Benders Decomposition for Large-Scale Prescriptive Evacuations

Authors: Julia Romanski, Pascal Van Hentenryck

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that Benders decomposition provides significant improvements in solution quality in reasonable time: It finds provably optimal solutions to scenarios considered in prior work, closing these instances, and increases the number of evacuees by 10 to 15% on average on more complex flood scenarios. This section presents experimental results for a case study of the evacuation of the Hawkesbury-Nepean (HN) floodplain, which is located near Sydney.
Researcher Affiliation Academia 1Brown University, Providence, RI. 2University of Michigan, Ann Arbor, MI.
Pseudocode Yes Algorithm 1 The two-stage algorithm for clearance time minimization
Open Source Code No The paper does not provide an explicit statement or link indicating that the source code for the methodology described is publicly available.
Open Datasets Yes Experimental results for a case study of the evacuation of the Hawkesbury-Nepean (HN) floodplain, which is located near Sydney. The HN evacuation graph has 80 evacuation nodes, 184 transit nodes, 5 safe nodes, and 580 edges. The Benders decomposition was evaluated on the real-case study from (Even, Pillac, and Van Hentenryck 2015).
Dataset Splits No The paper does not explicitly provide training/test/validation dataset splits, as it focuses on solving an optimization problem for a case study rather than training a machine learning model.
Hardware Specification Yes The algorithms were implemented using JAVA 8 and GUROBI 6.0 and the results were obtained on a 64 bit machine with a 1.4 GHz Intel Core i5 processor and 4 GB of RAM.
Software Dependencies Yes The algorithms were implemented using JAVA 8 and GUROBI 6.0 and the results were obtained on a 64 bit machine with a 1.4 GHz Intel Core i5 processor and 4 GB of RAM.
Experiment Setup Yes We use time horizons of 600 min for scenarios without flooding and 1000 min for scenarios with flooding, discretized into 5 minute time-steps. Each instance was run for one hour, unless the algorithm converged earlier. Each TDP iteration in the initial dichotomic search is solved with a time limit of 300 s. For the Benders decomposition dichotomic search, each run is given for up to 20 iterations, using the tree from the TDP-DS as a seed.