Traveling salesman problem across dense cities

Abstract

Consider~\(n\) nodes~\(\{X_i\}_{1 \leq i \leq n}\) distributed independently across~\(N\) cities contained with the unit square~\(S\) according to a distribution~\(f.\) Each city is modelled as an~\(r_n \times r_n\) square contained within~\(S\) and let~\(TSPC_n\) denote the length of the minimum length cycle containing all the~\(n\) nodes, corresponding to the traveling salesman problem (TSP). We obtain variance estimates for~\(TSPC_n\) and prove that if the cities are well-connected and densely populated in a certain sense, then~\(TSPC_n\) appropriately centred and scaled converges to zero in probability. We also obtain large deviation type estimates for~\(TSPC_n.\) Using the proof techniques, we alternately obtain corresponding results for the length~\(TSP_n\) of the minimum length cycle in the unconstrained case, when the nodes are independently distributed throughout the unit square~\(S.\)