A multi-gas source positioning method based on improved teaching and learning and group optimization algorithm

Abstract

The application discloses a multi-gas source positioning method based on improved teaching and learning and group optimization algorithm, in order to make up for the deficiency of the existing multi-gas source positioning technology, the application innovatively combines multi-unmanned aerial vehicle group and improved teaching and learning and group optimization algorithm, and proposes the improved teaching and learning and group optimization algorithm.In the group method, group formation, limited group size, group merging and group dissolving operation are included, and the distance and fitness are considered when evaluating the unmanned aerial vehicle to join which group, which helps the unmanned aerial vehicle to selectively join the group when multiple signals are received at the same time; the limited group size improves the unmanned aerial vehicle distribution efficiency; the group merging is performed when the y position coordinate difference of the unmanned aerial vehicle of the global optimal position of two groups is less than the crosswind distance, so that the system performance can be improved and local redundant coverage can be avoided; the group dissolving is considered, so that the group can be prevented from staying in the found gas source all the time.The method efficiently realizes the positioning of the multi-gas pollution source in a multi-gas pollution source concentration field simulation experiment.

Description

TECHNICAL FIELD

[0001] The application relates to a multi-gas source positioning method based on an improved teaching and learning and group optimization algorithm, and belongs to the field of multi-rotor unmanned aerial vehicles and environmental monitoring. BACKGROUND

[0002] At present, China is in a period of rapid economic development, and the industrialization process is constantly advancing. The amount of industrial waste gas emission is increasing, and air pollution problems are becoming increasingly serious. Therefore, it is necessary to strengthen the monitoring and treatment of air pollution. At present, the main body for studying the positioning of air pollution sources in China is mostly ground mobile robots or fixed monitoring stations. However, these methods have certain limitations: ground mobile robots are prone to failure and difficult to work continuously, and have poor flexibility; and monitoring stations are not evenly distributed and are not necessarily located near accident pollution sources, so the applicability of monitoring stations is low and they lack mobility and flexibility. At present, unmanned aerial vehicles have been widely used in atmospheric monitoring. Compared with ground mobile robots, unmanned aerial vehicles are more flexible to operate and can save detection costs; compared with fixed monitoring stations, unmanned aerial vehicles can reduce the risk of human monitoring and improve the accuracy of measurement. Unmanned aerial vehicles have strong applicability, accurate positioning, accurate measurement and convenient maintenance. Using unmanned aerial vehicles to locate air pollution sources can ensure accurate and accurate collection results, and has superior performance and higher automation, which can realize the monitoring and tracing of air pollutants in a large area.

[0003] At present, most gas source positioning researches are aimed at single point gas sources, such as the "Multi-UAV Atmospheric Pollution Source Positioning Method Based on Improved Teaching and Learning Algorithm" with the patent number CN113640470A, which is based on the improvement of swarm intelligence algorithm to locate single gas source. However, due to insufficient environmental regulation, some industrial parks still have the phenomenon of simultaneous direct discharge or illegal discharge by multiple enterprises, and toxic / hazardous gas leaks caused by factors such as aging of containers or pipelines and failure of landfill sealing layer may be multiple, i.e. there may be point or planar odor sources releasing the same gas (component) in number and location unknown in the search area. Obviously, to make this research more practically valuable, it is not enough to only locate single point gas sources, and it is necessary to conduct positioning research on multiple gas sources. SUMMARY

[0004] The purpose of the present application is to make up for the shortcomings of existing multi-gas source positioning technology. The present application innovatively combines multi-unmanned aerial vehicles with an improved teaching and learning algorithm, and proposes a multi-gas source positioning method based on an improved teaching and learning and group optimization algorithm. This method has the characteristics of high positioning accuracy, strong convergence and avoidance of local optimum, and can quickly and efficiently realize the positioning of multiple gas pollution sources.

[0005] The present application considers a system consisting of n unmanned aerial vehicles R i (i=1...n) equipped with sensors for measuring plume concentration and wind direction and moving in an unobstructed space. It is assumed that the unmanned aerial vehicles have the ability of odor perception, localization, communication and movement within the search space. There are m odor sources S j (j=1...m) placed at random locations in the search space. These gas sources diffuse odor plumes by releasing some chemicals. A multi-gas source localization method based on improved teach and learn and group optimization algorithm is used to quickly track the team consisting of n unmanned aerial vehicles to the m gas sources.

[0006] A multi-gas source localization method based on improved teach and learn and group optimization algorithm, the flow as Figure 1 shown, comprising the following steps:

[0007] Step 1: set n unmanned aerial vehicles R i (i=1...n) in the area to be monitored, there are m odor sources S j (j=1...m) in the area to be monitored;

[0008] Step 2: initialization setting, that is, setting the initial position of the unmanned aerial vehicle;

[0009] Step 3: search for multiple gas pollution sources using improved teach and learn and group optimization algorithm;

[0010] Step 4: output all gas pollution source positions.

[0011] The improved teach and learn and group optimization algorithm proposed in the step 3, the flow as Figure 2 shown, comprising the following steps:

[0012] Step 1: initialize parameter setting, threshold Th r located near the plume, threshold Th g for successful gas source localization;

[0013] Step 2: calculate the fitness value of the initial unmanned aerial vehicle group individual R i ;

[0014] Step 3: determine whether the fitness value of the unmanned aerial vehicle R i exceeds the threshold Th r , if so, execute step 4, otherwise, execute step 5;

[0015] Step 4: use group method for multi-unmanned aerial vehicle cooperative search;

[0016] Step 5: the unmanned aerial vehicle performs random search;

[0017] Step 6: determine whether the unmanned aerial vehicle R iwhether the fitness value exceeds a threshold Th g If yes, go to step 7, otherwise, go back to step 2.

[0018] Step 7: find a gas source, and then determine whether the UAV R i belongs to a group, if yes, disperse the group, otherwise, determine that the UAV R i is not assigned to a group.

[0019] Step 8: determine whether all gas sources are successfully located, if yes, end, otherwise, go back to step 2.

[0020] The fitness value calculation formula in step 2 of the improved teaching and learning and group optimization algorithm is:

[0021] fit i = C i

[0022] Wherein, C i is the measured gas concentration value of the UAV R i .

[0023] The group method proposed in step 4 of the improved teaching and learning and group optimization algorithm, the flow is shown as follows, including the following steps: Figure 3

[0024] Step 1: form a group by using a group formation method;

[0025] Further, the group formation method is:

[0026] One or more UAVs can enter the plume vicinity area at the same time, and the UAVs in the plume vicinity area start to send signals, which are received by other UAVs within the radius G r of the signal generating UAV, and the UAV can receive multiple signals at the same time. In this case, the UAV compares the signal strengths it receives, and joins the group whose signal it receives the strongest. The signal strength depends on the fitness value and the distance of the UAV, and the calculation formula is as follows:

[0027] V ij = fit i ·exp(-d ij )

[0028] Where V ij is the signal generated by the UAV R i within the radius G r and received by the UAV R j , fit i is the fitness value of the UAV R i , and d ij is the distance between the UAV R i and R j ​The distance between the UAVs and the signal strength equation set are used to consider the distance and fitness at the same time, which helps the UAVs to decide which group to join when multiple signals are received at the same time.

[0029] Step 2: Determine whether the size of the group is greater than G maxsize If it is greater, the group size limiting operation needs to be performed before Step 3 is executed, otherwise Step 3 is directly executed.

[0030] Further, the parameter G maxsize is defined to limit the maximum number of UAVs in the group, which helps to achieve better resource utilization. As the number of UAVs increases, the size of the group also increases, but the increase in the number of gas sources will lead to the decrease in the size of the group, so the number of UAVs needed to find additional gas sources should also increase. Considering this, a function is defined to limit the size of the group, and the function is as follows:

[0031]

[0032] w m = |0.5 - (1 / (exp(m)))|

[0033] where n is the total number of UAVs, m is the number of gas sources, and λ is a constant. When a group is formed, if the size of the group formed is greater than G maxsize , P worst-performing UAVs will be removed from the group, and the calculation formula of P is as follows:

[0034] P = gk size -G maxsize

[0035] where gk size is the size of the group gk, and P is the number of UAVs removed from the group gk;

[0036] Step 3: Update the positions of the UAVs in the group using the improved teach-and-learn algorithm;

[0037] Further, the improved teach-and-learn algorithm proposed in Step 3 has a flow as shown in Figure 4 , which includes the following steps:

[0038] Step 1: Initialization setting, i.e. setting the population size N = 6, the population dimension D = 2, and the maximum number of iterations maxgen = 500. The set of all UAVs in the search space is called a class, and a UAV in the class is called a student;

[0039] Step 2: Only in the first iteration, randomly set the step size d0, and the UAV swarm flies d0 using the concentration gradient method, and the individual ability value of each UAV is calculated After that, the individual ability value of the UAV is calculated directly according to the individual ability value model calculation formula before each iteration

[0040] The individual ability value model calculation formula is:

[0041]

[0042] wherein, is the remaining power of the ith unmanned aerial vehicle in the jth iteration, U max is the full power of the unmanned aerial vehicle, is the stability of the flight height of the ith unmanned aerial vehicle in the jth iteration, f i j is the odor pollution concentration value searched by the ith unmanned aerial vehicle in the jth iteration, is the minimum value of the pollution gas concentration searched by the N unmanned aerial vehicles in the jth iteration, is the maximum value of the pollution gas concentration searched by the N unmanned aerial vehicles in the jth iteration, β1, β2, and β3 are weight coefficients and β1+β2+β3=1.

[0043]

[0044] wherein is the height of the jth iteration path node q, Q is the number of nodes on the flight path of the ith unmanned aerial vehicle, h j is the average value of the flight height in the jth iteration.

[0045] Step 3: updating the positions and the pollution gas concentration values searched by the students according to the “teaching” stage updating rule in the improved teaching and learning algorithm;

[0046] Step 4: updating the positions and the pollution gas concentration values searched by the students according to the “learning” stage updating rule in the improved teaching and learning algorithm;

[0047] Step 5: determining whether the iteration number is reached, if the iteration number is reached, going to Step 6, otherwise returning to Step 2;

[0048] Step 6: outputting the algorithm optimization result, and ending the algorithm.

[0049] The “teaching” stage updating rule in the improved teaching and learning algorithm in Step 3 is that: first, selecting the top three students called “Teacher A”, “Teacher B”, and “Teacher C” according to the individual ability value ranking from large to small calculated in Step 2, and each student learns according to the difference between the teacher and the student average value.

[0050] The following formula is used to realize the “teaching” stage:

[0051]

[0052]

[0053] wherein, and are the values of the i-th student before and after learning, respectively, is the individual ability value of the i-th student, X A ,X B ,X C is the position of "Teacher A", "Teacher B", "Teacher C", X ABC is the barycentric position of the three teachers, teaching factor t is the current iteration number, is the average value of all students;

[0054] After the "teaching" is completed, the students are updated, and each student is compared according to the learning results after learning and the learning results before learning:

[0055] If

[0056]

[0057] End

[0058] The updating rule in the "learning" stage of the improved teaching and learning algorithm in step 4 is that the "learning" stage is divided into classes, and when there are two or more students searching for odor pollution concentration values exceeding the threshold f0, the students in the intensive class learn according to the intensive class student learning rule, and the students not exceeding the threshold f0 learn according to the ordinary class student learning rule;

[0059] The intensive class student learning rule is:

[0060] When f(X mean ) > f(X i ),

[0061] When f(X i ) > f(X mean ),

[0062] wherein X mean is the barycentric position of the intensive class students, m is the number of intensive class students;

[0063] The ordinary class student learning rule is: for each student X i (i = 1, 2,..., N), a student X k (k = 1, 2,..., N, i ≠ k) is randomly selected in the class, and X i learns by analyzing the differences between itself and the student X k :

[0064] When f(X k ) > f(X i ),

[0065] When f(X i ) > f(X k ),

[0066] Step 4: Calculate whether the crosswind distance between the two groups is less than the sum of their radii, if less, then combine and operate before output, otherwise directly output;

[0067] Further, the combination and operation method is:

[0068] During plume tracking, multiple UAV groups may search for the same odor source, resulting in performance degradation and local redundant coverage. To avoid this problem, the present application uses a combination and operation method, which measures the crosswind distance C d between the best performing UAVs in the two UAV groups, i.e. the distance in the y direction between the best performing UAVs in the two UAV groups. If the crosswind distance between the two groups is less than the sum of their radii, then the two groups are combined, and the mathematical expression is as follows:

[0069] C d (Y gk -Y gl )<(R gk +R gl )

[0070] Where C d is the crosswind distance between the best performing UAVs in group gk and group gl, Y gk and Y gl are the y position coordinates of the globally best performing UAVs in group gk and group gl, respectively, and R gk and R gl are the radii of group gk and group gl, respectively.

[0071] The group dissolving method in step 7 of the improved teach and learn and group optimization algorithm is: if a group finds a gas source, then the group is dissolved, and the members of the group are distributed in other groups. During the distribution process, it is important to maintain a balance between the number of UAVs in the plume discovery and plume tracking phases. To this end, a threshold Th p is defined and given by the following equation:

[0072]

[0073] Where β is a constant, Population size is the size of the group; compare the number of UAVs in other groups with the threshold Th p , if the number is less than the threshold Thp , the UAVs join other groups according to the benefit of each group. The benefit of a group is defined according to the need of the group and the cost of joining the group. The benefit of a group is:

[0074] U gk = Need gk - Cost

[0075] Need gk = exp(-(β4)·(fit gk +(β5)·(N m ))

[0076] Cost = d(|X gkmin - X j |)

[0077]

[0078] Where, Need gk is the need of the group, Cost is the cost of joining the group, N m is the number of members in the group, β4 and β5 are weight coefficients, d is the Euclidean distance function, X gkmin is the position of the UAV in the group gk closest to the position of the UAV R j belonging to the dissolved group, X j is the position of the UAV R j , i.e. belonging to the dissolved group, fit gk is the average fitness value of the group gk, H is the number of UAVs in the group gk, is the fitness value of the UAV R h in the group gk. Each unassigned UAV in the dissolved group calculates the benefit of each group and joins the group with the highest benefit. The above calculation procedure is repeated until there is no unassigned UAV in the dissolved group.

[0079] The present application has the following advantages:

[0080] The present application innovatively combines the multi-UAV group method, the improved teaching and learning algorithm, and proposes a multi-gas source positioning method based on the improved teaching and learning and group optimization algorithm. The method has the characteristics of high positioning accuracy, strong convergence, and can avoid falling into local optimum, and can quickly and efficiently realize the positioning of multiple gas pollution sources.

[0081] The group method includes group formation, limiting group size, group merging, and group dissolution operations. Considering the group formation operation can make the UAV group form various groups, divide the work and cooperate to search, and improve the search efficiency. Moreover, when evaluating which group the UAVs join, the distance and fitness are considered, which helps the UAVs to selectively join the group when receiving multiple signals at the same time.

[0082] The reason for considering limiting the group size is that as the number of UAVs increases, the group size also increases, but the increase in the number of gas sources will lead to a decrease in the group size, because the number of UAVs required to simultaneously search for additional gas sources should also increase. Therefore, taking this into account, a function for limiting the group size is defined.

[0083] The reason for considering group merging is that during plume tracking, multiple UAV groups may travel towards the same odor source, so a method for measuring the crosswind distance is proposed, and when the y-position coordinate difference between the globally optimal positions of the UAVs of two groups is less than the sum of their radii, the two groups are merged, which can improve system performance and avoid local redundant coverage.

[0084] The reason for considering group dissolution is that when a group finds a gas source, the UAVs in this group are to be dissolved and distributed in other groups to search for the next gas source, which helps to improve the system search efficiency and avoid the group staying at the found gas source. BRIEF DESCRIPTION OF DRAWINGS

[0085] Figure 1 is a whole flow chart of multi-gas pollution source positioning

[0086] Figure 2 is a flow chart of improved teach and learn and group optimization algorithm searching for gas pollution sources

[0087] Figure 3 is a flow chart of using group method searching

[0088] Figure 4 is a flow chart of improved teach and learn algorithm

[0089] Figure 5 is a merging diagram

[0090] Figure 6 is a multi-gas source positioning process DETAILED DESCRIPTION

[0091] The specific implementation of the present application is as follows:

[0092] A multi-gas source positioning method based on an improved teach and learn and group optimization algorithm, comprising the following steps:

[0093] Step 1: Set 12 UAVs in the area to be monitored;

[0094] Step 2: initialization setting, i.e. setting the initial position of the UAV, located in the threshold value Th r , the threshold value Th g of gas source positioning success;

[0095] Step 3: search for multi-gas pollution sources using an improved teach and learn and group optimization algorithm;

[0096] Step 4: Output the locations of all gaseous pollution sources.

[0097] like Figure 5 As shown, during plume tracking, multiple drone groups may search towards the same odor source, leading to performance degradation and localized redundant coverage. To avoid this problem, this invention uses a combination method, employing crosswind distance C. d The measurement method involves the distance in the y-direction between the best-performing drones in two drone groups. If the crosswind distance between the two groups is less than the sum of their radii, the two groups are merged. The mathematical expression for this is as follows:

[0098] C d (Y gk -Y gl ) < (R gk +R gl )

[0099] like Figure 6 As shown, in the multi-gas pollution source concentration field simulation experiment, at t=0s, 12 UAVs are initialized, and the improved teaching and learning and group optimization algorithms are used to search for multiple gas pollution sources; at t=2015s, each UAV gradually approaches the gas pollution source under the guidance of the group method; at t=5254s, a group has found the first gas pollution source and performed a group disbanding operation, assigning the disbanded UAVs to other groups to continue the search for gas pollution sources; at t=7295s, all four gas pollution sources are successfully located, each UAV stays near the gas pollution source, and the positions of all gas pollution sources are output, ending the gas pollution source localization process.

[0100] The above are preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions made by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-gas source localization method based on improved teach and learn and group optimization algorithm, characterized in that, The method comprises the following steps: Step 1: Set n drones R in the area to be monitored i (i = 1...n), there are m odor sources S in the area to be monitored j (j = 1...m); Step 2: initialization setting, that is, setting the initial position of the unmanned aerial vehicle; Step 3: searching for multiple gas pollution sources by using the improved teach and learn and group optimization algorithm; Step 4: outputting the positions of all gas pollution sources; The improved teach and learn and group optimization algorithm in step 3 comprises the following steps: Step 1: initialization of parameter settings, threshold value Th located near the plume r , threshold value Th for successful positioning of the gas source g ; Step 2: Calculate initial drone swarm individual R i of the fitness value; Step 3: Determine if the drone R i whether the fitness value exceeds a threshold Th r If so, proceed to Step 4, otherwise proceed to Step 5. Step 4: searching for multiple unmanned aerial vehicles by using the group method; Step 5: randomly searching for the unmanned aerial vehicle; Step 6: Determine if the drone R i whether the fitness value exceeds a threshold Th g If so, proceed to Step 7, otherwise, return to Step 2. Step 7: find a gas source, and then determine the UAV R i whether it belongs to the group, if it belongs to the group, the group is dissolved, otherwise determine the UAV R i not assigned to a group; Step 8: judging whether all gas sources are successfully positioned, if yes, ending, otherwise, returning to step 2; The multi-gas source positioning method based on the improved teaching and learning and group optimization algorithm further has the characteristics that the group is dissolved in step 7, if a group finds a gas source, the group is dissolved, and the members of the group are distributed in other groups. During the distribution process, it is important to maintain a balance between the number of unmanned aerial vehicles in the plume discovery and plume tracking stages. Therefore, a threshold Th is defined and given by the following equation: p Th = (N - 1) / N wherein β is a constant, Population size is the size of the group, the number of drones of other groups is compared with the threshold value Th p , if the number is less than the threshold value Th p , the drones join other groups according to the benefit of each group, the benefit of the group is defined according to the demand of the group and the cost of joining the group, and the benefit of the group is: U gk = Need gk -Cost Need gk = exp(-(β4) · (fit gk )+(β5) · (N m )) Cost = d(|X gkmin - X j |) where Need gk is the requirement of the group, Cost is the cost of joining the group, N m is the number of members in the group, β4 and β5 are weight coefficients, d is the Euclidean distance function, X gkmin is the position of the closest unmanned aerial vehicle in the group gk to the unmanned aerial vehicle R j that is to be dissolved, X j is the position of the unmanned aerial vehicle R j that is to belong to the dissolved group, fit gk is the average fitness value of the group gk, H is the number of unmanned aerial vehicles of the group gk, is the fitness value of the unmanned aerial vehicle R h in the group gk; The unmanned aerial vehicle not allocated in the group calculates the benefit of each group, and is added to the group with the highest benefit; the above calculation procedure is repeated until there is no unmanned aerial vehicle not allocated in the group.

2. A multi-gas source localization method based on improved teaching and learning and group optimization algorithm according to claim 1, characterized in that The group method in step 4 comprises the following steps: Step 1: forming a group by using a group forming method; Step 2: Determine if the size of the group is greater than G maxsize If yes, then limit the group size and perform Step 3. Otherwise, perform Step 3 directly. Step 3: updating the position of the unmanned aerial vehicle in the group by using the improved teach and learn algorithm; Step 4: calculating whether the crosswind distance between two groups is less than the sum of the radii, if yes, the combination operation needs to be performed and then output, otherwise, directly output.

3. A multi-gas source localization method based on improved teaching and learning and group optimization algorithm as claimed in claim 1, wherein The group forming in step 1 comprises the following steps: The unmanned aerial vehicle compares the signal strength received, and is added to the group with the strongest signal received, and the signal strength depends on the fitness value and distance of the unmanned aerial vehicle: V ij = fit i · exp(-d ij ) Where V ij It is made by drone R i That is, within radius G r Internally generated and by drone R j The received signal, fit i It is a drone R i fitness value, d ij It is a drone R i and R j The distance between signals is determined using a set of signal strength equations that take into account both distance and fitness, helping the drone decide which group to join when receiving multiple signals simultaneously.

4. The method of claim 1, wherein The limit group in step 2, this operation defines the parameter G maxsize To limit the maximum number of drones in the group, which helps to achieve better resource utilization, as the number of drones increases, the size of the group also increases, but the increase in the number of gas sources will lead to the decrease of the group size, so the number of drones needed to find additional gas sources should also increase, so the function to limit the size of the group is defined, the function is as follows: w m = |0.5 - (1 / (exp(m)))| Wherein, n is the total number of UAVs, m is the number of gas sources, λ is a constant, when a group is formed, if the size of the formed group is greater than G maxsize , P worst-performing UAVs will be removed from the group, and the calculation formula of P is as follows: P = gk size - G maxsize where gk size is the size of group gk, P is the number of drones removed from group gk.

5. The method of claim 1, wherein The combination of the two groups in Step 4 is measured by the crosswind distance C d between the two drones that perform best in each group in the y direction, and if the crosswind distance between the two groups is less than the sum of their radii, the two groups are combined: C d (Y gk -Y gl )<(R gk +R gl ) wherein C d is the crosswind distance between the best performing drones of group gk and group gl, Y gk and Y gl are the y position coordinates of the globally best drones belonging to group gk and group gl, respectively, R gk and R gl are the radii of group gk and group gl, respectively.

Images