A method for dispatching and path planning of AGV vehicles

Abstract

The present invention discloses an AGV trolley scheduling and path planning method, including a given AGV trolley layout and data constraints, a given starting position and a target position of the AGV trolley movement, a given AGV trolley path task, data training based on an algorithm combining a genetic algorithm and a neural network algorithm to obtain its parameters, analyzing the data fed back by program monitoring, and making optimization; this scheme determines the guide line and the work station position, as well as the allowed running direction of the AGV on each path segment by given constraints such as workshop layout, AGV size, AGV performance, and path handling load, so as to optimize a certain performance of the system; this scheme improves the fitness of the genetic algorithm based on a smoothness function on the basis of a traditional genetic algorithm, and solves the problem by using an algorithm formed by mixing the improved genetic algorithm with a deep learning neural network algorithm, so as to achieve accurate, fast, and efficient solutions to the AGV scheduling problem.

Description

Technical Field

[0001] The present invention relates to the field of AGV path planning, and in particular to an AGV trolley scheduling and path planning method. Background Art

[0002] In a multi-AGV system, multiple AGVs are often used to work together to complete logistics handling tasks. If the number is too small, the handling system efficiency is not high, which may affect the normal operation of production and reduce product output efficiency. However, if the number is too large, it will increase the burden of path planning and traffic management, increase the time of material transportation tasks, and also reduce product output efficiency. Multi-AGV system scheduling technology involves many technologies, but researchers mainly focus on the three key technologies of task allocation, path planning and traffic management. In addition to considering the shortest path, multi-vehicle path planning often needs to consider the impact of traffic management waiting time on the total driving time. Therefore, designing an improved genetic algorithm and deep learning neural network algorithm for path planning is an urgent problem that researchers in related fields need to solve. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a technical solution: an AGV car scheduling and path planning method, comprising the following steps:

[0004] (1) Given the layout and data constraints of the AGV;

[0005] (2) Given the starting position and target position of the AGV;

[0006] (3) Given an AGV path task;

[0007] (4) The AGV selects a task and starts executing it. First, the starting position of the AGV is determined. Then, the algorithm based on the genetic algorithm combined with the neural network algorithm is trained to obtain its parameters.

[0008] (5) When the AGV selects a task and starts executing it, the algorithm in step (4) is used to find the shortest directed distance between the AGV and the target position to reflect the minimum AGV task;

[0009] (6) After the task is completed, the data fed back by the program monitoring is analyzed and optimized.

[0010] Furthermore, the implementation of the algorithm combining the genetic algorithm with the neural network algorithm comprises the following steps:

[0011] (4.1) Establish the fitness function f(p): Take the total distance as the target and select the target value, which is the inverse of the path length L(p), as follows:

[0012]

[0013] (4.2) Add the smoothness function s(θ) to the fitness function in step (4.1), and the updated fitness function is:

[0014]

[0015] Among them, α and β are the penalty coefficients of the path length L(p) and the smoothness function s(θ), which restrict the relationship between the path length and the path smoothness:

[0016] S(θ)=ω·θ

[0017] θ<90°,ω→∞

[0018] 90°≤θ<180°,ω→1

[0019] When the turning angle is θ<90°, the AGV is not easy to turn. At this time, ω→∞ makes its overall fitness smaller, reducing the probability of chromosomes continuing to be inherited. When 90°≤θ<180°, the AGV can turn safely, ω→1, and the smoothness of the path can be calculated according to the normal angle.

[0020] Furthermore, the conditions for the completion of the task execution include: determining whether the destination has been reached, and when the AGV reaches the destination, the task is completed; when the AGV does not reach the destination, the next node is selected for movement.

[0021] Furthermore, the condition for the task execution to be completed also includes: determining whether to end the program, when the AGV car ends the program, the task is completed; when the AGV car does not end the program, select to execute another task.

[0022] Furthermore, the layout and data constraints include: given workshop layout, AGV size, AGV performance, and path handling load.

[0023] After adopting the above scheme, the present invention has the following advantages: this scheme determines the guide line and workstation position, as well as the allowed running direction of the AGV on each path segment by given constraints such as workshop layout, AGV size, AGV performance, and path handling load, so as to optimize a certain performance of the system; this scheme improves the fitness of the genetic algorithm based on the smoothness function on the basis of the traditional genetic algorithm, and solves the problem by applying an algorithm formed by mixing the improved genetic algorithm with the deep learning neural network algorithm, so as to achieve accurate, fast and efficient solution to the AGV scheduling problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 It is a flow chart of an AGV trolley scheduling and path planning method of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0027] Example

[0028] Genetic algorithms can modify the initial population generation, simplify the path selection and avoid the appearance of repeated paths. The use of improved genetic algorithms can effectively improve the quality of path solutions. Designing an improved genetic algorithm mixed with a deep learning neural network algorithm can avoid the problem of local bracelets of a single algorithm and improve the calculation accuracy of the path.

[0029] On the basis of the traditional genetic algorithm, the fitness of the genetic algorithm is improved based on the smoothness function, and the problem is solved by a hybrid algorithm of the improved genetic algorithm and the deep learning neural network algorithm, so as to achieve the purpose of accurately, quickly and efficiently solving the AGV scheduling problem.

[0030] The fitness function drives the operation of the genetic algorithm numerically, and the fitness function is generally non-negative. When solving the minimum problem, the reciprocal of the target value is usually selected, which forms an inverse relationship with the target value. That is, the smaller the fitness function, the larger the target value and the better the individual.

[0031] Combined with Figure 1 , an AGV car scheduling and path planning method, comprising the following steps:

[0032] (1) Given the layout and data constraints of the AGV, the layout and data constraints include: given workshop layout, AGV size, AGV performance, and path handling load;

[0033] (2) Given the starting position and target position of the AGV;

[0034] (3) Given an AGV path task;

[0035] (4) The AGV selects a task and starts executing it. First, the starting position of the AGV is determined. Then, the algorithm based on the genetic algorithm combined with the neural network algorithm is trained to obtain its parameters.

[0036] The implementation of the algorithm combining the genetic algorithm and the neural network algorithm includes the following steps:

[0037] (4.1) First, establish the fitness function f(p): Taking the total distance as the target, select the target value, that is, the inverse of the path length L(p) as follows:

[0038]

[0039] in:

[0040]

[0041] (4.2) This setting method is simple and easy to operate. Although it meets the inverse proportional requirement, it is too simple and cannot well reflect some of the details. Therefore, the smoothness function s(θ) is added to the fitness function in step (4.1). The updated fitness function is:

[0042]

[0043] Among them, α and β are the penalty coefficients of the path length L(p) and the smoothness function s(θ), which restrict the relationship between the path length and the path smoothness:

[0044] S(θ)=ω·θ

[0045] θ<90°,ω→∞

[0046] 90°≤θ<180°,ω→1

[0047] When the turning angle is θ<90°, the AGV is not easy to turn. At this time, ω→∞ makes its overall fitness smaller, reducing the probability of chromosomes continuing to be inherited. When 90°≤θ<180°, the AGV can turn safely, ω→1, and the smoothness of the path can be calculated according to the normal angle.

[0048] (5) When the AGV selects a task and starts executing it, the algorithm in step (4) is used to find the shortest directed distance between the AGV and the target position to reflect the minimum AGV task;

[0049] The conditions for the end of task execution include: judging whether the destination has been reached. When the AGV reaches the destination, the task ends; when the AGV does not reach the destination, the next node is selected for movement.

[0050] The conditions for the end of task execution also include: judging whether to end the program. When the AGV car ends the program, the task ends; when the AGV car does not end the program, choose to execute another task.

[0051] (6) After the task is completed, the data fed back by the program monitoring is analyzed and optimized.

[0052] The problem with neural networks is that they ignore the study of training path data. An algorithm that is a mixture of an improved genetic algorithm and a deep learning neural network algorithm is used to solve the problem. Path planning is performed based on the hybrid algorithm in a simulation environment, making the planned path more reliable.

[0053] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. An AGV car scheduling and path planning method, characterized in that: The following steps are involved: (1) Given the layout and data constraints of the AGV; the layout and data constraints include: given workshop layout, AGV size, AGV performance, and path handling load; (2) Given the starting position and target position of the AGV; (3) Given an AGV path task; (4) The AGV selects a task and starts executing it. First, the starting position of the AGV is determined. Then, the algorithm based on the genetic algorithm combined with the neural network algorithm is trained to obtain its parameters. The implementation of the algorithm combining the genetic algorithm with the neural network algorithm comprises the following steps: (4.1) Establish the fitness function f(p): Take the total distance as the target and select the target value, which is the inverse of the path length L(p), as follows: (4.2) Add the smoothness function s(θ) to the fitness function in step (4.1), and the updated fitness function is: Among them, α and β are the penalty coefficients of the path length L(p) and the smoothness function s(θ), which restrict the relationship between the path length and the path smoothness: S(θ)=ω·θ θ<90°,ω→∞ 90°≤θ<180°,ω→1 When the turning angle is θ<90°, the AGV is not easy to turn. At this time, ω→∞ makes its overall fitness smaller, reducing the probability of chromosomes continuing to be inherited. When 90°≤θ<180°, the AGV can turn safely, ω→1, and the smoothness of the path can be calculated according to the normal angle; (5) When the AGV selects a task and starts executing it, the algorithm in step (4) is used to find the shortest directed distance between the AGV and the target position to reflect the minimum AGV task; (6) After the task is completed, the data fed back by the program monitoring is analyzed and optimized; (6.1) The conditions for the completion of the task execution include: determining whether the destination has been reached. When the AGV reaches the destination, the task is completed; when the AGV does not reach the destination, the next node is selected for movement; (6.2) The conditions for the completion of the task execution also include: determining whether to complete the program. When the AGV completes the program, the task is completed. When the AGV does not complete the program, another task is selected for execution.

Images