Study of variable neighborhood descent and tabu search algorithm in VRPSDP

Abstract

Abstract The application of graph theory can be used to solve distribution optimization problems, especially the Vehicle Routing Problems (VRP). One variant of VRP is the Vehicle Routing Problem with Simultaneous Delivery and Pickup (VRPSDP), which has particular constraints, namely requests, and returns, which are done simultaneously. The VRPSDP problems can be solved by a variety of algorithms, such as the tabu search algorithm and the Variable Neighborhood Descent (VND) algorithm. The basic principle of the VND algorithm is to determine the initial solution and repair the solution using the neighborhood operator. In this article, the initial solution is determined by the insertion heuristic algorithm and makes improvements using six neighborhood operators. The optimum condition of the VND algorithm is reached if there is the most optimal solution for all operators, and the solution has converged to a minimum value. The application of the VND algorithm on VRPSDP was made using Borland Delphi 7 software. The VND algorithm application program was tested using standardized data sets with 100 and 200 customers, then compared multiple routes, solutions, and gaps (%). For many customers, 100 and 200 on the two types of data sets, there are many more VND algorithm routes than the tabu search algorithm. The results of the distance solution for the four data sets show that the tabu search algorithm is smaller than the VND algorithm. The type c data set has a smaller gap compared to the type r data set for both 100 and 200 customers to test tabu search algorithms and VND algorithms.