“…The SVRP can be broadly classified based on the following three criteria: (1) where the uncertainty is present in the problem, e.g., the presence of the customers (Jézéquel [35], Jaillet [34], Bertsimas [9,10]), the demand level (Bertsimas and Simchi-Levi [13]), or the travel time (Kao [37], Laporte et al [40], Jula et al [36]) and the service time at customer sites (Hadjiconstantinou and Roberts [31]); (2) how the problem is modeled, e.g., by stochastic programming technique (Stewart and Golden [44], Bertsimas [10]), by Markov decision process (Dror and Trudeau [20], Dror [19], Dror et al [21], Secomandi [41]), or by robust optimization methodology (Sungur et al [45]); and (3) how to solve the model, which heavily depends on the modeling method and can be broadly classified into two categories: exact methods (branch and cut, integer L-shaped method (Gendreau et al [26]) and generalized dynamic programming (Carraway et al [16])) and heuristic methods such as saving algorithms (Clarke and Wright [17]), sweep algorithms, genetic algorithms, tabu search (Gendreau et al [28]) to name a few. For a more detailed discussion based on this classification, please refer to Shen et al [43].…”