Authors:
Amirali Zarrinmehr
and
Yousef Shafahi
Affiliation:
Sharif University of Technology, Iran, Islamic Republic of
Keyword(s):
Transportation Network Design Problem, Parallel Branch-and-Bound Algorithm, Depth-First-Search, Greedy Algorithm, Super-linear Speedup.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Network Optimization
;
Operational Research
;
OR in Transportation
;
Pattern Recognition
;
Software Engineering
Abstract:
Transportation Network Design Problem (TNDP) aims at selection of a subset of proposed urban projects in budget constraint to minimize the network users’ total travel time. This is a well-known resource-intensive problem in transportation planning literature. Application of parallel computing, as a result, can be useful to address the exact solution of TNDP. This paper is going to investigate how the performance of a parallel Branch-and-Bound (B&B) algorithm with Depth-First-Search (DFS) strategy can be accelerated. The paper suggests assigning greedy solutions to idle processors at the start of the algorithm. A greedy solution, considered in this paper, is a budget-wise feasible selection of projects to which no further project can be added while holding the budget constraint. The paper evaluates the performance of parallel algorithms through the theoretical speedup and efficiency which are based on the number of parallel B&B iterations. It is observed, in four cases of TNDP in Siou
x-Falls transportation network, that achieving high-quality solutions by idle processors can notably improve the performance of parallel DFS B&B algorithm. In all four cases, super-linear speedups are reported.
(More)