Improved Peel-and-Bound: Methods for Generating Dual Bounds with Multivalued Decision Diagrams

Decision diagrams are an increasingly important tool in cutting-edge solvers for discrete optimization. However, the field of decision is relatively new, and still incorporating library techniques that conventional have had decades to build. We drew inspiration from warm-start technique used address one major challenges faced by diagram based methods. become more useful wider they allowed be, but also costly generate, especially with large numbers variables. In original version this paper, we presented a method peeling off sub-graph previously constructed using it as initial subsequent iterations call peel-and-bound. tested on sequence ordering problem, our results indicate peel-and-bound scheme generates stronger bounds than branch-and-bound same propagators, at significantly less computational cost. extended propose new methods relaxed improve solutions found restricted diagrams, discuss heuristic decisions involved parallelization peel-and-bound, how can be hyper-optimized sequencing problems. Furthermore, test problem traveling salesman time-windows (TSPTW), include updated generalized implementation algorithm capable handling any optimization problem. The show outperforms ddo (a solver) TSPTW. close 15 open benchmark instances