Polynomial time approximation schemes for the traveling repairman and other minimum latency problems

We give a polynomial time, (1 + ∊)-approximation algorithm for the traveling repairman problem (TRP) in the Euclidean plane, on weighted planar graphs, and on weighted trees. This improves on the known quasi-polynomial time approximation schemes for these problems. The algorithm is based on a simple technique that reduces the TRP to what we call the segmented TSP. Here, we are given numbers l1, …, lK and n1, …, nK and we need to find a path that visits at least nh points within path distance lh from the starting point for all h ∊ {1, …, K}. A solution is α-approximate if at least nh points are visited within distance αlh. It is shown that any algorithm that is α-approximate for every constant K in some metric space, gives an α(1 + ∊)-approximation for the TRP in the same metric space. Subsequently, approximation schemes are given for this segmented TSP problem in different metric spaces. The segmented TSP with only one segment (K = 1) is equivalent to the k-TSP for which a (2 + ∊)-approximation is known for a general metric space. Hence, this approach through the segmented TSP gives new impulse for improving on the 3.59-approximation for TRP in a general metric space. A similar reduction applies to many other minimum latency problems. To illustrate the strength of this approach we apply it to the well-studied scheduling problem of minimizing total weighted completion time under precedence constraints, 1|prec|Σ wjCj, and present a polynomial time approximation scheme for the case of interval order precedence constraints. This improves on the known 3/2-approximation for this problem. Both approximation schemes apply as well if release dates are added to the problem.