate per 10,000 vehicles in a certain road network unit within a certain period of time, and measures the safety of road travel. In the embodiment of the present application, the maximum value of the accident rates of all road sections in the path is selected as the path accident rate.

[0059]

[0060] Among them, H p represents the accident rate of 10,000 vehicles on path p; n ep represents the number of accidents that occurred on section e of path p during the key observation period, v eprepresents the average daily traffic volume (pcu / h) of the road segment e on the path p during the observation period; l ep represents the length (km) of segment e on path p.

[0061] In one implementation of the present application, a road section importance model is constructed based on information entropy theory.

[0062] By integrating the traffic flow and accident rate indicators of road sections, a road section importance model based on information entropy theory is constructed to identify traffic flow bottlenecks or accident-prone sections, providing a scientific basis for the layout optimization of traffic detectors.

[0063] (1) Index normalization

[0064]

[0065] Where r eq represents the original value of the index q corresponding to the road segment e, λ eq represents the normalized value of the index q corresponding to the road segment e, r q Indicates the value of indicator q in the statistical interval n.

[0066] (2) Calculation of indicator information entropy

[0067]

[0068] Where s q The information entropy of the indicator q, a is the number of road section importance indicators.

[0069] (3) Calculation of indicator weights

[0070]

[0071] Where, ω q Represents the weight of indicator q.

[0072] (4) Calculation of indicator weights

[0073]

[0074] Where, ω q Represents the weight of indicator q.

[0075] (5) Calculation of road section importance

[0076]

[0077] Where W e Indicates the importance of road segment e; ω H represents the accident rate indicator weight; ω V represents the weight of traffic flow index; represents the information entropy of the accident rate index of road section e; Represents the information entropy of traffic flow index of road section e.

[0078] In one implementation of the present application, the site selection of UAV take-off and landing facilities is based on mean shift clustering.

[0079] This embodiment of the application integrates the road section importance index and uses mean shift clustering to select the location of drone take-off and landing facilities. The main steps are as follows:

[0080] (1) Section node set Bandwidth parameter h, moving distance threshold τ.

[0081] (2) Randomly select a road section node x as the initial cluster center.

[0082] (3) Search for the road segment nodes within the cluster center within the bandwidth h, calculate the kernel density estimate of these nodes at the cluster center based on Equations 7 and 8, and find the direction of increasing density.

[0083] The mean drift of each road node is calculated according to Formula 9.

[0084]

[0085] Where x represents the node of the current iteration; h represents the bandwidth, where the UAV's range is used as the bandwidth; K(x) represents the node kernel function.

[0086]

[0087] Where f(x) is the kernel density estimation function, n h is the number of nodes included in the current bandwidth; S h (x) is the high-dimensional ball area with a search radius of h centered on x;

[0088]

[0089] Where M h (x) represents x i The mean shift relative to node x; W(x i ) represents the importance of the road section where the node is located.

[0090] (4) Update the position of the cluster center based on formula 10 and move it to the new position x′ along the direction of increasing density. Check the moving distance of the cluster center. If the moving distance || M h If (x)|| is less than the threshold τ, the cluster center is considered to have converged; otherwise, return to (2).

[0091] x′=M h (x)+x (10)

[0092] (5) Repeat (3)-(4) until all cluster centers converge to their density peaks, calculate the distances from all nodes to all cluster centers, and group the nodes that converge to the same position into the same cluster.

[0093] (6) Output the cluster set and cluster center location, and use the cluster center location as the site selection plan for UAV take-off and landing facilities.

[0094] S102. Construct a multi-objective optimization model based on the site selection plan for UAV take-off and landing facilities.

[0095] In one implementation of the present application, an air-ground integrated traffic detector deployment model is constructed taking into account deployment costs and emergency rescue response efficiency.

[0096] The embodiments of the present application consider factors such as the cost-effectiveness and time-effectiveness of the traffic sensing network, the performance of different types of traffic sensing equipment, the continuity of the drone's flight path, etc., to construct an air-ground integrated traffic detector deployment optimization model and optimize the detector layout.

[0097] The embodiments of this application comprehensively consider the complexity and difficulty of solving actual scenarios, improve the efficiency of model solving, and make reasonable assumptions about actual scenarios, as follows:

[0098] ① It is assumed that drone take-off and landing facilities can be deployed at any location on both sides of the highway without considering the impact of the surrounding environment.

[0099] ② Assume that the power supply, water supply and drainage of facilities around the road are in good condition, which can ensure the effective operation of the drone take-off and landing facilities.

[0100] ③ The signal blocking caused by terrain factors such as mountains and buildings is not taken into account, that is, drone communication is not restricted.

[0101] ④ Assume that all drones can autonomously avoid obstacles and do not interfere with each other within the spatial range during the inspection process.

[0102] ⑤ It is assumed that the model uses the same type of UAV take-off and landing facilities and that the deployment costs at different locations are consistent.

[0103] ⑥ Assume that the drone take-off and landing facilities can accommodate multiple drones of different sizes, and the capacity meets the drone landing requirements.

[0104] ⑦ Assume that the UAV only performs one-way inspection during path monitoring.

[0105] In one implementation of this application, a cost optimization model is established based on the site selection plan for drone takeoff and landing facilities, with minimizing the deployment cost of traffic sensing equipment as the optimization objective. An emergency rescue efficiency optimization model is also established based on the site selection plan for drone takeoff and landing facilities, with minimizing the total emergency rescue time as the optimization objective. Based on the cost optimization model and the emergency rescue efficiency optimization model, a multi-objective optimization model is constructed, and constraints are established based on drone-related data and ground detector-related data.

[0106] Specifically, the cost optimization model is:

[0107]

[0108] Among them, F1 is the cost optimization model; Ω U Assemble for drones;Ω N is a set of road section nodes; i is any road section node or a candidate point for a UAV landing and take-off facility; j is another road section node or a candidate point for a UAV landing and take-off facility; is the depreciation cost of the drone per unit distance; is the maintenance cost per unit distance of the drone; The working time of the drone; n u is the number of inspections per day by the drone u; d ij is the distance between node i and node j; x uij Is UAV u flying from node i to node j? G A collection of candidate points for drone take-off and landing facilities; The purchase cost of the UAV take-off and landing facilities g; The installation cost of the UAV take-off and landing facilities g; is the maintenance cost of the UAV landing facility g; l g is the service life of the UAV take-off and landing facility g; i Whether the drone has established a drone take-off and landing facility at node i; Ω D is the ground detector set; is the purchase cost of the ground detector d; is the installation cost of ground detector d; is the maintenance cost of ground detector d; x dij Is the ground detector d covering the road section (i, j); r d is the coverage of the ground detector d.

[0109] Furthermore, the emergency rescue efficiency optimization model is:

[0110]

[0111] Among them, F2 is the emergency rescue efficiency optimization model; Ω Jis the set of all candidate points of UAV take-off and landing facilities and road section nodes; d gi is the distance between the UAV take-off and landing facility g and node i; d gj is the distance between the UAV take-off and landing facility g and node j; d ij is the distance between node i and node j; Ω D is a set of ground detectors; d is any ground detector; d ci is the distance between road administration vehicle c and node i, d cj is the distance between the road administration vehicle and node j, x dij Is the ground detector covering the road section (i, j); v u is the flight speed of the UAV; v c is the speed of the road maintenance vehicle.

[0112] Furthermore, based on the function:

[0113] x uij ∈{0,1} (13)

[0114] x dij ∈{0,1} (14)

[0115] z i ∈{0,1} (15)

[0116] y u ∈{0,1} (16)

[0117] Construct integer constraints;

[0118] Function-based:

[0119]

[0120] Construct a constraint that the total flight time of the drone should not exceed the maximum endurance of the drone;

[0121] Function-based:

[0122]

[0123] Construct the constraint that the total flight mileage of the drone is no greater than the maximum cruising range;

[0124] Function-based:

[0125]

[0126] Construct a constraint that each road segment (i, j) must be covered at least once by a drone or ground detector; based on the function:

[0127]

[0128] Construct the constraint that any ground detector can only cover one road section;

[0129] Function-based:

[0130]

[0131] Construct a constraint that the number of drones entering a node is equal to the number of drones leaving the node;

[0132] Function-based:

[0133]

[0134] Construct sub-loop constraints that eliminate the UAV’s flight path;

[0135] Function-based:

[0136]

[0137] Construct constraints for the construction of drone take-off and landing facilities at the drone take-off point;

[0138] Function-based:

[0139]

[0140] Construct constraints for the construction of drone take-off and landing facilities at the drone landing site;

[0141] Among them, y u Whether the drone u is selected and covers the road section; t u The maximum flight time of the drone; r u is the cruising range of UAV u; x uij For the drone to fly from point i to point j; x uji The drone flies from point j to point i; t i is the order in which the drone visits node i; t j is the order in which the drone visits node j; N node is the number of nodes in the set N.

[0142] S103. By solving the multi-objective optimization model, a traffic perception network layout plan for aerial drones and ground detectors is obtained.

[0143] In one implementation of the present application, the optimization model constructed in the embodiment of the present application is a 0-1 integer programming model, and the branch and bound method is used to solve the model. Through the branch and prune strategy, the combinatorial explosion problem of the solution space is effectively controlled. The algorithm flow is as follows:

[0144] Step 1: Set the optimal solution of the traffic sensing equipment deployment optimization problem to X *, represents the optimal layout of traffic sensing equipment, and the relaxed subproblem of this optimization problem is for The optimal solution of .

[0145] Step 2: Solve The optimal solution and the minimum value of the objective function like If the 0-1 integer requirement is met, the algorithm terminates and outputs the optimal traffic sensing equipment layout plan; otherwise The minimum value of the objective function of the problem is used as the lower bound of IP Z , take the objective function value of the 0-1 feasible solution of IP as the initial upper bound

[0146] Step 3: From Select a solution that does not satisfy the 0-1 integer constraint make Join separately We can obtain two sub-problems and find the optimal linear programming solution for the sub-problems.

[0147] Step 4: Select the minimum value of the objective function corresponding to the solution as the new upper bound Determine whether each subproblem has a 0-1 integer solution. If so, the minimum value of the objective function that satisfies the 0-1 integer solution is used as the new lower bound. Z , otherwise go to Step 5.

[0148] Step 5: Compare the solutions of each branch subproblem. If there is no feasible solution or the objective function is greater than the new upper bound, Pruning.

[0149] Step 6: If If there is no sub-problem to be considered in , the optimal feasible solution of the current objective function is the optimal solution of the original problem IP, and the final traffic sensing equipment deployment plan is output; otherwise, go to Step 3.

[0150] S104: Construct a spatiotemporal traffic data sequence based on historical traffic data and current traffic data to predict traffic flow change trends in the future.

[0151] In one implementation of the present application, a traffic network topology graph is constructed by taking road sections as graph nodes and the connection relationships corresponding to the road sections as edges. The traffic network topology graph is convolved by a spatiotemporal graph convolutional network to capture the spatial features between road sections. Based on historical traffic data and current traffic data, a temporal convolutional network is used to extract traffic data sequences. The traffic data sequence is input into a preset long short-term memory network to predict the traffic flow corresponding to the road section and obtain a reference prediction result. The obtained spatial features and reference prediction results are weighted and fused to obtain the spatiotemporal feature vectors corresponding to each road section. The spatiotemporal feature vectors corresponding to each road section at different times are arranged in sequence to construct a spatiotemporal traffic data sequence to obtain a prediction result of the traffic flow change trend in the future.

[0152] Specifically, a traffic network topology is constructed using a graph data structure based on the identified road segment nodes and connection relationships. This constructed traffic network topology is then fed into a spatiotemporal graph convolutional network. During the convolution operation, the spatiotemporal graph convolutional network learns the spatial features between road segments and maps these features into a low-dimensional vector space. Each road segment node has a corresponding vector representation in this vector space, which contains information about the spatial associations between that segment and other road segments. This approach effectively extracts the spatial features of road segments, providing important spatial dimension information for subsequent analysis and prediction.

[0153] Furthermore, historical and real-time traffic data are collected through traffic monitoring devices installed on different road sections, as well as monitoring devices installed on drones. This data can include information such as traffic flow, vehicle speed, and vehicle density at different time points. The processed traffic data sequence is input into a temporal convolutional network. By applying convolution operations on the time dimension, the temporal convolutional network can automatically extract temporal features from the traffic data. The traffic data sequence processed by the temporal convolutional network is further input into a pre-set long-short-term memory network. Based on the input traffic data sequence, the long-short-term memory network learns the dependencies between different time points and outputs a prediction of future traffic flow, namely a reference prediction result. This reference prediction result contains information about the changing trends and characteristics of traffic flow in the temporal dimension.

[0154] Furthermore, the spatial and temporal feature vectors are weighted and summed according to the configured weights to obtain the spatiotemporal feature vector corresponding to each road segment. The spatiotemporal feature vectors corresponding to each road segment at different time points are then arranged in chronological order. By sequentially arranging the spatiotemporal feature vectors for all road segments at different time points, a complete spatiotemporal traffic data sequence is constructed. Ultimately, this spatiotemporal traffic data sequence is output as the basis for predicting traffic flow trends over a period of time.

[0155] S105. When the predicted traffic flow change trend does not meet the preset conditions, the weights of each objective function in the multi-objective optimization model are dynamically adjusted to re-output the adjusted traffic perception network layout plan of the aerial drones and ground detectors.

[0156] In one implementation of the present application, when the predicted traffic flow change trend does not meet preset conditions, key traffic indicators corresponding to the road section are obtained; wherein the key traffic indicators include at least congestion index, accident frequency, and emergency rescue response delay. The correlation between each objective function and the key traffic indicators is determined. Based on the correlation, the first indicator weight corresponding to each objective function is determined. And, based on the numerical value of the key traffic indicator, the second indicator weight corresponding to each objective function is determined. Based on the first indicator weight and the second indicator weight, the weight of each objective function in the multi-objective optimization model is dynamically adjusted.

[0157] Specifically, real-time traffic data, including road section traffic flow, vehicle speed, and lane occupancy, is collected from multiple sources, including ground traffic sensors and drone inspection video data, to determine key traffic indicators. For example, a congestion index for each road section is calculated based on a ratio of traffic flow to road capacity, or a complex model that takes into account factors such as vehicle speed and vehicle density.

[0158] Furthermore, determine how the congestion index, accident frequency, and emergency rescue response delays affect the cost of equipment deployment. For example, sections with a high congestion index may require more monitoring equipment to monitor traffic conditions in real time, thereby increasing costs; sections with a high accident frequency may require more advanced equipment to improve accident monitoring and early warning capabilities, which will also lead to increased costs. Analyze the relationship between the congestion index, accident frequency, and the total time for emergency rescue. A high congestion index will cause rescue vehicles to slow down, extending rescue time; a high accident frequency may require more rescue resources to be deployed, which will also affect the total time for emergency rescue. Collect a large amount of historical traffic data, including the actual values ​​of key traffic indicators and objective functions, and use correlation analysis to determine the quantitative relationship between each objective function and each key traffic indicator, and obtain a correlation value.

[0159] Normalize the correlation values ​​between each objective function and each key traffic indicator so that their sum is 1. Assign the first indicator weight to each objective function based on the normalized correlation value. For each road section, compare its congestion index value with the set threshold. If the congestion index is in the light congestion range, the second indicator weight of the corresponding objective function can be set to a lower value; if it is in the moderate congestion range, the weight is appropriately increased; if it is in the heavy congestion range, the weight is further increased. Adjust the second indicator weight of each objective function based on the comparison result of the accident frequency value and the threshold. The higher the accident frequency, the greater the weight of the corresponding objective function. Determine the second indicator weight of each objective function based on the comparison of the emergency rescue response delay value and the threshold.

[0160] Furthermore, the weighted sum of the first and second indicator weights for each objective function is calculated to obtain a final weight adjustment value. This final weight adjustment value is then applied to the multi-objective optimization model to update the weights of each objective function. Using the updated weights, the multi-objective optimization model is re-solved to obtain an adjusted traffic sensing network layout for aerial drones and ground detectors to accommodate situations where the predicted traffic flow trend does not meet the preset conditions.

[0161] Figure 2 This is a schematic diagram of a traffic sensing network deployment system for aerial drones and ground detectors provided in an embodiment of the present application. Figure 2 As shown, the traffic perception network deployment system 200 for aerial drones and ground detectors includes a drone take-off and landing facility site selection unit 201, a multi-objective optimization model construction unit 202, a deployment plan solving unit 203, a prediction unit 204 and a dynamic adjustment unit 205. The UAV take-off and landing facility site selection unit 201 selects the site of the UAV take-off and landing facility in the preset road section node set based on the road section importance index through mean shift clustering to obtain the UAV take-off and landing facility site selection plan; the multi-objective optimization model construction unit 202 constructs a multi-objective optimization model based on the UAV take-off and landing facility site selection plan; the layout plan solving unit 203 obtains the traffic perception network layout plan of the aerial UAV and ground detector by solving the multi-objective optimization model; the prediction unit 204 constructs a spatiotemporal traffic data sequence based on historical traffic data and current traffic data to predict the traffic flow change trend in the future; the dynamic adjustment unit 205 dynamically adjusts the weights of each objective function in the multi-objective optimization model when the predicted traffic flow change trend does not meet the preset conditions to re-output the adjusted traffic perception network layout plan of the aerial UAV and ground detector.

[0162] The various embodiments in this application are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simple. For relevant portions, refer to the descriptions of the method embodiments.

[0163] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. It will be apparent to those skilled in the art that various modifications and variations may be made to the embodiments of the present application. However, such modifications or substitutions do not deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A method for deploying a traffic sensing network of aerial drones and ground detectors, characterized in that: The method comprises: Based on the road section importance index, the location of UAV take-off and landing facilities is selected in the set of preset road section nodes through mean shift clustering, and the location selection plan of UAV take-off and landing facilities is obtained; Based on the UAV take-off and landing facility site selection plan, a multi-objective optimization model is constructed; By solving the multi-objective optimization model, a traffic sensing network layout plan for aerial drones and ground detectors is obtained; Based on historical and current traffic data, a spatiotemporal traffic data sequence is constructed to predict traffic flow trends in the future. When the predicted traffic flow change trend does not meet the preset conditions, the weights of each objective function in the multi-objective optimization model are dynamically adjusted to re-output the adjusted traffic perception network layout plan of the aerial drone and ground detectors.

2. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 1 is characterized in that: The method of selecting the location of UAV take-off and landing facilities in a set of preset road section nodes based on the road section importance index by using mean shift clustering can obtain a UAV take-off and landing facility location scheme, which specifically includes: In the set of road segment nodes, a road segment node is randomly selected as the initial cluster center point; Searching for reference road segment nodes within the area of ​​the initial cluster center within a preset bandwidth, and determining kernel density estimates of the reference road segment nodes at the initial cluster center to find a direction of increasing density; and, based on the importance of the road section where each reference road section node is located, determining the mean drift of each reference road section node relative to the initial cluster center point; updating the position of the initial cluster center point based on the mean shift along the direction of increasing density; Until all cluster center points converge to corresponding density peaks, the distances from all reference road segment nodes to all cluster centers are determined, and the reference road segment nodes that converge to the same position are grouped into the same cluster; The cluster set and cluster center location are output, and the cluster center location is used as the site selection plan for the UAV take-off and landing facility.

3. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 1 is characterized in that: The multi-objective optimization model is constructed based on the site selection plan of the UAV take-off and landing facilities, specifically including: Based on the site selection plan for the drone take-off and landing facilities, and taking minimizing the deployment cost of traffic sensing equipment as the optimization goal, a cost optimization model is established; Based on the site selection plan for the drone take-off and landing facilities, and taking minimizing the total emergency rescue time as the optimization goal, an emergency rescue efficiency optimization model is established; and, establishing constraints based on the drone-related data and the ground detector-related data; The multi-objective optimization model is constructed based on the cost optimization model, the emergency rescue efficiency optimization model and the constraint conditions.

4. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 3 is characterized in that: Based on the site selection plan for the UAV take-off and landing facilities and taking minimizing the deployment cost of traffic sensing equipment as the optimization goal, a cost optimization model is established, specifically including: The cost optimization model is: Among them, F1 is the cost optimization model; Ω U Assemble for drones;Ω N is a set of road section nodes; i is any road section node or a candidate point for a UAV landing and take-off facility; j is another road section node or a candidate point for a UAV landing and take-off facility; is the depreciation cost of the drone per unit distance; is the maintenance cost per unit distance of the drone; The working time of the drone; n u is the number of inspections per day by the drone u; d ij is the distance between node i and node j; x uij Is UAV u flying from node i to node j? G A collection of candidate points for drone take-off and landing facilities; The purchase cost of the UAV take-off and landing facilities g; The installation cost of the drone take-off and landing facilities g; is the maintenance cost of the UAV landing facility g; l g is the service life of the UAV take-off and landing facility g; i Whether the drone has established a drone take-off and landing facility at node i; Ω D is the ground detector set; is the purchase cost of the ground detector d; is the installation cost of ground detector d; is the maintenance cost of ground detector d; x dij Is the ground detector d covering the road section (i, j); r d is the coverage of the ground detector d.

5. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 3 is characterized in that: Based on the site selection plan for the UAV take-off and landing facilities, and taking minimizing the total emergency rescue time as the optimization goal, an emergency rescue efficiency optimization model is established, specifically including: The emergency rescue efficiency optimization model is: Among them, F2 is the emergency rescue efficiency optimization model; Ω J is the set of all candidate points of UAV take-off and landing facilities and road section nodes; d gi is the distance between the UAV take-off and landing facility g and node i; d gj is the distance between the UAV take-off and landing facility g and node j; Ω D is a set of ground detectors; d is any ground detector; d ij is the distance between node i and node j; d ci is the distance between road administration vehicle c and node i; d cj is the distance between the road administration vehicle and node j; x dij Is the ground detector covering the road section (i, j); v u is the flight speed of the UAV; v c is the speed of the road maintenance vehicle.

6. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 3 is characterized in that: The constraint conditions are established based on the drone-related data and the ground detector-related data, specifically including: The constraints are established based on integer constraints, the total flight time of the drone does not exceed the maximum endurance time of the drone, the total flight mileage of the drone does not exceed the maximum endurance mileage, each road section is covered by a drone or a ground detector at least once, any ground detector covers at most one road section, the number of drones entering nodes is equal to the number of leaving nodes, sub-loops of the drone flight path are eliminated, drone take-off and landing facilities are constructed at the drone take-off point, and drone take-off and landing facilities are constructed at the drone landing point.

7. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 6, characterized in that: The constraints are based on integer constraints, the total flight time of the drone does not exceed the maximum endurance of the drone, the total flight mileage of the drone does not exceed the maximum endurance, each road section is covered at least once by the drone or the ground detector, any ground detector covers at most one road section, the number of drone entry nodes is equal to the number of exit nodes, sub-loops of the drone flight path are eliminated, drone take-off and landing facilities are built at the drone take-off point, and drone take-off and landing facilities are built at the drone landing point. Specifically, the constraints include: Based on function: x uij ∈{0,1};x dij ∈{0,1};z i ∈{0,1};y u ∈{0,1}; Construct integer constraints; Function-based: Construct a constraint that the total flight time of the drone should not exceed the maximum endurance of the drone; Function-based: Construct the constraint that the total flight mileage of the drone is no greater than the maximum cruising range; Function-based: Construct a constraint that each road segment (i, j) must be covered at least once by a drone or ground detector; Function-based: Construct the constraint that any ground detector can only cover one road section; Function-based: Construct a constraint that the number of drones entering a node is equal to the number of drones leaving the node; Function-based: Construct sub-loop constraints that eliminate the UAV’s flight path; Function-based: Construct constraints for the construction of drone take-off and landing facilities at the drone take-off point; Function-based: Construct constraints for the construction of drone take-off and landing facilities at the drone landing site; Among them, y u Whether the drone u is selected and covers the road section; t u The maximum flight time of the drone; r u is the cruising range of UAV u; x uij For the drone to fly from point i to point j; x uji The drone flies from point j to point i; t i is the order in which the drone visits node i; t j is the order in which the drone visits node j; N node is the number of nodes in the set N.

8. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 1, characterized in that: The construction of a spatiotemporal traffic data sequence based on historical traffic data and current traffic data to predict traffic flow trends in the future includes: Using road segments as graph nodes and corresponding connection relationships of the road segments as edges to construct a traffic network topology graph; Performing a convolution operation on the traffic network topology graph through a spatiotemporal graph convolutional network to capture the spatial features between the road sections; Based on the historical traffic data and the current traffic data, a temporal convolutional network is used to extract traffic data sequences; Inputting the traffic data sequence into a preset long short-term memory network to predict the traffic flow corresponding to the road section to obtain a reference prediction result; Performing weighted fusion on the obtained spatial features and the reference prediction results to obtain the spatiotemporal feature vectors corresponding to each of the road sections; The spatiotemporal feature vectors corresponding to the road sections at different times are arranged in sequence to construct the spatiotemporal traffic data sequence, so as to obtain a prediction result of the traffic flow change trend in a future period of time.

9. The method for deploying a traffic sensing network of aerial drones and ground detectors according to claim 1, characterized in that: When the predicted traffic flow change trend does not meet the preset conditions, dynamically adjusting the weights of the objective functions in the multi-objective optimization model specifically includes: When the predicted traffic flow change trend does not meet the preset conditions, obtaining key traffic indicators corresponding to the road section; wherein the key traffic indicators include at least congestion index, accident frequency, and emergency rescue response delay; Determining the correlation between each of the objective functions and the key traffic indicators; Determining first indicator weights corresponding to the objective functions based on the correlation degree; and, based on the numerical values ​​of the key traffic indicators, determining the second indicator weights corresponding to the respective objective functions; Based on the first indicator weight and the second indicator weight, the weight of each objective function in the multi-objective optimization model is dynamically adjusted.

10. A traffic sensing network deployment system for aerial drones and ground detectors, characterized by: Applied to the method for deploying a traffic sensing network of aerial drones and ground detectors as claimed in any one of claims 1 to 9, the system comprises a drone take-off and landing facility site selection unit, a multi-objective optimization model construction unit, a deployment solution solution unit, a prediction unit, and a dynamic adjustment unit; The UAV landing facility site selection unit selects the location of UAV landing facilities in a set of preset road section nodes based on the road section importance index through mean shift clustering to obtain the UAV landing facility site selection plan; A multi-objective optimization model building unit, which builds a multi-objective optimization model based on the UAV take-off and landing facility site selection plan; A deployment solution unit, which obtains a deployment solution for the traffic sensing network of aerial drones and ground detectors by solving the multi-objective optimization model; The prediction unit constructs a spatiotemporal traffic data sequence based on historical and current traffic data to predict the traffic flow trend in the future. A dynamic adjustment unit dynamically adjusts the weights of each objective function in the multi-objective optimization model when the predicted traffic flow change trend does not meet the preset conditions, so as to re-output an adjusted traffic perception network layout plan for aerial drones and ground detectors.

Images