Design And Research Of AGV Path Planning Problem

AGV(Automated Guided Vehicle,AGV)is a core component of the factory intelligent transportation system.It can replace manual handling scenes(such as the transportation of semi-finished products and finished products,storage of finished products,etc.).in the manufacturing floor,and realize the automatic,intelligent and efficient transportation of goods in the factory,thus making the logistics system operate intelligently.Therefore,it is widely used in the production and warehousing systems of automobile manufacturing,3C,heavy industry and other automation industries.With the large number of applications of AGV,enterprises must also consider their investment costs under the premise of guaranteeing the efficiency of production and logistics.Therefore,how to carry out reasonable path planning for AGV has become a research hotspot.The existing AGV path planning algorithms are mainly divided into two major categories: traditional algorithms and intelligent algorithms.Among them,the intelligent algorithm is easier to implement than the traditional algorithm,and its search ability is stronger,so it is widely used.However,existing intelligent algorithms(including genetic algorithms,neural network algorithms,reinforcement learning algorithms,etc.)still have some problems,such as the "premature" problem of the genetic algorithm and the problem of low learning efficiency of the reinforcement learning algorithm.In response to these problems,the paper proposes corresponding improvement methods.The main contents include:1.The path planning of AGV in static environment under the guidance of free path(without magnetic strips,landmarks,etc.)is studied,and a genetic algorithm model is applied to this problem.In the selection of fitness function of genetic algorithm,some restrictive conditions are introduced to optimize the choice of path.Secondly,for the problem of “ premature ” genetic algorithm,the concept of layered inheritance is introduced into the population to increase the diversity of the population and effectively improve the search performance of the algorithm.2.The path planning problem of AGV in dynamic environment guided by free path is studied,and a path planning model based on deep reinforcement learning is designed.In the improved algorithm,the excellent behavior value obtained by the neural network algorithm is taken as the reference behavior of the reinforcement learning,and the problem that the reinforcement learning algorithm randomly selects the behavioral action caused by the low learning efficiency is optimized.In order to verify the effectiveness of the proposed algorithm,this paper carries out a simulation study of the improved intelligent algorithm and the traditional intelligent algorithm,and compares the path planning results and iterative curves respectively.The results show that the improved genetic algorithm has more global search capability than the original algorithm,and its path planning results are better.The improved depth reinforcement learning algorithm is faster than the traditional reinforcement learning algorithm,and effectively shortens the learning time of the optimal path planning.Finally,the main innovations and research results of this paper are summarized.The difficulty of AGV route planning needs to be further solved.The forecasting of AGV scheduling problems is expected.It is expected that it can be gradually solved and perfected in the future research process.