Parallel Beam Search Algorithms for Domain-Independent Dynamic Programming
DOI:
https://doi.org/10.1609/aaai.v38i18.30062Keywords:
SO: Heuristic Search, SO: Combinatorial OptimizationAbstract
Domain-independent dynamic programming (DIDP), a model-based paradigm based on dynamic programming, has shown promising performance on multiple combinatorial optimization problems compared with mixed integer programming (MIP) and constraint programming (CP). The current DIDP solvers are based on heuristic search, and the state-of-the-art solver, complete anytime beam search (CABS), uses beam search. However, the current DIDP solvers cannot utilize multiple threads, unlike state-of-the-art MIP and CP solvers. In this paper, we propose three parallel beam search algorithms and develop multi-thread implementations of CABS. With 32 threads, our multi-thread DIDP solvers achieve 9 to 39 times speedup on average and significant performance improvement over the sequential solver, finding the new best solutions for two instances of the traveling salesperson problem with time windows. In addition, our solvers outperform multi-thread MIP and CP solvers in four of the six combinatorial optimization problems evaluated.Downloads
Published
2024-03-24
How to Cite
Kuroiwa, R., & Beck, J. C. (2024). Parallel Beam Search Algorithms for Domain-Independent Dynamic Programming. Proceedings of the AAAI Conference on Artificial Intelligence, 38(18), 20743-20750. https://doi.org/10.1609/aaai.v38i18.30062
Issue
Section
AAAI Technical Track on Search and Optimization