Average Point Pursuit using the Greedy Algorithm: Theory and Applications
Abstract
This paper considers a discrete-time decision problem wherein a decision maker has to track, on average, a sequence of inputs selected from a convex set by choosing actions from a possibly non-convex feasible set , where is in fact the convex hull of . We study some generalized variants of this problem, in which: (i) and vary with time, and (ii) there might be a delay between them, in the sense that is the convex hull of the previous . We investigate the conditions under which the greedy algorithm that minimizes, in an online fashion, the accumulated error between the sequence of inputs and decisions, is able to track the average input asymptotically. Essentially, this comes down to proving that the accumulated error, whose evolution is governed by a non-linear dynamical system, remains within a bounded invariant set. Applications include control of discrete devices using continuous setpoints; control of highly uncertain devices with some information delay; and digital printing, scheduling, and assignment problems.
Cite
@article{arxiv.1811.07734,
title = {Average Point Pursuit using the Greedy Algorithm: Theory and Applications},
author = {Andrey Bernstein and Niek J. Bouman},
journal= {arXiv preprint arXiv:1811.07734},
year = {2018}
}
Comments
arXiv admin note: text overlap with arXiv:1612.07287