AGV route optimization and real-time scheduling is a research hotspot in the current AGV field. In practice, the main methods adopted by people are:

1. Mathematical programming method

Choosing the best task and best path for AGV can be summarized as a task scheduling problem. Mathematical programming method is a traditional method to solve the optimal solution of scheduling problem. The solution process of this method is actually a process of optimization under resource constraints. The practical methods mainly include integer programming, dynamic programming, and petri methods. In the case of small-scale scheduling, this type of method can get good results, but with the increase in scheduling size, the time spent solving problems increases exponentially, which limits the application of this method in responsible, large-scale real-time route optimization and scheduling .

2. Simulation method

The simulation method simulates the implementation of an AGV scheduling scheme by modeling the actual scheduling environment. Users and researchers can use simulation to test, compare, and monitor certain scheduling schemes to change and select scheduling strategies. The methods used in the practice include discrete event simulation methods, object-oriented simulation methods and 3D simulation technology. There are many softwares that can be used for AGV scheduling simulation. Among them, Lanner Group ’s Witness software can quickly establish simulation models to achieve three-dimensional simulation process Demonstration and analysis of results.

3. Artificial intelligence methods

The artificial intelligence method describes the AGV scheduling process as a process of searching for the optimal solution in a solution set that satisfies the constraints. It uses knowledge representation technology to include human knowledge, while using various search techniques to try to give a satisfactory solution. Specific methods include expert system methods, genetic algorithms, heuristic algorithms, and neural network algorithms. Among them, the expert system method is often used in practice. It abstracts the experience of scheduling experts into scheduling rules that the system can understand and execute, and uses conflict resolution techniques to solve the rule expansion and conflict problems in large-scale AGV scheduling.

Since neural network has the advantages of parallel operation, knowledge distribution storage, and strong adaptability, it becomes a promising method to solve large-scale AGV scheduling problems. At present, the TSP-NP problem is successfully solved by the neural network method. In the solution, the neural network can convert the solution of the combined optimization problem into an energy function of a discrete dynamic system. The optimization problem can be obtained by minimizing the energy function. solution.

Genetic algorithm is an optimization solution method formed by simulating the inheritance and mutation in the evolution process of natural organisms. When solving the optimal scheduling problem of AGV, the genetic algorithm first expresses a certain number of possible scheduling schemes by coding to the appropriate chromosomes, and calculates the fitness of each chromosome (such as the shortest running path), and repeats replication, crossover, and mutation Look for chromosomes with large fitness, that is, the optimal solution to the AGV scheduling problem.

Solving the scheduling problem with a single method often has certain flaws. At present, the fusion of multiple methods to solve the AGV scheduling problem is a research hotspot. For example, the expert system and genetic algorithm are fused to integrate the expert knowledge into the formation of the initial chromosome group to speed up the solution speed and quality.