Improving the search of near-optimal solution of vehicle routine problems with parallelable modified algorithms

• Main Authors: Hsin-yun Hsu, 許心芸
• Other Authors: Chieh-Sen Huang
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/ar6nws

碩士 === 國立中山大學 === 應用數學系研究所 === 105 === We consider finding the near-optimized solution of logistic''s vehicle routing problem includes grouping of customers and travelling salesman problem. We try to balance the number of customers for each vehicle. According to k-mean clustering algorithm, we add restrictions of the number of each cluster and conditions of distributing each customer to achieve our target. After that, we want to find a near-shortest route passing through all the customers for each cluster. This problem is a travelling salesman problem, has been proved to be an NP-hard problem that exact solution should be got by exhaustion method or branch and bound method. Therefore, we use Diagonalize Complete Algorithm to construct a feasible Hamiltonian path, and then using 2-opt and Feiring algorithm to get a shorter path. Among these algorithm, if the number of cluster is n, then the computing of $2$-opt algorithm is O( n^2 ) and can not be parallelable, so we consider two different way to modify the algorithm to do parallel computing with MPI. That is, if we use p processes in MPI, then it can reducing the computing to O( n^2/p ).