Hybrid parking path planning method based on improved particle swarm and artificial fish swarm algorithm

Abstract

The present application provides a hybrid parking path planning method based on an improved particle swarm and an artificial fish swarm algorithm, comprising: establishing a parking environment model according to an actual environment of a parking area, and numbering the road section nodes in the parking environment model to obtain a particle swarm, performing algorithm improvement on the inertia weight, learning factor, and fitness function of the particle swarm, searching for an optimal parking point in the parking environment model according to the improved particle swarm algorithm, and forming an optimal parking point set T with the optimal parking points, introducing the inertia weight of the particle swarm into an artificial fish swarm to obtain an improved artificial fish swarm algorithm, mixing the improved particle swarm and the artificial fish swarm algorithm to obtain a hybrid algorithm, and planning an optimal parking path according to the hybrid algorithm to solve the problem of inaccurate and inconvenient parking path planning caused by the limitations of a single algorithm.

Description

Technical Field

[0001] The present application relates to the technical field of parking path planning, and in particular to a hybrid parking path planning method based on improved particle swarm and artificial fish swarm algorithms. Background Art

[0002] As the number of cars increases, the parking problem is becoming increasingly serious. Statistics show that the ratio of cars to parking spaces in China is only 0.5, far below the average level of developed countries. Car owners need to spend a lot of time looking for parking spaces, which not only increases the burden on roads and causes urban traffic congestion, but also brings many unfavorable factors to social development. At present, an effective way to solve the parking problem is to plan an optimal parking route for car owners, which can not only save car owners' parking time but also reduce the burden on roads, effectively improving parking efficiency.

[0003] The traditional parking method uses AGV (Automated Guided Vehicle), which is an "automatic guided transport vehicle" equipped with electromagnetic or optical automatic guidance devices, capable of traveling along a specified guide path, and having safety protection and various transfer functions. The method of using AGV to plan parking paths can effectively plan convenient parking paths for indoor parking lots with simple paths, but AGV cannot make convenient plans for outdoor environments and complex road conditions. The optimal outdoor environment parking path refers to finding a path with the shortest time from the owner's location to the parking area and from the parking area to the final parking point. In the full outdoor environment, the optimal parking path planning usually encounters many uncertain factors, such as: the road congestion is different in each time period and each road. The greater the traffic volume, the lower the parking efficiency of the owner. When finding the optimal parking path for the owner, the congestion of each road is often given priority; road quality. The worse the road quality, the higher the bumpiness, and the lower the parking efficiency. Therefore, under such complex road conditions, the traditional form of using AGV for parking specifications cannot meet the owner's needs for fast and convenient parking. Summary of the invention

[0004] The present application provides a hybrid parking path planning method based on improved particle swarm and artificial fish swarm algorithms to solve the problem of inaccurate and inconvenient parking path planning due to the limitations of a single algorithm.

[0005] A hybrid parking path planning method based on improved particle swarm and artificial fish swarm algorithm includes:

[0006] According to the actual environment of the parking area, a parking environment model is established, and the road section nodes in the parking environment model are numbered to obtain a particle swarm;

[0007] Improving the algorithm for the inertia weight, learning factor and fitness function of the particle swarm;

[0008] Searching for optimal parking points in the parking environment model according to the improved particle swarm algorithm, and forming an optimal parking point set T from the optimal parking points;

[0009] Introducing the particle swarm inertia weight into the artificial fish swarm to obtain an improved artificial fish swarm algorithm;

[0010] Mixing the improved particle swarm algorithm and artificial fish swarm algorithm to obtain a hybrid algorithm;

[0011] An optimal parking path is planned according to the hybrid algorithm.

[0012] Preferably, the inertia weight improvement formula is:

[0013]

[0014]

[0015]

[0016] Among them, W min is the minimum value of inertia weight; W max is the maximum value of the inertia weight; W ce is the maximum variable value of W, which is W max and W min The difference between r is the initial value of inertia weight; g i is the current iteration number; g min is the minimum number of iterations; g max is the maximum number of iterations; W will change with the current number of iterations g of the particle i The inertia weight W increases from small to large in the early stage and decreases from large to large in the later stage. W needs to be kept at W max With W min During the rising phase, W increases to a value greater than W max , then W is equal to W max .

[0017] Preferably, a dynamic adjustment method is used to improve the learning factor, and the improved formula is:

[0018]

[0019]

[0020] Among them, C1 and C2 represent individual learning factors and social learning factors respectively, and C r1 , C r2They represent the initial values ​​of individual learning factor and social learning factor, respectively, i1 , g i2 Indicates the current iteration number of the particle, g max1 , g max2 represents the maximum number of iterations of the particle, C max1 , C max2 Represents the final value of the particle learning factor.

[0021] Preferably, the improvement steps of the fitness function are:

[0022] Get the initial fitness function:

[0023]

[0024] Obtain the road congestion coefficient δ, the actual total length of the path s, the weighted total length of the path affected by road quality S, and the parking quantity influence coefficient θ;

[0025] The improved fitness function obtained by combining the initial fitness function is:

[0026]

[0027] Among them, the function f(xy) is the distance traveled by the owner to the parking spot, D ij represents the node i(x i ,y i ) and node j(x j ,y j ), d ij represents the node i(x i ,y i ) and node j(x j ,y j ) between the two.

[0028] Preferably, the formula for improving the artificial fish swarm algorithm is:

[0029] Visual next =Visual i *W

[0030] Where W is the particle swarm inertia weight, Visual next Indicates the next fish school field of view during the iteration of the artificial fish school. i Indicates the current fish school field of view.

[0031] It can be seen from the above technical solutions that the present application provides a hybrid parking path planning method based on an improved particle swarm and artificial fish swarm algorithm, including: establishing a parking environment model according to the actual environment of the parking area, and numbering the road section nodes in the parking environment model to obtain a particle swarm, performing algorithm improvement on the inertia weight, learning factor and fitness function of the particle swarm, searching for an optimal parking point in the parking environment model according to the improved particle swarm algorithm, and forming an optimal parking point set T with the optimal parking points, introducing the inertia weight of the particle swarm into an artificial fish swarm to obtain an improved artificial fish swarm algorithm, mixing the improved particle swarm and artificial fish swarm algorithm to obtain a hybrid algorithm, and planning an optimal parking path according to the hybrid algorithm to solve the problem of inaccurate and inconvenient parking path planning caused by the limitations of a single algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0033] Figure 1 A process framework diagram of the parking path planning method;

[0034] Figure 2 This is a simulation diagram of parking area distribution. DETAILED DESCRIPTION

[0035] The following embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following embodiments do not represent all implementations consistent with the present application. They are only examples of systems and methods consistent with some aspects of the present application detailed in the claims.

[0036] The optimal outdoor parking path refers to finding a path with the shortest time from the owner's location to the parking area and from the parking area to the final parking point. The full parking path problem in an outdoor environment has its own unique characteristics and challenges, and more environmental uncertainties need to be considered.

[0037] In the study of the full parking path problem in outdoor environments, the commonly used methods include particle swarm optimization and artificial fish swarm optimization. Particle swarm optimization (PSO) is a group intelligent optimization algorithm that imitates the foraging behavior of birds. The algorithm has the characteristics of simple structure, easy implementation and fast convergence speed, but it is also easy to fall into the local optimum, resulting in low convergence accuracy. Artificial fish swarm algorithm (AFSA) is a new type of intelligent optimization algorithm with good global optimization ability and can quickly jump out of the local optimum, but the algorithm also has the defects of slow convergence in the later stage and low convergence accuracy.

[0038] The technical solution provided in this application provides a hybrid parking path planning method based on improved particle swarm and artificial fish swarm algorithm, see Figure 1 and Figure 2 , we can see that the method steps are:

[0039] S1: establishing a parking environment model according to the actual environment of the parking area, and numbering the road section nodes in the parking environment model to obtain a particle swarm.

[0040] In practical applications, it is necessary to establish an abstract model between the car owner and the parking area. Since the car owner may encounter some uncertainties during the parking process, such as the temporary closure of the road or parking lot due to weather factors, the following assumptions are made in the model:

[0041] a. The parking location information of the parking lot and whether there are any parked vehicles at the parking location are known;

[0042] b. The parking space location information of the parking lot is shared with the car owner;

[0043] c. The final parking route is accessible, and there are no parking lots or road sections that are temporarily closed;

[0044] d. The parking space that the owner is searching for while driving is not occupied.

[0045] According to the location information of the car owner, N parking areas nearby the car owner are searched for, and then the parking spaces in the selected N areas are planned, and each area is judged whether there is a car parked in the parking space, so as to establish an abstract data model between the car owner and the parking area. For example, see Figure 2, is a parking area distribution simulation map. First, the owner's location information is located, and then the parking area is searched, and finally four different parking areas are found. A parking space simulation map is established in each parking area. The hollow circle represents an empty parking space, the solid circle represents a parking space with a car, and the solid line represents a feasible path. An undirected connected weighted graph is established for these four areas and the parking spaces in the area. The nodes of each section from the owner's location to the final parking space are represented by (x i ,y i ) indicates that according to the characteristics of an undirected connected graph, all parking spaces are interconnected, so the following model is established:

[0046]

[0047]

[0048]

[0049] Among them, D ij represents the node i(x i ,y i ) and node j(x j ,y j ), d ij represents the node i(x i ,y i ) and node j(x j ,y j ), and D represents the total distance traveled by the owner from the current position to the final parking space.

[0050] S2: Improve the algorithm for the inertia weight, learning factor and fitness function of the particle swarm.

[0051] In the existing particle swarm algorithm, particles have the defects of strong local search ability and weak global search ability. In order to improve the global search ability of the particle swarm, improvements are made to the inertia weight, learning factor and fitness function of the particle swarm.

[0052] The particle swarm algorithm is an intelligent optimization algorithm that simulates bird foraging. It uses the sharing of individual information to the group so that the entire group can move from disorder to order. During the search process, if the particle swarm updates the latest individual optimal and group optimal values, please refer to the following formula for the update rule:

[0053]

[0054]

[0055] Where i = 1, 2, 3, ..., N (N is the number of particles), d = 1, 2, 3, ..., D (D is the dimension of the search space); W is the inertia weight factor; C1 and C2 are the social learning factor and the individual learning factor respectively; r1 and r2 are random numbers between (0, 1); P i is the optimal solution found by the particle itself, which is the individual extreme value; P g It is the optimal solution found by the entire population so far and is the global extreme value; V represents the particle velocity, and x represents the position of the particle.

[0056] Furthermore, since the inertia weight W affects the convergence speed and search ability of the particle swarm, in the technical solution provided in this application, the inertia weight is optimized by first increasing and then decreasing, and the improved formula is as follows:

[0057]

[0058]

[0059]

[0060] Among them, W min is the minimum value of inertia weight; W max is the maximum value of the inertia weight; W ce is the maximum variable value of W, which is W max and W min The difference between r is the initial value of inertia weight; g i is the current iteration number; g min is the minimum number of iterations; g max is the maximum number of iterations; W will change with the current number of iterations g of the particle i The inertia weight W increases from small to large in the early stage and decreases from large to large in the later stage, and W needs to be kept at W max With W min During the rising phase, W increases to a value greater than W max , then W is equal to W max .

[0061] The larger W is, the faster the particle convergence speed is and the stronger the global search capability is, but the particle's local search capability is reduced, which ultimately leads to a low global convergence accuracy. The smaller W is, the higher the particle convergence accuracy is, but the particle convergence speed is reduced. During the search process, the particle is easily trapped in the local optimum and cannot jump out of the local extreme value. The traditional inertia weight optimization process adopts a linear decreasing optimization method, which can speed up the convergence speed of the particle swarm, but it is easy to ignore the optimal value in the early stage of convergence, which reduces the global search capability. However, the technical solution provided by the present application adopts an optimization method of increasing first and then decreasing. In the early stage, the inertia weight increases from a small value to effectively ensure the search capability of the searched particle, and it is not easy to ignore the local solution. In the later stage, the inertia weight gradually decreases from a large value, which speeds up the particle convergence speed.

[0062] Furthermore, the learning factor is improved. During the optimization process, the particle swarm will adjust its speed and position according to the motion state of itself and its companions so as to quickly obtain the optimal value. In the standard particle swarm algorithm, the update speed of the particle swarm is mainly adjusted by the individual learning factor C1 and the social learning factor C2. However, due to the lack of changes in the learning factor in the standard algorithm, it is easy for particles to fall into the local optimum. Therefore, in the technical solution provided in this application, the learning factor of the particle swarm is improved by a dynamic adjustment method as follows:

[0063]

[0064]

[0065] Among them, C r1 , C r2 They represent the initial values ​​of individual learning factor and social learning factor, respectively, i1 , g i2 Indicates the current iteration number of the particle, g max1 , g max2 represents the maximum number of iterations of the particle, C max1 , C max2 Represents the final value of the particle learning factor

[0066] Furthermore, the fitness function is improved.

[0067] The fitness function can be regarded as the path length from the optimal parking point to the car owner. Ignoring the road congestion and various influencing factors, the fitness function of the distance to the car owner is defined as:

[0068]

[0069] Where: f(xy) is the distance D that the owner travels to the parking spot, D ijrepresents the distance between nodes i and j in the path from the current location to the entrance of the parking area, d ij Represents the distance between node i and node j in the internal path of the parking area.

[0070] In practical applications, the shorter the parking distance of the car owner, the better the fitness function. Since there are many uncertain factors in the driving process of the car owner, the driving distance alone cannot accurately measure the quality of the parking spot. Therefore, the technical solution provided in this application introduces the road congestion coefficient, road quality, and parking quantity to improve formula (11).

[0071] The road congestion coefficient refers to the proportion of the number of cars in the area where the road is located, expressed as:

[0072]

[0073] Where δ is the road congestion coefficient, N is the number of vehicles on the road, d is the width of the road, and S is the length of the road.

[0074] Road quality. Different path planning will result in different road quality. Therefore, we combine the path planning of flat surface and bridge curved surface to improve the path length. Please refer to the following formula:

[0075]

[0076] Among them, s is the actual total length of the path, S is the weighted total length of the path after being affected by this factor, l is the straight line length of the path, c is the curve length of the path, and β is the curve density, which is the ratio of the curve length c of the path to the actual total length s of the path.

[0077] The number of parking lots in a parking area affects the parking quality in the parking area. Therefore, the number of parking lots in the parking area is taken as the influence coefficient and expressed as θ. Please refer to the following formula:

[0078]

[0079] Wherein, n represents the number of cars currently parking in the parking area, M represents the total number of parking spaces in the parking area, and N represents the number of occupied parking spaces in the parking area.

[0080] Combining the above three influencing factors, the initial fitness function is improved as follows:

[0081]

[0082] Among them, D ij represents the node i(x i ,y i ) and node j(xj ,y j ) and the distance between nodes, d ij represents the node i(x i ,y i ) and node j(x j ,y j ). Formula (15) shows that in practical applications, in the process of particle swarm and artificial fish swarm optimization, due to the influence of road congestion coefficient and road quality, even if the initial parking path length is small, the larger road congestion coefficient will make the final fitness function higher than the planning path with a longer initial parking path length.

[0083] S3: searching for optimal parking points in the parking environment model according to the improved particle swarm algorithm, and forming an optimal parking point set T from the optimal parking points.

[0084] In practical applications, a parking area around the car owner is selected, the particle swarm in the area is initialized, the length of the parking path of the hollow point in the area is calculated, and the position and speed of the particles are updated according to equations (6)(7)(8)(9)(10) and (4)(5). The fitness value of the individual particle is calculated again according to the fitness function of equation (15), and the optimal position of the individual particle and the global optimal position are updated according to the fitness value. Then, the nodes of the shortest path found are put into the set T; until all parking areas are searched. The parking spaces in the set T are initialized and the optimal path is planned.

[0085] S4: Introducing the particle swarm inertia weight into the artificial fish swarm to obtain an improved artificial fish swarm algorithm.

[0086] The artificial fish swarm algorithm simulates the foraging process of fish, and constructs the best optimization strategy by foraging, clustering and tail-chasing. Through the optimization process of each fish, the purpose of finding the global optimal value is achieved. In the artificial fish swarm algorithm, the foraging behavior of the fish swarm is sometimes a blind search, which makes the algorithm search efficiency low, and the limitation of the visual domain makes the algorithm easy to fall into the local optimum. In the technical solution provided by the present application, according to the advantages of the particle swarm algorithm, the improved particle swarm algorithm is combined with the artificial fish swarm algorithm, and the dynamic inertia weight W of the particle swarm is introduced into the field of view of the artificial fish swarm, so that the next target search of the fish swarm has direction and target and improves the convergence speed, please refer to the following formula:

[0087] Visual next =Visual i *W (16)

[0088] Among them, Visual next Indicates the next fish school field of view during the iteration of the artificial fish school.i Indicates the current fish school field of view.

[0089] S5: Mixing the improved particle swarm algorithm and artificial fish swarm algorithm to obtain a hybrid algorithm.

[0090] S6: Planning an optimal parking path according to the hybrid algorithm.

[0091] When the hybrid algorithm seeks the optimal path, it first uses the improved particle swarm algorithm to calculate the position, speed and fitness value of each particle in each iteration, records the global optimal value of the optimal path and updates the bulletin board status; then sets the search target of the artificial fish swarm, and the inertia weight changes the field of view of the artificial fish. The artificial fish searches for the optimal value within the field of view and compares different paths. After each action, the artificial fish checks its own state and compares it with the state of the bulletin board. If it is better than the state of the bulletin board, it replaces it with its own state. The size of the inertia weight W determines the change of the particle speed. The gradual increase of W allows the particle swarm to gradually explore a larger area and locate the approximate position of the optimal solution at a faster speed; as W gradually decreases, the particle optimization speed slows down, and the particles begin to search accurately locally. The size of W adopts the strategy of increasing the inertia weight first and then decreasing it. When the algorithm is optimized, the path becomes increasingly complex, the artificial fish are randomly distributed, the distance between the artificial fish individuals is large, and the field of view of the artificial fish individuals gradually increases. It can effectively overcome the influence of local extreme values ​​and converge quickly. As the algorithm iterates, the artificial fish gradually gather together. At this time, the distance between the artificial fish gradually decreases, and the field of view also decreases with the change of W, which improves the accuracy of the algorithm and can achieve global optimization.

[0092] In summary, the present application provides a hybrid parking path planning method based on an improved particle swarm and artificial fish swarm algorithm, including: establishing a parking environment model according to the actual environment of the parking area, and numbering the road section nodes in the parking environment model to obtain a particle swarm, performing algorithm improvement on the inertia weight, learning factor and fitness function of the particle swarm, searching for an optimal parking point in the parking environment model according to the improved particle swarm algorithm, and forming an optimal parking point set T with the optimal parking points, introducing the inertia weight of the particle swarm to an artificial fish swarm to obtain an improved artificial fish swarm algorithm, mixing the improved particle swarm and artificial fish swarm algorithm to obtain a hybrid algorithm, and planning an optimal parking path according to the hybrid algorithm to solve the problem of inaccurate and inconvenient parking path planning caused by the limitations of a single algorithm.

[0093] Similar parts between the embodiments provided in this application can be referenced to each other. The specific implementation methods provided above are only a few examples under the general concept of this application and do not constitute a limitation on the protection scope of this application. For those skilled in the art, any other implementation methods expanded based on the scheme of this application without creative work belong to the protection scope of this application.

Claims

1. A hybrid parking path planning method based on improved particle swarm and artificial fish swarm algorithm is characterized by: include: According to the actual environment of the parking area, a parking environment model is established, and the road section nodes in the parking environment model are numbered to obtain a particle swarm; Improving the algorithm for the inertia weight, learning factor and fitness function of the particle swarm; Searching for optimal parking points in the parking environment model according to the improved particle swarm algorithm, and forming an optimal parking point set T from the optimal parking points; Introducing the particle swarm inertia weight into the artificial fish swarm to obtain an improved artificial fish swarm algorithm; Mixing the improved particle swarm algorithm and artificial fish swarm algorithm to obtain a hybrid algorithm; Planning an optimal parking path according to the hybrid algorithm; The inertia weight improvement formula is: Among them, W min is the minimum value of inertia weight; W max is the maximum value of the inertia weight; W ce is the maximum variable value of W, which is W max and W min The difference between r is the initial value of inertia weight; g i is the current iteration number; g min is the minimum number of iterations; g max is the maximum number of iterations; W will change with the current number of iterations g of the particle i The inertia weight W increases from small to large in the early stage and decreases from large to large in the later stage. W needs to be kept at W max With W min During the rising phase, W increases to a value greater than W max , then W is equal to W max ; The learning factor is improved by a dynamic adjustment method, and the improved formula is: Among them, C1 and C2 represent individual learning factors and social learning factors respectively, and C r1 , C r2 They represent the initial values ​​of individual learning factor and social learning factor, respectively, i1 , g i2 Indicates the current iteration number of the particle, g max1 , g max2 represents the maximum number of iterations of the particle, C max1 , C max2 represents the final value of the particle learning factor; The improvement steps of the fitness function are: Get the initial fitness function: Obtain the road congestion coefficient δ, the actual total length of the path s, the weighted total length of the path affected by road quality S, and the parking quantity influence coefficient θ; The improved fitness function obtained by combining the initial fitness function is: Among them, the function f(xy) is the distance traveled by the owner to the parking spot, D ij represents the node i(x i ,y i ) and node j(x j ,y j ), d ij represents the node i(x i ,y i ) and node j(x j ,y j ) between the two.

2. The method according to claim 1, characterized in that The formula for improving the artificial fish swarm algorithm is: Visual next =Visual i *W Where W is the particle swarm inertia weight, Visual next Indicates the next fish school field of view during the iteration of the artificial fish school. i Indicates the current fish school field of view.

Images