[01036] Progress in Mathematical Programming Methods and Applications
Session Date & Time :
01036 (1/2) : 3C (Aug.23, 13:20-15:00)
01036 (2/2) : 3D (Aug.23, 15:30-17:10)
Type : Proposal of Industrial Minisymposium
Abstract : (Mixed-)integer (non-)linear optimization has been one of the biggest successes in transferring mathematical insight into real-world impact. Due to the generality of the integer programming model combined with continuous improvement in solving capability, the list of industrial applications is virtually endless. While the problems are usually NP-hard in theory, in practice, an incredible number of real-world instances can be solved within seconds. The algorithmic progress outpaced the increase in computer performance by far; combined, the solver speed has exponentially grown over the last 40 years. We present the latest state-of-the-art and a glimpse into the future.
Thorsten Koch (Technische Universität Berlin / Zuse Institute Berlin)
Shoji Shimizu (NTT DATA Mathematical Systems Inc.)
Robert Luce (Gurobi GmbH)
Timo Berthold (Fair Isaac Deutschland GmbH)
Gerald Gamrath (Cardinal Operations)
Junko Hosoda (Hitachi Ltd.)
Jaap Pedersen (Zuse Institute Berlin)
Hiroki Ishikura (Kyushu University)
Talks in Minisymposium :
[02009] Benders' decomposition approach for the integrated long-haul and local VRP
Author(s) :
Junko Hosoda (Hitachi, Ltd.)
Stephen J. Maher (Quantagonia GmbH)
Yuji Shinano (Zuse Institute Berlin )
Christoffer Villumsen (Hitachi, Ltd.)
Abstract : A supply chain management problem that integrates the determination of consolidation locations with the coordination of long-haul and local vehicle routing is a complicated problem. A Benders' decomposition approach is used to solve this problem. The delivery area and consolidation locations are computed in the master problem, and the long-haul and local vehicle routes are computed in the subproblems. The effectiveness of the decomposition is discussed in the presentation.
[02259] Techniques and advances for solving MINLPs
Author(s) :
Robert Luce (Gurobi)
Abstract : We consider solving mixed-integer nonlinear optimization problems to global
optimality. Our solver is based on the common branch-and-bound paradigm, but
includes a number of specialized techniques to deal with nonconvex constraints
and nonconvex objective functions. In this talk we will outline a few of these
components from a theoretical and computational point of view.
[02270] Steepest-Edge Simplex Algorithms for Quadratic Programming
Author(s) :
Shoji Shimizu (NTT DATA Mathematical Systems Inc.)
Koichi Fujii (NTT DATA Mathematical Systems Inc.)
Julian Hall (University of Edinburgh)
Abstract : We present steepest-edge simplex algorithms for quadratic programming problems. It is well known that in linear programming problems, the steepest-edge rule or Devex rule greatly reduces the total number of iterations in the simplex method. We extend these rules to the simplex method for quadratic programming and show their effectiveness through numerical experiments.
[02296] New MIP presolving techniques in the Cardinal Optimizer
Author(s) :
Gerald Gamrath (COPT GmbH)
Abstract : Presolving is an essential component of modern MIP solvers. Besides model cleanup, it identifies structures in the problem and tightens the formulation before the branch-and-cut search starts.
In this talk, we discuss common structures in real-world instances and show how a mathematical analysis of those structures resulted in new presolving reductions implemented in the Cardinal Optimizer (COPT). The impact of the new techniques is demonstrated in computational experiments.
[03882] Realization of smart factories using MIP
Author(s) :
Hiroki Ishikura (Kyushu University)
Abstract : Smart factories have become widely used for more efficient production activities recently. In collaboration with Rohto Pharmaceutical Co. (Rohto), we have conducted research to realize smart factories. In this talk, we will introduce mobility optimization related to automated warehouses. Rohto uses an automated warehouse to manage a large volume of various items. By optimizing the mobility of automated warehouses, production activities can be streamlined, and factory operations can be made more efficient.
