This paper presents a special type of VRP consisting of CVRP and HVRP. The objective is to minimize the
fleet cost and to maximize the capacity utilization. A heuristic based on the nearest neighbor is presented to
generate of initial solutions named NNBH. The results show that NNBH is very efficient to find the optimal
solution for small-scale problems. The SA algorithm is used for solving the proposed model in large-scale
problems and acceptable results are obtained. The CPU time of optimal solution is very sensitive to V and
c parameters, in which the CPU time of SA solution is not sensitive to them, expect to the number of nodes