New policies for the dynamic traveling salesman problem

The dynamic traveling salesman problem is a variant of the classical traveling salesman problem in which service requests arrive in a random fashion. In this paper we propose three new routing policies minimizing the expected total waiting time of all the demands in the system. In particular, we show that two policies are asymptotically optimal in heavy traffic, whereas the third one has a constant factor guarantee of two. We also perform extensive simulations in order to compare the three policies under various arrival rates.