CN110704259A

CN110704259A - Efficient test path optimization method for digital microfluidic biochip

Info

Publication number: CN110704259A
Application number: CN201910846939.2A
Authority: CN
Inventors: 黄喜军; 许川佩; 张龙; 曾莹; 李翔; 胡聪
Original assignee: Guilin University of Electronic Technology
Current assignee: Guilin University of Electronic Technology
Priority date: 2019-09-09
Filing date: 2019-09-09
Publication date: 2020-01-17
Anticipated expiration: 2039-09-09
Also published as: CN110704259B

Abstract

The invention relates to a high-efficiency digital microfluidic biochip test path optimization method, which solves the technical problem of high blindness in random search direction, and comprises the following steps of defining an array unit of a test chip as a vertex, connecting adjacent vertexes by using edges, converting the test chip into a non-directional connected graph, and completing the construction of a test model; allocating a different decimal to each edge in the undirected connected graph for representing the priority level of the edge; defining a priority strategy as a path selection strategy, generating a path of a traversal edge as a test path by using the priority strategy, wherein the test path is a vertex set which is sequentially walked by the traversal edge; step three, optimizing the test path in the step two by using a genetic algorithm; the priority strategy comprises a technical scheme of an off-line test priority strategy and an on-line test priority strategy, the problem is well solved, and the method can be used for testing the digital microfluidic biochip.

Description

Efficient test path optimization method for digital microfluidic biochip

Technical Field

The invention relates to the field of off-line and on-line testing of digital microfluidic biochips, in particular to a high-efficiency method for optimizing a testing path of a digital microfluidic biochip.

Background

The digital microfluidic biochip is generally composed of two layers of plates, wherein the top layer is a continuous grounding electrode, the bottom layer is a control electrode capable of being independently driven, and liquid drops between the plates are moved, mixed and the like by applying different high and low levels to the control electrode. When a chip has a catastrophic failure such as short circuit, the liquid drops will be stagnated before passing through the array unit with the failure, so that the biochemical experiment can not be normally carried out. Because the reliability of the digital microfluidic chip is highly required by biochemical experiments, the chip needs to be fully tested, and not only needs to be tested off-line after the chip is produced and before the experiment is carried out, but also needs to be tested on line during the experiment is carried out, so that the chip fault can be found in time, and the reliability of the experiment result can be ensured.

Aiming at the problem of testing catastrophic failures of a digital microfluidic biochip, the method of traversing all array units of the chip by using test droplets is adopted in the prior art, and whether the chip has failures is judged by judging whether the test droplets successfully reach a detection area through a capacitance detection circuit. However, this method cannot detect all short-circuit failures, which cannot be detected when the droplet advancing direction does not coincide with the wiring direction of the short-circuited array unit. Therefore, researchers have proposed traversing not only all array elements but also the boundaries between array elements, that is, converting vertex-based detection into edge-based detection, and have proposed implementing the method using euler loops. The method comprises the steps of firstly converting a chip into an undirected graph model, then enabling edges in the undirected graph to form an Euler loop in a mode of adding the edges, and traversing all the edges in the Euler loop by using a test liquid drop according to a Fleury algorithm. However, the method needs more waiting time during on-line testing, and the detection efficiency is not high. In addition, researchers put forward a parallel scanning method for the chip by adopting multiple liquid drops to carry out fault detection, and also put forward a built-in self-test method on the basis of the parallel scanning method, so that the offline detection efficiency is further improved. However, the above methods are not suitable for online testing, and some researchers propose to optimize the testing path of the chip by using the ant colony algorithm, and although the online testing efficiency is improved to a certain extent, the testing path is searched and optimized only by using the ant colony algorithm, so that the searching randomness is high under the condition of not fully utilizing heuristic information, the convergence speed is low, and the optimal value cannot be converged.

In order to further reduce the test time and optimize the test path, the invention combines heuristic information with a random search algorithm to reduce the blindness of the search of the test path, adopts a hybrid method (PS-GA) combining a Priority Strategy (PS) and a Genetic Algorithm (GA) to search and optimize the test path, and seeks the shortest test path on the basis of traversing the boundaries between the chip array units and the array units, namely traversing all the edges in an undirected graph.

Disclosure of Invention

The technical problem to be solved by the invention is the technical problem of high blindness in the random search direction in the prior art. The method for optimizing the testing path of the digital microfluidic biochip has the characteristic of high optimization efficiency.

In order to solve the technical problems, the technical scheme is as follows:

an efficient method for optimizing a test path of a digital microfluidic biochip, the method comprising:

defining an array unit of a test chip as a vertex, connecting adjacent vertices by using edges, converting the test chip into a multidirectional connection graph, and completing construction of a test model; allocating a different decimal to each edge in the undirected connected graph for representing the priority level of the edge;

defining a priority strategy as a path selection strategy, generating a path of a traversal edge as a test path by using the priority strategy, wherein the test path is a vertex set which is sequentially walked by the traversal edge;

and step three, optimizing the test path of the step two by using a genetic algorithm.

The priority policy comprises an offline test priority policy and an online test priority policy.

In the above scheme, to implement test path optimization, further, the offline test priority policy is:

step A, if the vertex where the test liquid drop is located has an adjacent edge which is not traversed, selecting the vertex corresponding to the edge with the highest priority level as the next searching vertex;

and step B, if the adjacent edges are traversed, determining the shortest path from the test liquid drop to the untested edge in the test model according to the Floyd algorithm, guiding the test liquid drop to move along the shortest path, and finishing selecting the top point with the shortest distance from the untested edge as the next search point from the test liquid drop.

Further, the online test priority policy is:

step a, defining constraint conditions, judging the number of effective adjacent points when the adjacent vertex of the test liquid drop meets the constraint conditions and is an effective adjacent point exists, and executing the step d if the number of the effective adjacent points is 0; otherwise, the test liquid drop does not return, and the step b is executed;

b, if the number of the effective adjacent points is 1, selecting the vertex as the vertex to be passed by at the next moment; if the number of the effective adjacent points is more than 2, judging the number of the non-traversed adjacent edges;

c, if the number of the non-traversed adjacent edges is 0, selecting an effective adjacent point closest to the non-traversed edges as a vertex to be passed by at the next moment according to a Floyd algorithm; if the quantity of the adjacent edges which are not traversed is 1, selecting a vertex corresponding to the adjacent edge; if the quantity of the adjacent edges which are not traversed is larger than 1, selecting a vertex corresponding to the edge with the highest priority level as a vertex to be passed next;

d, if the number of the effective adjacent points is 0, making a backspacing mark back _ flag equal to 1, and executing the step e;

step e, deleting the current vertex from the effective adjacent points of the previous vertex of the current vertex, namely deleting the adjacent points which do not meet the constraint condition; deleting the mark of the edge corresponding to the current vertex, and subtracting 1 from the mark edge _ flag of the edge where the current vertex is located; and returning the time pointer from the time t to the time t-1 to point to a vertex before the current vertex.

Furthermore, the constraint conditions are static constraint and dynamic constraint;

static constraints are that the droplets cannot be directly adjacent or diagonally adjacent:

the dynamic constraint is that a droplet is not located adjacent to another droplet at the same time, neither at the previous nor the next time.

Furthermore, the test model construction also comprises the step of adding an additional virtual edge, converting the vertex with the odd degree into the vertex with the even degree, and converting the undirected connected graph into the Euler loop graph.

Further, the second step comprises:

generating an offline test path based on the offline test priority policy:

step 2.1, appointing a vertex in the undirected connected graph as a search starting point;

step 2.2, judging whether all adjacent edges of the vertex are traversed or not according to an offline test priority strategy, and determining the next adjacent vertex to pass through;

step 2.3, continuously repeating the step 2.2 until all edges are traversed, recording sequentially passed vertexes as a feasible solution of traversing edges of the undirected connected graph, and obtaining a test path of the offline test;

generating an online test path based on the online test priority strategy:

step 2.4, appointing a vertex meeting the constraint condition as a search starting point, and marking the adjacent vertex meeting the constraint condition at the next moment as an effective adjacent point;

step 2.5, judging whether the adjacent vertex of the current vertex meets the constraint condition of the next moment or not according to the online test priority strategy, and selecting the vertex to pass through at the next moment or executing rollback operation;

and 2.6, continuously repeating the step 2.5 until all edges are traversed, and recording the sequentially passed vertexes as a test path of the online test.

Further, the third step comprises:

step 3.1, initializing the genetic algorithm, and setting the total iteration times N of the genetic algorithm_ITPopulation size N_PSChromosome length L_CCross probability p_cAnd the mutation probability p_m；

Step 3.2, generating an initial population according to the population size and the chromosome length;

step 3.3, solving the test path according to the priority strategy, and calculating the length of each path

And solving a local optimal solution;

step 3.4, calculating the fitness function value

Step 3.5, calculating the selection probability

Step 3.6, selecting the group according to the roulette mode and the elite reservation strategy;

step 3.7, performing cross operation on the group according to the cross probability;

step 3.8, carrying out mutation operation according to the mutation probability;

and 3.9, repeating the steps 3.3 to 3.8 until the iteration times are finished.

The invention transforms the test chip into a non-directional connected graph, namely a test model, and allocates a different decimal to each edge in the graph as the priority level of the edge, wherein the larger the decimal value is, the higher the priority is. Designing a priority strategy as a path selection strategy according to the test characteristics of the chip, and generating a path of the traversal edge by using the designed priority strategy, wherein the path is a vertex set which is sequentially walked by the traversal edge.

The designed offline test priority policy is as follows: (1) and if the vertex where the test liquid drop is located has the adjacent edges which are not traversed, selecting the vertex corresponding to the edge with the highest priority level as the next searching vertex.

(2) If the adjacent edges are traversed, determining the shortest path from the test liquid drop to the untested edge in the test model according to the Floyd algorithm in the graph theory, and guiding the test liquid drop to move along the shortest path, namely, the test liquid drop selects the vertex with the shortest distance to the untested edge as the next search point.

The designed online test priority policy is as follows: (1) if the adjacent vertex of the test droplet meets the constraint condition, that is, a valid adjacent point exists, the test droplet does not need to retreat, and the retreat flag back _ flag is 0. If there are only 1 valid neighbors, then the vertex is selected as the vertex to be passed by at the next time. If there are more than 2 valid adjacency points, then: a. and if the adjacent edges are traversed, selecting the effective adjacent point closest to the non-traversed edge as a vertex to be passed by at the next moment according to the Floyd algorithm. b. If the adjacent edges which are not traversed exist, when only 1 adjacent edge which is not traversed exists, the vertex corresponding to the edge is selected, and if more than 1 adjacent edge which is not traversed exists, the next corresponding vertex is selected according to the priority level, namely the priority coefficient.

(2) If the adjacent vertex of the test droplet does not satisfy the constraint condition, the backoff flag back _ flag needs to be set to 1, and the following backoff operation is performed. Firstly, in the valid adjacent points of the previous vertex, deleting the current vertex, namely deleting the adjacent points which do not meet the constraint condition. Then, deleting the mark of the corresponding edge, namely subtracting 1 from the mark edge _ flag of the current edge (the mark edge _ flag of the edge is automatically added with 1 after the edge is passed, so that 1 is subtracted after the edge is retreated, and the edge is not marked). Finally, the time pointer is backed from time t to time t-1, and points to the previous vertex. So far, the rollback operation ends. Namely, the 'deadlock' problem generated in the path generation process is solved by the rollback operation of the liquid drop.

And carrying out intersection and variation operation on the priority levels of the edges in the selected test path through a genetic algorithm to generate a new priority level, and gradually obtaining a shorter test path in an iterative manner, thereby realizing the optimization of the test path.

The invention has the beneficial effects that: the invention can reduce the adverse effect generated in the random search aspect only depending on the intelligent algorithm, improve the optimization effect of the test path and enable the test path to approach or be equal to the shortest path by combining heuristic information with a random search algorithm, namely by combining a priority strategy with a genetic algorithm. For the phenomenon of deadlock of test liquid drops in the process of searching the online test path, a backspacing strategy is utilized to solve the problem.

Drawings

The invention is further illustrated with reference to the following figures and examples.

FIG. 1 is a schematic diagram of a test model of a digital microfluidic biochip and its Euler model.

Fig. 2 is a schematic diagram of test droplet routing during off-line testing.

FIG. 3 is a diagram illustrating deadlock.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Example 1

The embodiment provides an efficient method for optimizing a test path of a digital microfluidic biochip, which includes:

Fig. 1 shows a test model of a digital microfluidic biochip and its euler model, which includes 5 rows and 5 columns of array units. Fig. 1 (a) shows a physical model of a chip, and each array unit represents a control electrode. According to the graph theory, the chip is regarded as a non-directional connected graph G, the array units are converted into vertexes V in the non-directional connected graph G, and adjacent vertexes are connected by edges E, so that the graph theory model shown in (b) in FIG. 1, namely the test model of the chip, can be obtained.

Preferably, the undirected connected graph G can be eulerized to obtain an eulerized graph G1, and as shown in (c) of fig. 1, the undirected connected graph G is changed into a graph with eulerized loops by adding an additional virtual edge to convert the vertex with odd degree into the vertex with even degree. Since the start point and the end point of the euler loop must be the same vertex, and the euler path does not need to satisfy this condition, the euler path may have 1 or 2 fewer edges than the euler loop.

Defining the microfluidic chip as an array with m rows and N columns, wherein the total number of edges is N_EM × (n-1) + (m-1) × n. According to the Euler loop theory of the undirected graph, the total number of edges which go back to the starting point after traversing all the edges in the graph once and only once is N_ECThe total number of edges in the Euler path can be N, as shown in equation (1)_EC-1 or N_EC-2。

In order to avoid collision of test droplets with experimental droplets during on-line testing, further consideration of fluid constraints between droplets is required. Suppose that

Is a droplet l₁The row and column at time t: (

Is a droplet l₁The row and column at time t + 1), similarly, let

Is a droplet l₂The row and column at time t: ( Is a droplet l₂The row and column at time t + 1), and the fluid constraints can be divided into static and dynamic constraints, as follows:

1) static constraints

Or

It is indicated that two stationary droplets at the same time are in a row or column that differs by at least 2 positions, i.e. the droplets cannot be directly or diagonally adjacent, otherwise a droplet fusion will occur.

2) Dynamic constraints

Or

Or

Or

The constraint condition that two droplets cannot be adjacent to each other when the two droplets are different by one time is satisfied, that is, the position of one droplet at the time cannot be adjacent to the position of the other droplet at the previous time and the next time.

The method for realizing the test path optimization based on the priority strategy and the genetic algorithm specifically comprises the following steps:

1. generation of test paths based on priority policy

(1) The problem of how to select the next vertex by the test droplet is solved according to the foregoing priority strategy, and the design steps for generating the off-line test path in this embodiment can be briefly described as follows:

step 1: one vertex in the undirected graph is designated as the search starting point.

Step 2: and judging whether all adjacent edges of the vertex are traversed or not according to an offline test priority strategy, and determining the next adjacent vertex to be traversed.

And step 3: and (3) continuously repeating the step (2) until all the edges are traversed, and recording the sequentially passed vertexes as a feasible solution of the traversed edges of the undirected graph, namely obtaining a test path of the offline test.

FIG. 2 is a schematic diagram of the test drop selecting the next adjacent vertex according to a priority strategy during off-line testing. Wherein, fig. 2 (a) describes the case that the adjacent edges are not traversed, after the test droplet reaches the vertex 13, the test droplet will select to go to the vertex 14 because the priority coefficient of the edge from the vertex 13 to the vertex 14 is 0.60, which is greater than the priority coefficient of the other 2 edges, that is, the priority level of the edge is the highest;

fig. 2 (b) shows a case where all the adjacent edges have traversed, the path of the test droplet is (13,14,19,20,25,24,19,18,23,24), and after the droplet reaches the vertex 24, the adjacent edges around have traversed, so that the next vertex is no longer selected according to the size of the priority coefficient, but the vertex with the closest distance to the non-traversed edge is found according to the Floyd algorithm, and the resulting vertex is 23, and then the test droplet will reach the vertex 23 from the vertex 24.

(2) The process of generating the online test path according to the defined online test priority policy is as follows:

step 1: and appointing a vertex meeting the constraint condition as a search starting point, and marking the adjacent vertex meeting the constraint condition at the next moment as a valid adjacent point.

Step 2: and judging whether the adjacent vertex of the current vertex meets the constraint condition at the next moment or not according to the online test priority strategy, and selecting the vertex to be passed by at the next moment or executing the rollback operation.

And step 3: and (3) continuously repeating the step (2) until all edges are traversed, and recording the sequentially passed vertexes as a test path of the online test.

In the process of generating a path by using a test droplet, the situation that the test droplet cannot stay in place for waiting and cannot move forward due to the fluid constraint is called a deadlock phenomenon. This phenomenon also occurs when the test path is generated using the online test priority policy, and can be solved by a rollback operation.

Fig. 3 is a schematic diagram of a deadlock situation, where an experimental droplet 1 is located at the 39 th array unit at time t, and will go to the 38 th and 37 th array units in turn; experiment droplet 2 is located at the 12 th array element and will go to the 20 th and 28 th array elements again; and the test drop is located at the 36 th array element when the test drop satisfies the fluidic constraint. But at the next moment, the test droplet will not meet the fluid constraints, can not stop in place, and can not go to 4 adjacent vertices, resulting in a "deadlock". Once this occurs, the test drop must be retracted to the position of the previous time, i.e., the 35 th array element, and the 36 th array element is marked as an invalid adjacent vertex, and the routing is resumed.

2. Genetic algorithm implementation optimization of test path

Before test path optimization is carried out by using Genetic Algorithm (GA), the total iteration times N of the genetic algorithm is set according to the characteristics of the size and the like of a test chip_ITPopulation size N_PSChromosome length L_CCross probability p_cAnd the probability of variation p_mAnd the like.

1) Initial population generation

The generated initial population contains N_PSIndividual, individual length L_CCorresponding to the number of edges in the undirected graph. In the priority strategy, the priority coefficient of each edge in the undirected graph is a random number of 0-1 generated according to a chaotic operator, and L_CThe priority data constitutes the individuals encoded by real numbers in the genetic algorithm, i.e. the initial population is actually generated by N_PSEach length is L_CIs calculated.

2) Fitness function design

On the basis of generating the initial population, a group of test paths traversing all edges in the undirected graph can be solved according to a priority strategy, and the result of decoding by the genetic algorithm is corresponded. The feasible solution, i.e., the obtained test path, can be represented as

Wherein t is 1,2, N_ITDenotes the number of iteration, j ═ 1,2, ·, N_PSAnd indicates an individual number.

The formula for calculating the fitness function value is as follows:

wherein the content of the first and second substances,

in order to be a function of the fitness measure,

for testing the length of the path, L_EulerPathFor the length of the chip Euler path, ε is a constant (0.000001) that takes a minimum value to prevent computation overflow.

3) Improvement of selection probability

Firstly, according to the principle that the probability of individual selection is in direct proportion to the fitness value, obtaining the initial selection probabilityIs shown as formula (5), wherein

As a fitness function, N_PSIs the population size.

(5) The initial selection probability in the formula is also relatively small in general, the efficiency is low when group selection is carried out, more time is needed for selection operation, the difference of individual fitness values cannot be reflected during selection, and the selection probability is expanded according to the formula (6) similarly to random selection. Wherein the rand () random function generates random numbers of 0-1, so that the selection probability corresponding to the individual with the maximum fitness value in the group is about 0.6-0.93,for preliminary selection of probabilities

The maximum value of (a) is,

a selection probability for an individual. In this case, the individual selection probabilities are also amplified in the same proportion according to equation (6), and the efficiency of the selection operation can be improved.

After the selection probability is determined, the local optimum is retained according to the elite retention strategy, and then N is selected according to the roulette mode_PS-1 individuals, together constituting N_PSAnd (4) solving.

4) Crossover and mutation operation design

Firstly, to the groupThe lines are grouped into 2 individuals, each individual is a group of priority data with the length L_C. And each group of individuals adopts a single-point crossing mode, and judges whether 5 data after the randomly selected crossing point need to be crossed or not according to the comparison between the crossing probability and the random number. Because each individual length L_CLarger, to further increase the diversity of the population, so each set of priority data is processed

And (4) performing secondary crossing operation.

The mutation operation is similar to the crossover operation in processing mode, and whether each individual needs to mutate 5 data after the position is selected is judged according to the mutation probability. Also, each set of priority data is processed

And (5) performing mutation operation. After the selection, crossing and mutation operations are performed on the priority data, new priority data are generated, and a new feasible solution can be obtained according to the priority strategy.

5) Algorithm design flow

Through the foregoing description of the genetic algorithm, the design steps of the genetic algorithm for implementing path optimization are briefly described as follows:

step 1: the genetic algorithm is initially set.

Step 2: based on a given population size and chromosome length, an initial population is generated.

And step 3: solving the test path according to the designed priority strategy, and calculating the length of each pathAnd a local optimum solution is solved.

And 4, step 4: calculating a fitness function value

And 5: determining a selection probability

Step 6: and selecting the group according to the roulette mode and the elite reservation strategy.

And 7: and performing cross operation on the group according to the cross probability.

And 8: and carrying out mutation operation according to the mutation probability.

And step 9: and repeating the steps 3 to 8 until the iteration number is completed.

3. Simulation result of algorithm

(1) Off-line test simulation result

When the chip is tested off line, the size N of the genetic algorithm group is set_PS30, iteration number N_IT300, cross probability p_c0.6 and the probability of variation p_m0.2. Meanwhile, the test is carried out in a mode that rows and columns of the digital microfluidic chip are equal, namely m is equal to n, wherein m is 7, 9, 11, 13 and 15 respectively. The measured shortest test path length is shown in table 1, where the IACA algorithm is an improved ant colony algorithm and the PMF algorithm is an euler loop method, and it can be seen that the algorithm proposed herein can obtain a shorter test path length, and the shortest path of the traversal edge in the graph theory has been reached in the path length.

TABLE 1 shortest test Path Length during offline test

Chip scale	7×7	9×9	11×11	13×13	15×15
						IACA algorithm	100	165	254	350	470
PMF algorithm	96	160	240	336	448
						Text algorithm	94	158	238	334	446

(2) On-line test simulation result

In comparison with the offline test, the online test needs to consider the fluid constraint of the test droplet on the test droplet, so that the difficulty in finding the optimized path is further increased. When the algorithm is used for testing simulation, the iteration times N of the genetic algorithm corresponding to the 15 multiplied by 15 chip is set_ITThe other test parameters are the same as the off-line test at 500. The resulting shortest test path length is shown in table 2, from which it can be seen that the shortest test path length is already close to or equal to the limit of the euler path.

TABLE 2 shortest test Path Length in Online testing

Chip scale	7×7	9×9	11×11	13×13	15×15
						IACA algorithm	105	171	260	361	471
PMF algorithm	110	178	269	380	486
						Text algorithm	95	158	239	335	446

In the embodiment, by combining heuristic information with a random search algorithm, that is, by combining a priority strategy with a genetic algorithm, adverse effects generated in the random search aspect only depending on an intelligent algorithm can be reduced, the optimization effect of the test path is improved, and the test path approaches or equals to the shortest path. For the phenomenon of deadlock of test liquid drops in the process of searching the online test path, a backspacing strategy is utilized to solve the problem.

Although the illustrative embodiments of the present invention have been described above to enable those skilled in the art to understand the present invention, the present invention is not limited to the scope of the embodiments, and it is apparent to those skilled in the art that all the inventive concepts using the present invention are protected as long as they can be changed within the spirit and scope of the present invention as defined and defined by the appended claims.

Claims

1. An efficient test path optimization method for a digital microfluidic biochip is characterized by comprising the following steps: the method for optimizing the test path of the digital microfluidic biochip comprises the following steps:

step three, optimizing the test path in the step two by using a genetic algorithm;

2. The method for efficient digital microfluidic biochip test path optimization according to claim 1, wherein: the offline test priority policy is:

3. The method for efficient digital microfluidic biochip test path optimization according to claim 1, wherein: the online test priority policy is as follows:

4. The method for efficient optimization of digital microfluidic biochip test path according to claim 3, wherein: the constraint conditions are static constraint and dynamic constraint;

5. The efficient digital microfluidic biochip test path optimization method of claim 3, wherein: the test model construction also comprises the steps of adding additional virtual edges, converting vertexes with odd degrees into vertexes with even degrees, and converting the undirected connected graph into an Euler loop graph.

6. The method for efficient digital microfluidic biochip test path optimization according to claim 1, wherein: the second step comprises the following steps:

generating an offline test path based on the offline test priority policy:

generating an online test path based on the online test priority strategy:

7. The method for efficient digital microfluidic biochip test path optimization according to claim 1, wherein: the third step comprises:

step 3.3, solving the test path according to the priority strategy, and calculating the length of each pathAnd solving a local optimal solution;

step 3.4, calculating the fitness function value

Step 3.5, calculating the selection probability

and 3.9, repeating the steps 3.3 to 3.8 until the iteration times are finished.