We selected the first problem from each dataset (C101, C201, R101, R201, RC101, and RC201) with 50 and 100 customers to be the test problems so a total of 12 VRPTW problems are used. For each of these problems, three TTRPTW
problems are generated in the following manner. For each customer i in a Solomon’s problem, the distance between i and its nearest neighbor customer is calculated and denoted by Ai. In the first TTRPTW problem, 25% of the customers with the smallest Ai values are specified as truck customers. This percentage was increased to 50%
and 75% in the second and third TTRPTW problem respectively. See Chao (2002) for a more detailed discussion on the problem generating procedure.