CN108168558A - Unmanned aerial vehicle flight path planning algorithm applied to river target search task - Google Patents

Abstract

Applied to the UAV track planning algorithm for river target search tasks, the present invention abstracts the target river area into a two-dimensional curve, and uses a Gaussian mixture model to extract and search when the probability of the target in each curve segment is known. High-value curve segments, that is, high-value search areas, and sort the search order of high-value curve segments, and assign them to UAVs performing search tasks, so that UAVs can search for targets according to the sorted curve segments. This method can obtain near-optimal results and has a fast solution speed, which improves the target search efficiency of UAVs.

Description

UAV track planning algorithm applied to river target search task

technical field

The invention relates to the technical field of unmanned aerial vehicle navigation and control, and relates to an unmanned aerial vehicle track planning algorithm, in particular to an unmanned aerial vehicle track planning algorithm applied to river target search tasks.

Background technique

Unmanned Aerial Vehicle (UAV) refers to a type of aircraft that is driven by power, unmanned on board, controlled by radio remote control or controlled by its own program. As a product of the high integration of aviation technology and information technology, unmanned aerial vehicles (UAVs) have been widely used in military and civilian fields because of their advantages such as high cost performance, flexible use, high-risk tasks, and no restrictions on the pilot's physiological conditions. Over the past 30 years, countries around the world have continued to pay attention to and increase investment in the field of drones. UAV technology has made great progress and progress, representing the development direction of today's high-tech.

As one of the typical applications of unmanned aerial vehicles, target search has been widely used in scenarios such as emergency monitoring or rescue. Since the survival probability of the searched target (such as lost people, shipwreck, etc.) will decrease rapidly with the passage of time, it is required that the UAV find the target in the shortest possible time. The above problem can be regarded as a trajectory optimization problem in essence, so that the UAV can obtain the maximum search rate of return (ie, the cumulative probability) when flying along the optimal trajectory, and realize efficient reconnaissance coverage of the mission area and rapid target search.

Scholars at home and abroad have done a lot of research on the problem of target search and proposed a series of methods, mainly including geometric method, random search method, method based on search graph, etc. The geometric method realizes the traversal or full coverage of the mission area by the UAV by planning a search trajectory of a specific shape, such as parallel lines, spiral lines, etc. Although the principle of this type of method is simple, its search Significantly less efficient. The random search method guides the UAV to move randomly in the mission area, so as to gradually cover the area and search for the target. The biggest advantage of this type of method is that it does not require precise positioning and complex optimization decision-making process, but it also does not use the prior information of the target. And it is not suitable for large-scale complex areas. The method based on the search graph first discretizes the task area into a series of grid units, and then adopts an appropriate optimization strategy based on the target information stored in each unit to make the UAV move in the most promising direction, although this type of method can be flexible Handle all kinds of complex situations, but drones may linger in local areas for a long time and ignore other high-value areas, and the search efficiency needs to be improved. In addition, in order to solve the above local optimal problem and simplify the multi-UAV cooperative target search task, an effective idea is to decompose the task area into multiple sub-regions and assign them to each UAV, so as to transform the complex cooperative control problem into For multiple simple single-UAV search problems, the main region decomposition methods include centroid Voronoi segmentation sampling, fuzzy C-means clustering, polygon segmentation, etc.

Most of the existing research on target search is applied to two-dimensional horizontal areas (such as regular or irregular areas such as rectangles), but there are few researches on the problem of target search in river areas. As a special task area, rivers can be regarded as curves with certain terrain constraints. Therefore, compared with the multi-course or even full-course selection of UAVs in ordinary two-dimensional horizontal areas, river areas limit the flight of UAVs. mode, which increases the difficulty of the target search problem. For this type of problem, the existing solutions are mostly qualitative passive strategies, such as drones flying over rivers for full-coverage searches, or directly rushing to the previous location of the target for greedy searches, lacking quantitative analysis and heuristic strategies guide.

Contents of the invention

The purpose of the present invention is to propose a UAV track planning algorithm suitable for river target search tasks in combination with the natural conditions and characteristics of rivers, so as to realize the rapid search of river target tasks.

In order to achieve the above object, the present invention provides the following technical solutions:

The unmanned aerial vehicle track planning algorithm applied to the river target search task is characterized in that the field of view of the unmanned aerial vehicle is greater than the width of the river, and the planning algorithm includes the following steps:

S1: abstract the target river to be searched into a two-dimensional plane curve, take the forward distance L _s as an independent variable, and discretize the plane curve into M=L/L _s discretization units, where L is the total length of the river , L _s is the length of each discretization unit;

S2: In any sampling period, the UAV moves within two discretization units s _m ; the probability p(s _m ) of the target to be searched in each discretization unit s _m is known, and p(s _m )∈[0,1], and satisfy

S3: Any number of discretized units compose the search curve segment, S _k = {s _m ,s _m+1 ,...,s _n }, 1≤m<n≤M, use the Gaussian mixture model to describe the target to be searched in each The probability information existing in each search curve segment, extract the high-value search curve segment S _k with P _k,1 and P _k,2 as the boundary of the region;

S4: Define the sub-area transition time as the shortest flight time for the boundary point P _{i, p} of the current search curve segment to go to the boundary point P _{j, q} of the next search curve segment, and define the transition time as:

T(P _i,p ,P _j,q )=λ ₁ ·Dubins_cost(P _i,p ,P _j,q ), where i≠j and i,j∈{1,2,...K}, and p,q∈{1,2}; where λ ₁ represents the proportional coefficient of sub-region transition time; Dubins_cost(P _i,p ,P _j,q ) represents the length of Dubins curve, p represents the i-th search curve segment Flying point flag, q represents the entry point flag of the jth search curve segment, K represents the number of high-value search curve segments;

Define the sub-area coverage time as the time when the UAV searches for the flight in the search curve segment, which is C _k = λ ₂ L _k , where λ ₂ represents the proportional coefficient of the sub-region coverage time, and L _k represents the length of the search curve segment, L _k = (n-m+1)·L _s ;

Define the sub-area coverage return as the cumulative probability obtained by the drone completely covering the high-value search curve segment: R _k ;

S5: Using the above-mentioned sub-area transition time, sub-area coverage time and sub-area coverage return as evaluation indicators, adopt the nearest insertion method to iteratively sort the search sequence of each high-value search curve segment, and the sorting model is:

S6: Connect the optimized high-value search curve segments in sequence to obtain the final search path of the UAV.

Preferably: the method for extracting high-value search curve segments includes the following steps:

Using K one-dimensional Gaussian functions Form a Gaussian mixture model, where k=1,2,...K, u _k is the mean value of each Gaussian function, and σ _k is the standard deviation of each Gaussian function;

Let the weight coefficient of each Gaussian function be α _k , and satisfy

The target probability of the curve segment to be searched is

Estimate the weight coefficient α _k , mean value u _k , and standard deviation σ _k of each Gaussian function, and iteratively estimate p(s) until the convergence condition is met;

The corresponding intervals of 95.4% probability of each Gaussian function are obtained, and P _k,1 =u _k -2σ _k and P _k,2 =u _k +2σ _k are the high-value search curve segments of the area boundary points, then L _k =4σ _k ; R _k =0.954α _k .

As a preference: the convergence condition is |p(s)-p'(s)|<ξ, where p(s) and p'(s) are target probability values before and after iteration respectively, and ξ=10 ^-5 .

As a preference: if multiple unmanned aerial vehicles perform the search task, the following steps are further included: assigning the sorted high-value search curve segments to multiple unmanned aerial vehicles, and performing area allocation.

As preferred: the method for area allocation, comprising the following steps:

Assign a high-value search curve segment set to each unmanned aerial vehicle, obtain a set for each unmanned aerial vehicle and assign a high-value search curve segment set for A _i ;

According to the method in step S5, perform index calculation on the high-value search curve segment in each UAV set A _i to obtain J _i ;

Model the region assignment as Among them, _Nu represents the total number of UAVs, _ρ1 represents the proportional coefficient of the total search revenue, and _ρ2 represents the proportional coefficient of the task balance degree;

The iterative method is used to solve the modeling of area allocation and determine the area allocation strategy.

As a preference: the set of high-value search curve segments allocated by all drones satisfies the following constraints:

The beneficial effects of the present invention are:

(1) The present invention abstracts the target river area into a two-dimensional curve, which greatly reduces the complexity of the problem and the amount of calculation;

(2) The present invention utilizes the self-adaptive Gaussian mixture model to approximately describe the characteristics of the river region and extract high-value sub-regions, quantify and extract the corresponding interval of 95.4% probability of the Gaussian function, and assign search tasks to high-value sub-regions in a targeted manner. The accuracy of the quantitative results is high and it is conducive to the solution of subsequent problems;

(3) The present invention uses the nearest insertion method to sort regions, which can obtain approximately optimal results and has a faster solution speed;

(4) The algorithm provided by the invention is applicable to the planning of one or more unmanned aerial vehicle search paths.

Description of drawings

Fig. 1 is method flowchart of the present invention;

Figure 2 is a schematic diagram of abstracting the river into a two-dimensional curve;

Figure 3 is the target probability map in the river area;

Figure 4 is the approximate result of the target probability based on the adaptive Gaussian mixture model;

Fig. 5 is the iterative process of subregion sorting based on the nearest insertion method;

Fig. 6 is the unmanned aerial vehicle trajectory optimized by the method of the present invention with three unmanned aerial vehicles performing target search.

Detailed ways

The specific implementation manners of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Apparently, the embodiments described in the specific implementation are only some of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

The invention provides an algorithm for planning the search track of an unmanned aerial vehicle, specifically an algorithm for planning the unmanned aerial vehicle's track when the unmanned aerial vehicle is used for a river target search task. Using this planning algorithm can greatly improve the search efficiency of UAVs.

In order to improve the search efficiency of the UAV, the sight of the visual sensor on the UAV is always pointing directly below, and its field of view is larger than the width of the river. In this way, it can be guaranteed that the UAV can search The extent of the entire channel width. Refer to Figure 1 for the overall flowchart of the method.

This embodiment firstly provides a search track planning algorithm suitable for a UAV performing a search mission.

The UAV track planning algorithm applied to the river target search task includes the following steps:

S1: Abstract the target river into a two-dimensional plane curve and discretize it.

Referring to Figure 2, since the width of the UAV's search viewing angle is greater than the width of the river, the width of the river can be ignored, and the target river to be searched is abstracted into a two-dimensional plane curve. Taking the forward distance L _s as an independent variable, discretize the plane curve into M=L/L _s discretization units, where L is the total length of the target river, and L _s is the length of each discretization unit; After processing, the discretization units of M rivers will be obtained.

S2: Construct the probability information of each discretized unit within the unit sampling period.

In the unit sampling period, the search range of the UAV is the length of the discretization unit of the above-mentioned river. Specifically, in any sampling period, the UAV moves within two discretization units s _m ; the probability p(s _m ) of the target to be searched in each discretization unit s _m is known, and p( s _m )∈[0,1], and satisfy Refer to Figure 3 for the probability plot.

S3: Extract high-value search curve segments.

Since the probability of the target existing in each search curve segment is different, some are high and some are low, in order to improve the search efficiency, it is necessary to select the curve segment with a high target existence probability for searching, therefore, it is necessary to extract high-value search curve segments.

Any number of discretized units compose the search curve segment, S _k ={s _m ,s _m+1 ,...,s _n }, 1≤m<n≤M, use the Gaussian mixture model to describe the target to be searched in each search The probability information existing in the curve segment, extract the high-value search curve segment S _k with P _k,1 and P _k,2 as the boundary of the region; the number of high-value search curve segments is K, k∈{1,2,. .., K}.

The specific method is: using K one-dimensional Gaussian functions:

Form a Gaussian mixture model, where k=1,2,...K, u _k is the mean value of each Gaussian function, and σ _k is the standard deviation of each Gaussian function;

Let the weight coefficient of each Gaussian function be α _k , and satisfy

The target probability of the curve segment to be searched is

Estimate the weight coefficient α _k , mean value u _k , and standard deviation σ _k of each Gaussian function, define a training sample composed of D unit positions, and use the maximum expectation method to iteratively estimate p(s) until the convergence condition is met; The convergence condition is p(s)-p'(s)<ξ, where p(s) and p'(s) are the target probability values before and after iteration respectively, and the value of ξ depends on the specific convergence accuracy requirements. In the examples, ξ=10 ^-5 . Iteration results refer to Figure 4.

In the above parameter iterative estimation process, the proportion of the number of training individuals corresponding to each unit position is consistent with the target probability, that is, D _m = p(s _m )D; dynamic mechanisms such as elimination, merging and splitting can be introduced to adaptively adjust the number of models . If the weight of a Gaussian model is very small and there is a certain distance from other models, it means that the component is an redundant noise component, which can be eliminated directly; if the distance between two Gaussian models is very close and the weight is not large, they are considered to reflect the same feature distribution. Overfitting occurs, so they can be combined into one Gaussian component; if the weight and standard deviation of a certain Gaussian component are large, it indicates that underfitting occurs, and it needs to be split into two Gaussian components.

According to the parameter estimation results, the corresponding intervals of 95.4% probability of each Gaussian function can be quantified and extracted as high-value river sub-regions (that is, high-value search curve segments), and P _k,1 = u _k -2σ _k and P _k,2 =u _k +2σ _k is the high-value curve segment of the area boundary point, then the length of the high-value search curve segment L _k =4σ _k .

S4: Define the quantitative evaluation index, use the nearest insertion method to evaluate and sort the high-value search curve segments, and determine the search order of the high-value search curve segments by the UAV. This embodiment defines the following three evaluation indicators: sub-area transition time, sub-area coverage time, and sub-area coverage return.

Define the sub-region transition time as the shortest flight time from the boundary point P _i,p of the current search curve segment (or the search initial position P _initial ) to the boundary point P _j,q of the next search curve segment. High-value search curve segments are relatively separated curve segments, and the boundary points of each curve segment are discontinuous points, and the sub-region transition time represents the time for moving between mutually discontinuous curve segments. Usually the Dubins curve is the shortest route of the UAV between any two points considering the heading constraint, and the transition time is defined as:

T(P _i,p ,P _j,q )=λ ₁ ·Dubins_cost(P _i,p ,P _j,q ), where i≠j and i,j∈{1,2,...K}, and p,q∈{1,2}; where λ ₁ represents the proportional coefficient of sub-region transition time; Dubins_cost(P _i,p ,P _j,q ) represents the length of Dubins curve, p represents the i-th search curve segment Flying point flag, q represents the entry point flag of the jth search curve segment, K represents the number of high-value search curve segments;

Define the sub-area coverage time as the time when the UAV searches for the flight in the search curve segment, which is C _k = λ ₂ L _k , where λ ₂ represents the proportional coefficient of the sub-region coverage time, and L _k represents the length of the search curve segment, L _k =(n-m+1)·L _s ; since the length of the high-value curve segment extracted by the Gaussian model is L _k =4σ _k , therefore, C _k =λ ₂ ·4σ _k .

The relationship between λ ₁ and λ ₂ is inversely proportional to the relationship between the flight speed of the UAV in different stages, for example, the flight speed of the UAV in the transition between sub-areas is the flight speed when it covers the sub-area (that is, over the sub-area) 2 times the speed, the above parameters should satisfy

Define the sub-area coverage return as the cumulative probability R _k obtained by the UAV completely covering the high-value search curve segment, which is the sum of the target existence probability of each high-value search curve segment; and compared to the high-value curve extracted by the Gaussian model Since the river sub-region is the range corresponding to the 95.4% probability of each Gaussian function, and the weight coefficient is α _k , the sub-region coverage return is defined as follows: R _k =0.954α _k .

S5: Sort the search order of each high-value search curve segment to achieve an optimal and efficient search strategy.

Using the above-mentioned sub-area transition time, sub-area coverage time and sub-area coverage return as evaluation indicators, the nearest insertion method is used to iteratively sort the search order of each high-value search curve segment _Lp , and the sorting model is:

Among them, P _initial is the starting position of the search, Respectively represent the two endpoints of the first high-value curve segment to start searching, represents the subregion coverage return for the first high-value curve segment to start searching, Indicates the subregion coverage time for the first high-value curve segment to start searching. corresponding, Respectively represent the corresponding indicators of the k-th high-value curve segment.

What needs to be explained here is that the two boundary point flag bits of the sub-area are 1 and 2 respectively, and their sum is 3. 1 and 2 do not refer to the number, but to the mark, for example, the two points of the fifth sub-area The boundary points are P _5,1 , P _5,2 . Since the UAV's entry point flag for the k-1th sub-area is nk _-1 , the UAV's flight-out point flag after the sub-area is searched is 3-nk _-1 .

The sub-region sorting problem can be regarded as a typical traveling salesman problem, so the nearest insertion method is adopted to obtain an approximate optimal sorting result. During each iteration, from the remaining unsorted subregion collection Randomly select a sub-area in and insert it into any position of the sorted sub-area sequence {l ₁ ,...,l _I }, and calculate the index value J(I+1) of the new sorting area according to the formula of the sorting model, Then calculate the increment ΔJ=J(I+1)-J(I) of the index value, select the situation when ΔJ takes the maximum value from all the above-mentioned possible situations, and then insert the selected unsorted sub-region into the sorted in the sequence of subregions. The above iterative process repeats K steps, that is, the approximate sorting of K sub-regions is completed.

Figure 5 shows the relative three _sub -regions (P _1,1 ,P _1,2 ), (P _2,1, P _2,2 ) and (P _3,1 ,P _{3, 2} ), the process of approximately sorting the three sub-regions by using the nearest insertion method, in which the solid line represents the river sub-region, the dotted line represents all possible transition track segments, and the solid line represents the determined transition track segment. It can be seen that the possible transition paths from the starting position P _initial include (P _initial ,P _1,1 ), (P _initial ,P _1,2 ), (P _initial ,P _2,1 ), (P _initial , P _2,2 ), (P _initial ,P _3,1 ), (P _initial ,P _3,2 ); after one iterative calculation, determine (P _initial ,P _1,1 ) as the optimal path, and start For the next path iterative calculation, the possible transition paths include those shown in figure (c). After iterative calculation, (P _1,2 ,P _2,1 ) is determined to be the optimal path; after repeating this process, the figure ( f) The search path shown.

After sorting, the level of each river sub-area (that is, the search order of the sub-areas) and the flag bit of the entry point of each sub-area (that is, which of the two endpoints of each sub-area enters the sub-area) can be determined. , the UAV will enter the sub-area according to the flag, and search one by one in the order of the sub-areas.

The above search algorithm is suitable for the search trajectory planning of a UAV performing a search mission.

However, in the prior art, in order to achieve efficient search, usually a plurality of unmanned aerial vehicles jointly perform the search task. Therefore, this embodiment further provides an algorithm for UAV search trajectory planning in the case that multiple UAVs jointly perform a search task.

When multiple UAVs perform search tasks together, it is necessary to comprehensively consider the total search revenue and the balance of the task, and assign sub-areas to each UAV.

Specifically, if multiple unmanned aerial vehicles perform the search task, the following steps are further included: assigning the sorted value search curve segments to multiple unmanned aerial vehicles, and performing area allocation.

A method for area allocation, comprising the following steps:

The high-value search curve segment is assigned to each drone, and each drone obtains a collection of assigned high-value search curve segment sets for A _i , and the curve segment set includes multiple high-value search curve segments;

In order to avoid collisions between UAVs, each sub-area is required to be assigned to only one UAV, and in order to ensure the efficient use of UAVs, it is required to assign at least one sub-area to each UAV, so all UAVs The set of high-value search curve segments allocated by the machine satisfies the following constraints:

where _Nu denotes the number of drones and K denotes the number of high-value search curve segments.

For the process of allocating the search paths of multiple UAVs, refer to the method in steps S4 and S5, and perform index calculation (including sub-region transitions) for the high-value search curve segments in the curve segment set A _i obtained by each UAV. Time, sub-area coverage time and sub-area coverage return), get J _i ;

Model the region assignment as Among them, _Nu represents the total number of UAVs, ρ ₁ represents the proportional coefficient of the total search revenue, and ρ ₂ represents the proportional coefficient of the task balance degree; the values of ρ ₁ and ρ ₂ need to be assigned according to the specific task requirements, for example, each aircraft When the detection capability is similar, the value of _ρ2 can be increased; when the detection capability of each aircraft is quite different, the value of _ρ1 can be increased.

The iterative method is used to solve the regional allocation modeling and determine the regional allocation strategy, as shown in Figure 6. The final optimized track can be expressed as:

The algorithm of the invention is used to plan the track of the UAV, so that the efficient search of the target river section can be realized.

Claims (6)

1. be applied to the unmanned aerial vehicle track planning algorithm of river target search task, it is characterized in that, the field of view scope of described unmanned aerial vehicle is greater than river width, and described planning algorithm comprises the following steps: S1: abstract the target river to be searched into a two-dimensional plane curve, take the forward distance L _s as an independent variable, and discretize the plane curve into M=L/L _s discretization units, where L is the total length of the river s _m , L _s is the length of each discretization unit; S2: In any sampling period, the UAV moves within two discretization units s _m ; the probability p(s _m ) of the target to be searched in each discretization unit s _m is known, and p(s _m )∈[0,1], and satisfy S3: Any number of discretized units compose the search curve segment, S _k = {s _m ,s _m+1 ,...,s _n }, 1≤m<n≤M, use the Gaussian mixture model to describe the target to be searched in each The probability information existing in the search curve segments, extract the high-value search curve segment S _k with P _k,1 and P _k,2 as the boundary of the region, where the number of high-value search curve segments is K, k∈{1,2 ,...,K}; S4: Define the sub-area transition time as the shortest flight time for the boundary point P _{i, p} of the current search curve segment to go to the boundary point P _{j, q} of the next search curve segment, and define the transition time as: T(P _i,p ,P _j,q )=λ ₁ ·Dubins_cost(P _i,p ,P _j,q ), where i≠j and i,j∈{1,2,...K}, and p,q∈{1,2}; where λ ₁ represents the proportional coefficient of sub-region transition time; Dubins_cost(P _i,p ,P _j,q ) represents the length of Dubins curve, p represents the i-th search curve segment Flying point flag, q represents the entry point flag of the jth search curve segment, K represents the number of high-value search curve segments; Define the sub-area coverage time as the time when the UAV searches for the flight in the search curve segment, which is C _k = λ ₂ L _k , where λ ₂ represents the proportional coefficient of the sub-region coverage time, and L _k represents the length of the search curve segment, L _k = (n-m+1)·L _s ; Define the sub-area coverage return as the cumulative probability obtained by the drone completely covering the high-value search curve segment: R _k ; S5: Using the above-mentioned sub-area transition time, sub-area coverage time and sub-area coverage return as evaluation indicators, adopt the nearest insertion method to iteratively sort the search sequence of each high-value search curve segment, and the sorting model is: S6: Connect the optimized high-value search curve segments in sequence to obtain the final search path of the UAV. 2. the unmanned aerial vehicle track planning algorithm that is applied to river target search task as claimed in claim 1, is characterized in that, the method for extracting high-value search curve segment may further comprise the steps: Using K one-dimensional Gaussian functions Form a Gaussian mixture model, where k=1,2,...K, u _k is the mean value of each Gaussian function, and σ _k is the standard deviation of each Gaussian function; Let the weight coefficient of each Gaussian function be α _k , and satisfy The target probability of the curve segment to be searched is Estimate the weight coefficient α _k , mean value u _k , and standard deviation σ _k of each Gaussian function, and iteratively estimate p(s) until the convergence condition is met; The corresponding intervals of 95.4% probability of each Gaussian function are obtained, and P _k,1 =u _k -2σ _k and P _k,2 =u _k +2σ _k are the high-value search curve segments of the area boundary points, then L _k =4σ _k ; R _k =0.954α _k . 3. the unmanned aerial vehicle track planning algorithm that is applied to river target search task as claimed in claim 2, is characterized in that, described convergence condition is |p(s)-p'(s)|＜ξ, wherein p (s) and p'(s) are target probability values before and after iteration, respectively, and ξ=10 ^-5 . 4. the unmanned aerial vehicle track planning algorithm that is applied to river target search task as claimed in claim 1, it is characterized in that, if a plurality of unmanned aerial vehicles carry out search mission, then further comprise the following steps: to sorted value search Curve segments are assigned to multiple drones for area assignment. 5. the unmanned aerial vehicle track planning algorithm that is applied to river target search task as claimed in claim 4, is characterized in that, the method for area distribution, comprises the following steps: Assign high-value search curve segments to each UAV, and each UAV obtains a set of assigned high-value search curve segments whose set is A _i ; According to the method in step S5, perform index calculation on the high-value search curve segment in each UAV set A _i to obtain J _i ; Model the region assignment as Among them, _Nu represents the total number of UAVs, _ρ1 represents the proportional coefficient of the total search revenue, and _ρ2 represents the proportional coefficient of the task balance degree; The iterative method is used to solve the modeling of area allocation and determine the area allocation strategy. 6. the unmanned aerial vehicle track planning algorithm that is applied to river target search task as claimed in claim 5, it is characterized in that, the high-value search curve section set that all unmanned aerial vehicles are assigned meets the following constraints: