newpage
Faculty of Business, Economics and Statistics

Department of Business Administration

Chair of Production and Operations Management
(Lehrstuhl für Produktion und Logistik)

Consistent Vehicle Routing Problem

   

Extended benchmark instances

The instances are extensions of the ConVRP benchmark dataset of Groër et al. (2009). Those are available at http://www.rhsmith.umd.edu/faculty/bgolden/vrp_data.htm. The modifications to the original dataset only concern the visit frequency. So, the probabilities of a customer to be visited on each of the 5 days were set to p=0.5 and p=0.9, respectively. The instances are structured as follows: The header gives instance parameters like number of customers, planning horizon, vehicle capacity and maximum travel distance. The NODE_COORD_SECTION gives the coordinates for all customers. The DEMAND_SECTION gives the demand of each customer on each day. Service times are listed in the SVC_TIME_SECTION. Finally, the depot coordinates are given in the DEPOT_SECTION.

   
ConVRP_extended_benchmark.zip
   
   

Big instances

The big instances are obtained by adapting Gehring & Homberger's (1999) benchmark instances for the vehicle routing problem with time windows (see http://www.sintef.no/Projectweb/TOP/Problems/VRPTW/Homberger-benchmark). To turn the single period instances into periodic ones a customer request was assigned with p=0.5 over a planning horizon of 25 days (= five consecutive weeks with five days each). The demands and service times were chosen randomly according to a Poisson distribution with the mean equal to the demands and service times in the VRPTW instances. For the experiments with the big instances no maximum arrival time differences were set. The instances are structured as described above.

   
ConVRP_big.zip
   
   
    Last update: May 20, 2011