Autonomous Vehicle Storage And Retrieval Systems Modeling And Some Key Technologies Study

As an alternative of traditional crane-based automated storage and retrieval system(AS/RS),autonomous vehicle storage and retrieval system(AVS/RS) technology exploits the capabilities of autonomous vehicles within high-density storage systems.Comparing with AS/RS and automated guided vehicles systems (AGVS),AVS/RS is a relatively new technology and more flexible for unit load storage and retrieval.The researches are focused on design optimization models for AVS/RS now time.However,the loss of study for the dynamic process leads to difficulty of popularizing AVS/RS.This paper investigates such fields.Firstly,an approach based on Petri Net has been proposed to model the autonomous vehicle storage and retrieval system(AVS/RS) in which several autonomous vehicles running in bi-direction and one single rail.Then,cycle-deadlock control problem is addressed. Finally,optimization of deadlock-free process in AVS/RS is discussed in details.The subjects studied in thesis and several innovations are presented as follows.A notion of dual color is rendered to solve double need for conciseness of modeling and integrity of dynamic process in the system.Based on dual color and CPN(colored Petri net),an approach,DCDT-PN(Dual Colored Dynamic timed Petri Net),to build the dynamic model of AVS/RS is proposed.Moreover,the behaviors of active resources,such as the RGV and lifts in AVS/RS,are analyzed in details.Then,we propose the modular DCDT-PN model of AVS/RS.In the new model,one color is used to describe those systems with similar or redundant logical structures and another color is employed to represent the route of the token in it.For an AVS/RS with several autonomous vehicles running in bi-direction,deadlock control is one of the key issues in the implementation of AVS/RS.The cycle-deadlock is the main type of deadlock in AVS/RS.Based on the DCDT-PN model,a route digraph is built to detect cycle-deadlock in AVS/RS with digraph tools;the necessary and sufficient conditions of deadlock-free are established.Moreover deadlock-free control policies are proposed,the critical state in deadlock free is also identified and FCFS policy is applied to solve it.Finally,a case study is given to validate the policies.After deadlock control is addressed above,deadlock-free optimization should be considered next.So three optimization models are built to describe different deadlock conditions in AVS/RS,that is lift transferring RGV model(LO-Model),deadlock model in one layer(OLDO-Model) and hybrid deadlock model(BDO-Model) in which OLDO-Model linked with LO-Model.Then,the paper studies the algorithms to solve such three optimization models as follows.The improved genetic algorithm(GA) for optimizing OLDO-Model is discussed firstly.In which the infeasible degree is used to represent the degree that each individual violating constrained conditions.Then,based on such infeasible degrees,a heuristic uniform arithmetical crossover operator is constructed.Three typical OLDO-Model cases are provided to experiment.Compared with the GA with penalty function and feasible region,the algorithm proposed in the paper shows its faster convergence speed and good solution precision.Secondly,the paper addresses the methods for optimizing LO-Model.LO-Model is a TSP with two objectives and nonlinear constrained Conditions.In order to optimize it,two methods are given.On the one hand,an improved genetic algorithm(GA) is introduced,and corresponding genetic operators are constructed.A case is given to test the three scenes of different objective:lift runtime,all RGVs waiting time for lifting and their linear weighted sum.The result shows its validity.On the other hand, a modified GA from a popular multiobjectives GA,Non-Dominated Sorting Genetic Algorithms(NSGA-II),is introduced to optimize LO-Model.In which the infeasible degree is used to deal with the constraints,and placed into front rank calculating.The case is provided and shows its validity.Finally,BDO-Model optimization is discussed in detail.BDO-Model is such a hybrid model that there exist TSP and integer programming with non-linear constraints.In order to optimize such model,two techniques are adopted.One is using linear weighted sum to transform the multiobjectives to single objective.Then,based on the improved GA given above two paragraphs,a heuristic crossover operator and a heuristic mutation are given,and the adjustable ratio of them is also rendered. According to test,such methods can get better results than the traditional GAs. Another is using the multiobjectives GA to deal with BDO-Model.A NonF-HNSGA-â…¡,based on NSGA-â…¡,is proposed.In NonF-HNSGA-â…¡,two infeasible degrees are used to describe the constrains of TSP and integer programming respectively,and then put into non-dominated sorting.According to infeasible degree and front rank,the heuristic crossover operator and mutation operator,based on the improved GA given above two paragraphs,are given to select one or more objectives in BDO-Model to apply crossover or mutation operating.Moreover,combined with elitism,a rule that feasible individuals are priority to infeasible individuals is adopted, then the offspring population is combined with the current generation population and selection is perform to set the individuals of the next generation.According to the experience,NonF-HNSGA-â…¡shows faster convergence speed and practicability,and most of all,NonF-HNSGA-â…¡can also gain good pareto region in feasible practical interval.