Constant Integrality Gap LP Formulations of Unsplittable Flow on a Path

The Unsplittable Flow Problem on a Path (UFPP) is a core problem in many important settings such as network flows, bandwidth allocation, resource constraint scheduling, and interval packing. We are given a path with capacities on the edges and a set of tasks, each task having a demand, a profit, a source and a destination vertex on the path. The goal is to compute a subset of tasks of maximum profit that does not violate the edge capacities. In practical applications generic approaches such as integer programming (IP) methods are desirable. Unfortunately, no IP-formulation is known for the problem whose LP-relaxation has an integrality gap that is provably constant. For the unweighted case, we show that adding a few constraints to the standard LP of the problem is sufficient to make the integrality gap drop from Ω(n) to O(1). This positively answers an open question in [Chekuri et al., APPROX 2009]. For the general (weighted) case, we present an extended formulation with integrality gap bounded by 7+e. This matches the best known approximation factor for the problem [Bonsma et al., FOCS 2011]. This result exploits crucially a technique for embedding dynamic programs into linear programs. We believe that this method could be useful to strengthen LP-formulations for other problems as well and might eventually speed up computations due to stronger problem formulations.

[1]  A. Frieze,et al.  Approximation algorithms for the m-dimensional 0–1 knapsack problem: Worst-case and probabilistic analyses , 1984 .

[2]  Ronald L. Rardin,et al.  Polyhedral Characterization of Discrete Dynamic Programming , 1990, Oper. Res..

[3]  Sanjeev Khanna,et al.  On approximating rectangle tiling and packing , 1998, SODA '98.

[4]  Reuven Bar-Yehuda,et al.  A unified approach to approximating resource allocation and scheduling , 2000, STOC '00.

[5]  Reuven Bar-Yehuda,et al.  A unified approach to approximating resource allocation and scheduling , 2001, JACM.

[6]  Piotr Berman,et al.  Improved approximation algorithms for rectangle tiling and packing , 2001, SODA '01.

[7]  Yuval Rabani,et al.  Improved Approximation Algorithms for Resource Allocation , 2002, IPCO.

[8]  Alexander Schrijver,et al.  Combinatorial optimization. Polyhedra and efficiency. , 2003 .

[9]  Baruch Schieber,et al.  A quasi-PTAS for unsplittable flow on line graphs , 2006, STOC '06.

[10]  Amit Kumar,et al.  Approximation Algorithms for the Unsplittable Flow Problem , 2002, Algorithmica.

[11]  Chandra Chekuri,et al.  Multicommodity demand flow in a tree and packing integer programs , 2007, TALG.

[12]  Chandra Chekuri,et al.  Unsplittable Flow in Paths and Trees and Column-Restricted Packing Integer Programs , 2009, APPROX-RANDOM.

[13]  Parinya Chalermsook,et al.  Maximum independent set of rectangles , 2009, SODA.

[14]  Ulrich Pferschy,et al.  Resource allocation with time intervals , 2010, Theor. Comput. Sci..

[15]  Mark de Berg,et al.  Algorithms - ESA 2010, 18th Annual European Symposium, Liverpool, UK, September 6-8, 2010. Proceedings, Part I , 2010, ESA.

[16]  Paul S. Bonsma,et al.  A Constant Factor Approximation Algorithm for Unsplittable Flow on Paths , 2011, 2011 IEEE 52nd Annual Symposium on Foundations of Computer Science.

[17]  Marek Chrobak,et al.  Caching Is Hard—Even in the Fault Model , 2012, Algorithmica.

[18]  Sorin C. Popescu,et al.  Lidar Remote Sensing , 2011 .

[19]  Nikhil Bansal,et al.  A logarithmic approximation for unsplittable flow on line graphs , 2014, TALG.