Abstract: To improve the efficiency of sequential Simulated Annealing(SA)for solving Time Dependent Vehicle Routing Problem(TDVRP),a parallel-Simulated Annealing(p-SA)algorithm was proposed.Push Forward Insertion Heuristic(PFIH)was implemented to generate an initial solution,and four kinds of local search moves within the structure of master-slave p-SA were developed to optimize the initial solution.Computational results were reported for 56 test problems with 100 customers from Figliozzi's benchmark,and the results showed that p-SA could get 5 times speedup when 6 threads was used by comparing with traditional sequential SA,and could get better solutions than Figliozzi's algorithm within an accepted computational time.Therefore,p-SA could solve TDVRP effectively and could be extended to handle other variants of vehicle routing problems and other combinatorial optimization problems.

Key words: vehicle routing, time dependent speeds, parallel algorithm, simulated annealing

摘要： 为提高传统串行模拟退火算法求解时间依赖型车辆路径问题的效率,提出一种并行模拟退火算法。该算法首先使用前向插入启发式算法生成初始解,在主从式并行模拟退火算法框架下使用4种邻域搜索法对初始解进行优化。采用Figliozzi测试数据库(包含56个测试问题,顾客数均设定为100)对算法性能进行测试,结果表明在不同时间依赖型行驶函数情形下,当使用6个线程时,并行模拟退火算法相对于传统串行模拟退火算法可以得到近似于5倍的加速比,且均能在较快时间内得到比Figliozzi算法更优的解。因此,并行模拟退火算法能有效地求解时间依赖型车辆路径问题,并且可以灵活地扩展解决其他车辆路径问题和组合优化问题。

关键词: 车辆路径, 时间依赖型, 并行算法, 模拟退火