Research On Planning And Task Assignment Of Smart Parking AGV System

This project is based on a smart parking lot in Beijing.It mainly studies the dynamic path planning and static scheduling of multi-AGVs in parking lots.The indoor smart parking system of this article is designed to transform existing parking lots instead of constructing new ones to meet two indicators which are improving the utilization rate(a larger compatibilization rate)and reducing user's waiting time.Therefore,higher requirements are presented on real-time path planning and dynamic task scheduling.At present,some domestic companies have also begun to actively develop AGV smart parking services.Most of its parking lot routes are circular loops.Although the path planning and task scheduling algorithms are simple and easy to operate,AGV can only run on one-way loops,which is not conducive to increasing the compatibilization rate and reducing users' waiting time.Therefore,in the case of satisfying the maximum compatibilization rate(there is no circular route and the AGV can be driven bidirectionally),this paper proposes a new path planning time model,and adopts a static and dynamic combined path planning algorithm for real-time path planning.Multiple AGVs encounter deadlock problems.Aiming at this project,based on the Distributed Constraint Optimization Problem Model(DCOP),a new task assignment model is established,and the DPOP algorithm based on the FRODO framework is used to optimize the solution.The main research contents of this article are as follow:Firstly,the key issues of the AGV unmanned parking system are analyzed in detail,and mathematical models of path planning and task allocation in the AGV system are built.The optimization objectives are the minimum AGV running time and the minimum tardiness penalty cost respectively.At the same time,the topological electronic map model of the parking lot is established which provides the calculation basis for subsequent calculations;Secondly,an improved Dijkstra path planning algorithm is proposed to get the static shortest path for AGV,namely the original optimal path.The parking lot route in this article is regular,but there are many nodes.So the key node method and taboo table are used to reduce the actual computing nodes.This method is connected with the database to get a complete optimal path.Through the comparison with A* algorithm and Dijkstra algorithm,results show that the added taboo list can mask the irrelevant nodes and does not reduce the effective domain of the path search.Although the search results for the paths are same,the improved Dijkstra algorithm is more efficient;Thirdly,this paper uses two methods to model and solve the task assignment of AGVs.The first method builds centralized task scheduling model,which is solved by an improved algorithm combined with ATC rule,beam search and simulated annealing.The second method uses distributed constrained optimization problem model,and the DPOP algorithm based on the FRODO framework.The two methods are compared by an experiment.Its results show that the improved algorithm ATCSA proposed in this paper is better to get a initial solution than the DPOP algorithm also it has a higher optimal solution quality;Finally,this paper designs the simulation system of AGV scheduling system combining with the theoretical research algorithm.Based on the previous theoretical research,this paper uses C++ builder XE to achieve the AGV control simulation system,and mainly focusing on the system's functional modules and database design.