Approximation Algorithms for Correlated Knapsack Orienteering

We consider the correlated knapsack orienteering (CorrKO) problem: we are given a travel budget B, processing-time budget W, finite metric space (V,d) with root ρ ∈ V, where each vertex is associated with a job with possibly correlated random size and random reward that become known only when the job completes. Random variables are independent across different vertices. The goal is to compute a ρ-rooted path of length at most B, in a possibly adaptive fashion, that maximizes the reward collected from jobs that processed by time W. To our knowledge, CorrKO has not been considered before, though prior work has considered the uncorrelated problem, stochastic knapsack orienteering, and correlated orienteering, which features only one budget constraint on the sum of travel-time and processing-times. Gupta et al. [Gupta et al., 2015] showed that the uncorrelated version of this problem has a constant-factor adaptivity gap. We show that, perhaps surprisingly and in stark contrast to the uncorrelated problem, the adaptivity gap of CorrKO is is at least Ω(max{√log(B),√(log log(W))}). Complementing this result, we devise non-adaptive algorithms that obtain: (a) O(log log W)-approximation in quasi-polytime; and (b) O(log W)-approximation in polytime. This also establishes that the adaptivity gap for CorrKO is at most O(log log W). We obtain similar guarantees for CorrKO with cancellations, wherein a job can be cancelled before its completion time, foregoing its reward. We show that an α-approximation for CorrKO implies an O(α)-approximation for CorrKO with cancellations. We also consider the special case of CorrKO where job sizes are weighted Bernoulli distributions, and more generally where the distributions are supported on at most two points (2CorrKO). Although weighted Bernoulli distributions suffice to yield an Ω(√{log log B}) adaptivity-gap lower bound for (uncorrelated) stochastic orienteering, we show that they are easy instances for CorrKO. We develop non-adaptive algorithms that achieve O(1)-approximation, in polytime for weighted Bernoulli distributions, and in (n+log B)^O(log W)-time for 2CorrKO. (Thus, our adaptivity-gap lower-bound example, which uses distributions of support-size 3, is tight in terms of support-size of the distributions.) Finally, we leverage our techniques to provide a quasi-polynomial time O(log log B) approximation algorithm for correlated orienteering improving upon the approximation guarantee in [Bansal and Nagarajan, 2015].