It is normally assumed that goods
stored at some customer location cannot directly be transported to another customer. In other
words, all goods have to either originate from, or end up, at a depot. Thus, problems such as the
dial-a-ride problem are excluded from our consideration, although some authors do refer even to
these as pickup-and-delivery problems. It is also usually assumed that the number of vehicles is
not fixed in advance. The objective function of the VRPPD is to minimise the total distance
travelled by the vehicles, subject to maximum distance and maximum capacity constraints on the
vehicles. We also mention that the VRPPD is NP–hard, being a generalisation of the classical
VRP. It can be formulated as a mixed ILP (see Appendix for a possible formulation), and solved
by exact methods for small problems. Our focus, however, will exclusively be on heuristics.
Within the above assumptions, three important VRPPD models may be distinguished. In the
following, we briefly describe these models, and then present and explain our choice of model.