Forest fire monitoring method based on unmanned aerial vehicle auxiliary edge calculation

Abstract

The invention provides a forest fire monitoring method based on unmanned aerial vehicle auxiliary edge calculation, and belongs to the technical field of forest fire detection. Assisting sensor nodes in the forest to perform edge calculation through the hovering unmanned aerial vehicle group; taking the maximum calculation time delay in all the unmanned aerial vehicles and the maximum flight energy consumption in the minimum unmanned aerial vehicles as target functions; constructing a model by taking a two-dimensional region interval, an unmanned aerial vehicle hovering height interval, a sensor node transmission power interval, an interval of computing resources allocated to sensor nodes by the unmanned aerial vehicle and a delay threshold as constraint conditions; updating a continuous solution part of the model by utilizing a genetic algorithm, a grey wolf algorithm and a doliosis algorithm, updating a discrete solution part by utilizing a random mode, and selecting an optimal solution according to a Pareto dominating principle; and the hovering position of the unmanned aerial vehicle with the lowest flight energy consumption and the minimum time delay is obtained. The calculation efficiency and the transmission efficiency of the sensor nodes are improved through the unmanned aerial vehicle; and the hovering position with the minimum calculation time delay is selected, so that the data calculation efficiency is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of forest fire detection, and in particular to a forest fire monitoring method based on drone-assisted edge computing. Background Art

[0002] Forest fires can cause serious damage to forest ecosystems, leading to a decrease in forest biomass, reduced productivity, and even casualties of humans and animals. Therefore, effective forest fire monitoring is crucial for forest management.

[0003] In order to monitor forest fires in real time, multiple sensor nodes are deployed in the forest in the prior art. The computing delay of sensor nodes is an important measurement indicator, and the sensor computing delay will affect the response time of forest fire monitoring. However, in actual detection scenarios, the computing power of sensor nodes is limited, and the distance from the central computing device is far, resulting in a large computing delay, which affects the fire detection results.

[0004] Therefore, there is a need for a forest fire detection method that can reduce the computational delay of sensor nodes. Summary of the invention

[0005] In view of this, the present invention provides a forest fire monitoring method based on UAV-assisted edge computing. By utilizing the flexible, mobile, and cost-effective characteristics of UAVs, UAVs are used as aerial edge nodes to assist sensor nodes in edge computing, and a hovering position with the smallest computing delay is selected for the UAV to further reduce the computing delay of the sensor nodes.

[0006] To this end, the present invention provides the following technical solutions:

[0007] A forest fire monitoring method based on drone-assisted edge computing, comprising:

[0008] Using hovering drone swarms to assist sensor nodes in the forest to perform edge computing;

[0009] The objective function is to minimize the maximum computing delay among all drones and minimize the maximum flight energy consumption among drones;

[0010] The model is constructed with the two-dimensional area interval, the drone hovering height interval, the sensor node transmission power interval, the drone allocated computing resource interval to the sensor node, and the delay threshold as constraints;

[0011] The continuous solution part of the model is updated using genetic algorithm, gray wolf algorithm and salp algorithm, the discrete solution part of the model is updated in a random manner, and the optimal solution is selected according to the Pareto dominance principle; the hovering position of each UAV with the lowest flight energy consumption and the smallest calculation delay is obtained.

[0012] Furthermore, the objective function:

[0013] F = min{f1, f2}

[0014] in, is the local computation latency of the jth task, is the computation delay of the UAV of the jth task; f1 is the maximum computation delay among all sensor nodes; f2 is the maximum flight energy consumption among all UAVs, i is the index of the UAV; E i It represents the energy consumed by the drone from its initial position to the hovering position.

[0015] Furthermore, the energy consumed by the drone from the initial position to the hovering position is:

[0016]

[0017] in, Indicates the distance from the initial position to the hovering position of the drone.

[0018] Furthermore, the two-dimensional area interval, the drone hovering height interval, the sensor node transmission power interval, the drone allocated sensor node computing resource interval, and the delay threshold include:

[0019] L min ≤X i ≤L max

[0020] L min ≤Y i ≤L max

[0021] Z min ≤Z i ≤Z max

[0022]

[0023] TD j ≤T th

[0024] Among them, L min and L max are the lower and upper bounds of the two-dimensional region, Z min and Z max are the lower and upper bounds of the drone’s hovering height, respectively. min and p max are the lower and upper bounds of the sensor node transmission power, respectively, and f min and f max are the lower and upper bounds of the computing resources allocated by the UAV to the sensor nodes; Tth Indicates the latency threshold.

[0025] Furthermore, the constraints also include:

[0026]

[0027]

[0028] a ij ∈{0, 1}

[0029] Among them, Ca l and Ca u They represent the task size of the local sensor node and the UAV calculation, and a ij Indicates whether the sensor node offloads the computing task to the drone.

[0030] Furthermore, the model input parameters include:

[0031] The three-dimensional position X, Y, Z of the drone and the transmission power p of the sensor node t , the number of drones M, the computing frequency f assigned by the drone to each sensor node u , the task size Ca of the local sensor node l , the task size Ca calculated by the drone u , the number of sensors K and the subordinate relationship a between the UAV and the sensor nodes.

[0032] Furthermore, the method of updating the continuous solution part of the model by using a genetic algorithm, a gray wolf algorithm and a salp algorithm, updating the discrete solution part of the model by a random method and selecting the optimal solution according to the Pareto dominance principle includes:

[0033] Step 1: Randomly initialize the population and calculate the objective function; a population includes several groups of solutions;

[0034] Step 2: For each solution, a set of UAV and sensor affiliation relationships is regenerated, and the objective function is recalculated to obtain the current solution of this step. The solution in step 1 and the current solution of this step are compared according to the Pareto dominance principle, and the better solution is used as the solution of this step.

[0035] Step 3: Use the genetic algorithm to update the continuous solution part, recalculate the objective function, and obtain the current solution of this step. Compare the solution in step 2 with the current solution of this step according to the Pareto dominance principle, and the better solution is used as the solution of this step.

[0036] Step 4: Determine whether the current number of iterations is greater than 3 / 4 of the total number of iterations. If so, execute step 5; otherwise, execute steps 5 and 6 in sequence.

[0037] Step 5: Use the Gray Wolf Algorithm to update the continuous solution part, recalculate the objective function, and obtain the current solution of this step; compare the solution in step 3 and the current solution of this step according to the Pareto dominance principle, and the better solution is used as the solution of this step;

[0038] Step 6: Use the Salp Algorithm to update the continuous solution part, recalculate the objective function, and obtain the current solution of this step; compare the solution in step 5 and the current solution of this step according to the Pareto dominance principle, and the better solution is used as the solution of this step;

[0039] Step 7: Update the calculation parameters of the salp algorithm;

[0040] Step 8: Randomly update the discrete solution, obtain the Pareto front and update the archive;

[0041] Step 9: Determine whether the current iteration number is greater than the total iteration number. If so, exit the iteration and obtain the archive. Otherwise, execute steps 2 to 8.

[0042] Furthermore, the solution includes:

[0043]

[0044] Furthermore, the continuous solution part is updated using the Grey Wolf Algorithm:

[0045]

[0046] in, and Respectively represent the three selected leader salps, and is based on the selected leadership salp learning, and represents the coefficient vector, and the expression is in is the parameter vector, which decreases linearly from 2 to 0 after iteration; and are two randomly generated vectors between (0, 1).

[0047] Furthermore, the continuous solution part is updated by using the salp algorithm:

[0048]

[0049] in, and are the global best salp position and the position of the current leader salp in the jth decision variable, ub j and lb j are the upper and lower bounds of the j-th decision variable, respectively.

[0050] Advantages and positive effects of the present invention:

[0051] The present invention selects unmanned aerial vehicles as aerial nodes to assist edge computing of sensor nodes in the forest. Sensor monitoring data is divided into two pieces of data of different sizes, one piece is calculated locally by the sensor and the other piece is calculated by the unmanned aerial vehicle, which reduces the computing delay of the sensor node and improves the efficiency of fire alarm, and has the advantages of flexibility, mobility, and economic efficiency. Genetic algorithms, gray wolf algorithms, and salp algorithms are used to update the continuous solution part of the model, and the optimal solution is selected according to the Pareto dominance principle. The hovering position of the unmanned aerial vehicle with the lowest flight energy consumption and the shortest delay is obtained, which further reduces the computing delay of the sensor node, while reducing the energy consumption of the unmanned aerial vehicle, increasing the endurance of the unmanned aerial vehicle, and improving the auxiliary computing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0053] Figure 1 This is a schematic diagram of fire detection by a drone in conjunction with a sensor in an embodiment of the present invention;

[0054] Figure 2 The figure is a flow chart of a method for solving a model in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of 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 of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0056] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0057] The present invention provides a forest fire monitoring method based on drone-assisted edge computing, such as Figure 1 As shown in the figure, multiple drones provide low-latency edge computing services for ground sensor nodes. In order to further improve the efficiency of edge computing, the task monitored by the sensor (monitoring fire data) can be divided into two pieces of data of different sizes, one piece is calculated locally by the sensor and the other piece is calculated by the drone. The calculation delay of the task calculated by the drone = data calculation delay + data transmission delay, while the task calculated locally by the sensor node only has data calculation delay. The larger delay between the local calculation and the drone calculation is taken as the calculation delay of the sensor node. In addition, since the drone itself has airborne energy consumption, it is necessary to minimize the flight energy consumption of the drone to improve efficiency.

[0058] Therefore, a multi-objective optimization problem is considered to jointly minimize the maximum computational delay on all sensor nodes and minimize the maximum energy consumption of the UAV. Since this problem is a hybrid multi-objective optimization problem with continuous solution space and discrete solution space, an improved multi-objective Salp algorithm is used to solve the model, which has a continuous solution update mechanism and a discrete solution update mechanism to improve the performance of the algorithm.

[0059] Specifically, the present invention constructs a dual-objective optimization problem, by adjusting the three-dimensional position of the UAV (continuous solution), the transmission power of the sensor node (continuous solution), the calculation frequency allocated by the UAV to each sensor node (continuous solution), the task size calculated by the UAV (continuous solution), and the subordinate relationship between the UAV and the sensor node (which sensor is offloaded to which UAV, discrete solution), to simultaneously minimize the maximum computing delay among all UAVs and minimize the maximum flight energy consumption among the UAVs.

[0060] S1. Model input parameters include:

[0061] The three-dimensional position X, Y, Z of the drone, the transmission power pt of the sensor node, the number of drones M, and the computing frequency f assigned by the drone to each sensor node u , the task size Ca of the local sensor node l , the task size Ca calculated by the drone u , the number of sensors K and the subordinate relationship a between the UAV and the sensor node; where the superscript U represents the UAV.

[0062] S2.1 The first goal is as follows:

[0063]

[0064] in, is the local computation latency of the jth task, The calculation delay for the UAV of the jth task. f1 is the maximum calculation delay among all sensor nodes.

[0065] S2,2The second goal is as follows:

[0066]

[0067] f2 is the maximum flight energy consumption among all drones, and i is the index of the drone. i It represents the energy consumed by the drone from the initial position to the hovering position. The expression is:

[0068]

[0069] in, Indicates the distance from the initial position to the hovering position of the drone.

[0070] S3,3Therefore, the problem is modeled as:

[0071] F = min{f1, f2}

[0072] sL min ≤X i ≤L max

[0073] L min ≤Y i ≤L max

[0074] Z min ≤Z i ≤Z max

[0075]

[0076] a ij ∈{0, 1}

[0077] TD j ≤T th

[0078] Among them, L min and L max are the lower and upper bounds of the two-dimensional region, Z min and Z max are the lower and upper bounds of the drone’s hovering height, respectively. min and p max are the lower and upper bounds of the sensor node transmission power, respectively, and f min and f max are the lower and upper bounds of the computing resources allocated to the sensor nodes by the UAV, Ca l and Ca u They represent the task size of the local sensor node and the UAV calculation, and a ij Indicates whether the sensor node offloads the computing task to the drone, T th Indicates the latency threshold.

[0079] S3, solve the model through the algorithm, combined with Figure 1 The solution method in the present invention is further described:

[0080] Step 1: Randomly initialize a population (a population consists of several groups of solutions, an example of a solution is shown below), and calculate the objective function: F = min{f1, f2}.

[0081] The solution formula is:

[0082]

[0083] Among them, X, Y, Z are the three-dimensional positions of the UAV, M represents the number of UAVs, pt is the transmission power of the UAV, and K is the number of sensors.

[0084] Step 2: For each solution, regenerate a set of UAV and sensor affiliations, recalculate the objective function, obtain a new solution and compare it with the solution of step 1, and take the better one as the solution of step 2; (whether it is better follows the Pareto dominance principle).

[0085] Step 3: Use genetic algorithm to update the continuous solution part, recalculate the objective function, obtain the current solution, and compare it with the solution of step 2. Take the better one as the solution of step 3 (whether it is better follows the Pareto dominance principle).

[0086] Step 4: Determine whether the current number of iterations is greater than 3 / 4 of the total number of iterations. If so, execute step 5; otherwise, execute steps 5 and 6 in sequence.

[0087] Step 5: Use the Gray Wolf Algorithm to update the continuous solution, recalculate the objective function, compare the current solution with the solution in Step 3, and take the better one as the solution in Step 5 (whether it is better follows the Pareto dominance principle). Specifically,

[0088]

[0089] in, and Respectively represent the three selected leader salps, and is based on the selected leadership salp learning, and represents the coefficient vector, and the expression is in is the parameter vector, which decreases linearly from 2 to 0 after iteration; and are two randomly generated vectors between (0, 1).

[0090] Step 6: Use the Salp algorithm to update the continuous solution part.

[0091]

[0092] in, and are the global best salp position and the position of the current leader salp in the jth decision variable, ub j and lb j are the upper and lower bounds of the jth decision variable, respectively. Recalculate the objective function and compare it with the solution in step 5. Take the better one as the solution in step 6 (whether it is better follows the Pareto dominance principle).

[0093] Step 7: Update the calculation parameters of the salp algorithm;

[0094] Step 8: Randomly update the discrete solution, obtain the Pareto front and update the archive;

[0095] Step 9: Determine whether the current iteration number is greater than the total iteration number. If so, exit the iteration and obtain the archive. Otherwise, execute steps 2 to 8.

[0096] The present invention constructs the hovering position of the UAV as a multi-objective optimization problem, and utilizes a continuous solution update mechanism including the crossover operation of the genetic algorithm and the update operation of the gray wolf algorithm to obtain the hovering position of the UAV with the best computational efficiency, thereby effectively monitoring forest fires.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A forest fire monitoring method based on drone-assisted edge computing, characterized in that: include: Using hovering drone swarms to assist sensor nodes in the forest to perform edge computing; The objective function is to minimize the maximum computing delay among all drones and minimize the maximum flight energy consumption among drones; The model is constructed with the two-dimensional area interval, the drone hovering height interval, the sensor node transmission power interval, the drone allocated computing resource interval to the sensor node, and the delay threshold as constraints; The continuous solution part of the model is updated using genetic algorithm, gray wolf algorithm and salp algorithm, the discrete solution part of the model is updated in a random manner, and the optimal solution is selected according to the Pareto dominance principle; the hovering position of each UAV with the lowest flight energy consumption and the smallest calculation delay is obtained.

2. The method for forest fire monitoring based on drone-assisted edge computing according to claim 1, characterized in that: The objective function: F=min{f1,f2} in, is the local computation latency of the jth task, is the computation delay of the UAV of the jth task; f1 is the maximum computation delay among all sensor nodes; f2 is the maximum flight energy consumption among all UAVs, i is the index of the UAV; E i It represents the energy consumed by the drone from its initial position to the hovering position.

3. The method for forest fire monitoring based on drone-assisted edge computing according to claim 2, characterized in that: The energy consumed by the drone from the initial position to the hovering position: in, Indicates the distance from the initial position to the hovering position of the drone.

4. The forest fire monitoring method based on drone-assisted edge computing according to claim 1 is characterized in that: The two-dimensional area interval, the drone hovering height interval, the sensor node transmission power interval, the drone allocated sensor node computing resource interval, and the delay threshold include: L min ≤X i ≤L max L min ≤Y i ≤L max WITH min ≤Z i ≤Z max TD j ≤T th Among them, L min and L max are the lower and upper bounds of the two-dimensional region, Z min and Z max are the lower and upper bounds of the drone’s hovering height, respectively. min and p max are the lower and upper bounds of the sensor node transmission power, respectively, and f min and f max are the lower and upper bounds of the computing resources allocated by the UAV to the sensor nodes; T th Indicates the latency threshold.

5. The method for forest fire monitoring based on drone-assisted edge computing according to claim 1, characterized in that: The constraints also include: a ij ∈{0,1} Among them, Ca l and Ca u They represent the task size of the local sensor node and the UAV calculation, and a ij Indicates whether the sensor node offloads the computing task to the drone.

6. The method for forest fire monitoring based on drone-assisted edge computing according to claim 1, characterized in that: The model input parameters include: The three-dimensional position X, Y, Z of the drone and the transmission power p of the sensor node t , the number of drones M, the computing frequency f assigned by the drone to each sensor node u , the task size Ca of the local sensor node l , the task size Ca calculated by the drone u , the number of sensors K and the subordinate relationship a between the UAV and the sensor nodes.

7. The method for forest fire monitoring based on drone-assisted edge computing according to claim 1, characterized in that: The method of updating the continuous solution part of the model by using a genetic algorithm, a gray wolf algorithm and a salp algorithm, updating the discrete solution part of the model by a random method and selecting the optimal solution according to the Pareto dominance principle includes: Step 1: Randomly initialize the population and calculate the objective function; a population includes several groups of solutions; Step 2: For each solution, a set of UAV and sensor affiliation relationships is regenerated, and the objective function is recalculated to obtain the current solution of this step. The solution in step 1 and the current solution of this step are compared according to the Pareto dominance principle, and the better solution is used as the solution of this step. Step 3: Use the genetic algorithm to update the continuous solution part, recalculate the objective function, and obtain the current solution of this step. Compare the solution in step 2 with the current solution of this step according to the Pareto dominance principle, and the better solution is used as the solution of this step. Step 4: Determine whether the current number of iterations is greater than 3 / 4 of the total number of iterations. If so, execute step 5; otherwise, execute steps 5 and 6 in sequence. Step 5: Use the Gray Wolf Algorithm to update the continuous solution part, recalculate the objective function, and obtain the current solution of this step; compare the solution in step 3 and the current solution of this step according to the Pareto dominance principle, and the better solution is used as the solution of this step; Step 6: Use the Salp Algorithm to update the continuous solution part, recalculate the objective function, and obtain the current solution of this step; compare the solution in step 5 and the current solution of this step according to the Pareto dominance principle, and the better solution is used as the solution of this step; Step 7: Update the calculation parameters of the salp algorithm; Step 8: Randomly update the discrete solution, obtain the Pareto front and update the archive; Step 9: Determine whether the current iteration number is greater than the total iteration number. If so, exit the iteration and obtain the archive. Otherwise, execute steps 2 to 8.

8. A method for forest fire monitoring based on drone-assisted edge computing according to claim 7, characterized in that: The solution comprises:

9. The method for forest fire monitoring based on drone-assisted edge computing according to claim 7, characterized in that: The gray wolf algorithm is used to update the continuous solution part: in, and Respectively represent the three selected leader salps, and is based on the selected leadership salp learning, and represents the coefficient vector, and the expression is in is the parameter vector, which decreases linearly from 2 to 0 after iteration; and are two randomly generated vectors between (0, 1).

10. The method for forest fire monitoring based on drone-assisted edge computing according to claim 7, characterized in that: The continuous solution part is updated using the salp algorithm: in, and are the global best salp position and the position of the current leader salp in the jth decision variable, ub j and lb j are the upper and lower bounds of the j-th decision variable, respectively.

Images