Stochastic And Dynamic Logistics Networks Optimization

The researches on the vehicle routing problems and the facility location problems are very important parts of the logistics networks optimization. These two classes of problems are significant on the economic area and military area. In the classical logistics networks optimization, the researchers mainly focused on the deterministic problems. The problems are very simple, even though they considered the uncertainty elements. In the real life, there exist much uncertainty information and many complex constraints. Traditional models and algorithms show great limitations in solving the stochastic problems and dynamic problems. Therefore, it is necessary to do further research jobs to construct more effective models and algorithms for the stochastic logistics networks optimization and the dynamic logistics networks optimization. In this thesis, a series of stochastic vehicle routing problems and a dynamic location allocation problem are studied thoroughly. The main contributions of this thesis are outlined as follows:Firstly, the research of the vehicle routing problem with stochastic demands and time windows (VRPSDTW) is provided. In this problem, the demands of customers are stochastic and there exists a time window for each customer. A priori optimization strategy is provided for the problem, and a stochastic programming with recourse model is built. The computation of the expected cost of recourse is complex and is discussed in detail. An adpative large neigborhood search heuristic is developed for the solving of problem. The heuristic is tested on the modified Solomon instances. The computational results are compared with the results which are obtained by an alternative solving strategy. The effectiveness of the sub-heuristics is also evaluated.Secondly, the research of the vehicle routing problem with stochastic demands and split deliveries (VRPSDSD) is provided. A paired-vehicles recourse with split delivery strategy is given for the solving problem. The mathematical model is then built. The cooperations between the two paired vehicles and the non-cooperative operations are respectively considered when the expected cost of recourse is computed. A large neigborhood search heuristic is devised, and the heuristic is tested on the modified Solomon instances. The results show that when the average customer demand is larger than the half capacity of vehicle, the savings with the paired-vehicles recourse with split delivery strategy are obvious.Thirdly, the research of vehicle routing problem with stochastic service times (VRPSTT) is provided. In this problem, the customer demand and the travel time are deterministic, but the service times on the customers are stochastic. There also exists a limit on the route duration. The mathematical model is constructed and the expected cost of solution is discussed. The general variable neighborhood search heuristic is provided for the solving. In the experiments, the general variable neighborhood search heuristic is compared with two other heuristics. Also the computations are tested with different parameters. The computational results verify the performance of model and algorithm.Fourthly, a dynamic location allocation problem is studied. With a military background, the supported unit is associated with the multiple periodical demand. The mathematical model is constructed and a hybrid evolutionary algorithm is developed for the problem. The model and algorithm are verified by the experimental results.