The Vehicle Routing Problem with Backhauls and Time Windows:

This data set contains 15 problems. The problems were created by 
S. Gelinas, M. Desrochers, J. Desrosiers and M.M. Solomon and first
used in the paper:

[GDDS] S. Gelinas, M. Desrochers, J. Desrosiers and M.M. Solomon, "A new
branching strategy for time constrained routing problems with application 
to backhauling", Annals of Operations Research 61 (1995), 91-109.

The problems were created by modifying Solomons VRPTW problems.

Each data file starts with a header that looks like:
-----------------------------------------------------
 X DISTANCE      Y DISTANCE    NUMBER OF NODES
  100.00            90.00       100
 DEPOT X           DEPOT  Y
   35.00              35.00
-----------------------------------------------------
The "X DISTANCE" and "Y DISTANCE" numbers defines the bounding box of the 
locations, these numbers can safely be ignored. The "NUMBER OF NODES" 
field tells us how many customers the problem contains. "DEPOT X" and 
"DEPOT Y" defines the depot location. Next follows one line per customer.
Each line contains 8 numbers: The first number is the customer id. The 
second number indicates if the customer is a linehaul (0) or a backhaul (1)
customer. The two next numbers define the customer location (x and y 
coordinate). The fifth number indicates the amount of goods to be 
delivered to linehaul customers and picked up at backhaul customers.
The two next numbers define the time window of the customer and the
final number is the service time associated with the customer.

Notice that the file does not specify the vehicle capacity or the latest
allowable arrival time at the depot. In the experiements performed in:

[RP] S. Ropke, D. Pisinger, "A Unified Heuristic for Vehicle Routing 
Problems with Backhauls", submitted to European Journal of 
Operational Research (2003).

the numbers used in the original Solomon problems where used. That is,
the vehicle capacity were set to 200 and the depot deadline were 
set to 230.

Distances and travel times are calculated by using euclidean 
distances. 

Two objectives have been used in the literature:

1: Minimize the travelled distance. If this objective is used then 
the distances should be truncated to one decimal to make the results
comparable to the results in [GDDS].

2: Minimize the number of vehicles used as first priority and then 
minimize the travelled distance as second priority. There does not seem
to be a standard for how distances and travel times should be handled when 
using this objective. I suggest that double precision calculations are
used when handling distances and travel times, as this is the standard 
in the VRPTW community when dealing with the same type of objective. The
results presented in [RP] were obtained in this way.



Regards,

Stefan Ropke
DIKU, University of Copenhagen, Denmark
email: sropke@diku.dk