4 Presentations

• 
  02:45 PM - 03:07 PM

  CP-SAT, and high performance constraint programming and integer programming parallel solver on top of a SAT solver

  • Vitor Sessak, presenter, Google
  • Laurent Perrron, Google

  OR-Tools CP-SAT is a state of the art discrete optimization solver that combines SAT, MaxSAT, Constraint Programming, Integer Programming, Local Search and Large Neighborhood Search techniques in a single package.

  In this presentation, we will explain the architecture of this portfolio SAT based solver and showcase its versatility and performance. The key aspects are (a) the deep integration of the SAT engine with the Constraint Programming and the Integer Linear Programming solvers, and (b) the innovative parallel portfolio implementation that orchestrates multiple heterogeneous workers (full solvers, first solution heuristics, local improvement heuristics, dual oriented solvers).

  The combination offers enables fast primal solutions, good dual bounds, robust solving. This makes CP-SAT a solver of choice for solving demanding discrete optimization problems.

• 
  03:07 PM - 03:29 PM

  Generating Alternative Solutions with Benders Decomposition

  • Matthew Viens, presenter, University of Wisconsin-Madison/Sandia National Laboratory
  • William Hart, Sandia National Laboratory
  • Michael Ferris, University of Wisconsin-Madison

  Optimization problems can have multiple solutions that achieve an optimal or near-optimal objective value. We show how to identify multiple solutions for problems solved by Benders Decomposition. We provide a theoretical foundation that describes the iterative refinement of the approximate geometry used in Benders Decomposition, and we describe how this geometry defines sets of multiple solutions. We discuss how this theory can be applied to generate multiple solutions in two problem classes: stochastic programming and interdiction. Further, we demonstrate how multiple solutions can provide additional insights into solution structure and tradeoffs for both problems.

• 
  03:29 PM - 03:51 PM

  Solving MINLPs to global optimality with FICO Xpress

  • Tristan Gally, presenter, FICO

  This talk discusses the global optimization capability within FICO Xpress Solver, which allows to solve general mixed-integer nonlinear problems to global optimality. We will discuss the internal workings of the solver, its features, and recent performance improvements.

• 
  03:51 PM - 04:13 PM

  MaxiCP: A Constraint Programming Solver for Routing and Scheduling

  • Pierre Schaus, presenter, ucl
  • Pascal Van Hentenryck, Georgia Tech
  • Laurent Michel, University of Connecticut
  • Guillaume Derval, ULiège

  We present MaxiCP, an open-source constraint programming solver derived from MiniCP educational solver. MaxiCP offers unique features, including sequence variables for solving vehicle routing problems. It also provides an implementation of the conditional task intervals modeling paradigm for scheduling problems. The solver is designed to support the implementation of custom complex search strategies and enables parallel search execution.