Optimal Robot Configuration And Dynamic Path Planning In Intelligent Warehouse System

The development of the Internet has greatly promoted the popularity of e-commerce,which has brought great challenges for e-commerce logistics.Although China's logistics industry has developed rapidly,it still cannot meet the actual needs,and there is still a certain gap with the logistics industry in developed countries in terms of logistics intelligence,information level and human cost.As an important link in e-commerce logistics,warehousing affects the efficiency of the whole logistics process.How to improve the efficiency of the system and reduce the cost of the system is the key problem to be solved in the storage system.In recent years,the intelligent storage system represented by Amazon's‘Kiva' system has rapidly become a new storage mode favored by e-commerce warehouses due to its intelligent operation mode.This kind of system uses robots as the medium of picking up and transporting goods,and for the first time realizes the cargo transportation mode of "goods to people".The warehouse information and operation process all adopt the digital management method to realize the unified deployment of the system.As an important part of the system,quantity allocation of robot and path planning algorithm have a great impact on the system efficiency and cost.In the problem of quantity optimal allocation,it is difficult to establish proper warehouse model and objective function.In the path planning problem,it is necessary to use the dynamic programming algorithm to deal with the urgent order problem,which puts forward higher requirements for the efficiency of the algorithm.This thesis studies the above problems.Firstly,it introduces the research background.Then,according to the research literature content,it summarizes the relevant research on storage allocation and dynamic path planning at home and abroad.Aiming at the optimization of robot configuration,a two-stage SOQN network was used to model the system in order to better simulate the process of order matching with robot.The steady-state probability of the system is solved by state transition diagram and global equilibrium equation.The average waiting time of a single order in the system can be obtained by steady-state probability,including the waiting time in the external queue and the waiting time at the sorting platform.Based on this,the objective function is designed.Finally,the accuracy of the model and the solution of the optimal robot configuration are verified in the simulation.The experiments show that the theoretical analysis model in this thesis can simulate the real scene well and guide the warehouse to configure the number of robots.In order to solve the problem of robot dynamic path planning,a grid road network model is firstly established according to the system characteristics,and then a hierarchical programming method is adopted.At the bottom,a single agent Q-Learning algorithm is used to carry out static path planning to avoid the static obstacles in the map.The top layer is the dynamic path planning algorithm,and the obstacle avoidance strategy based on conflict classification is adopted.For non-reciprocal conflicts,stop and wait obstacle avoidance is carried out.For reciprocal conflicts,dual-agent Q-Learning algorithm combined with the protected area mechanism is adopted to avoid.The simulation results show that the algorithm in this thesis can improve the system efficiency to a certain extent,and the protection area mechanism can make a great contribution to reducing the robot's driving distance.