You are here

Analytics-Based Decomposition of a Class of Bilevel Problems

Authors: 
Publication Type: 
Refereed Conference Meeting Proceeding
Abstract: 
A new class of multiple-follower bilevel problems has been proposed. In this class of problems, the follower problems are allowed to be nonlinear, do not share constraints or variables, and are only weakly constrained. This allows the leader variables can be partitioned among the followers. This new problem was formalised and compared with existing problems in the literature, and it was seen that approaches currently in use for solving multiple-follower problems are not suitable for this new problem. Evolutionary algorithms can be used to solve this problem; however, these are computationally-intensive approaches which do not scale up efficiently. An analytics-based approach was therefore proposed in order to address this issue. Two example problems were solved using the decomposition approach, as well as two evolutionary algorithms. The decomposition approach is particularly useful for large-scale problems; it was seen that time as the size of the bilevel problem got larger, the decomposition approach produced much better results in a shorter amount of time.
Conference Name: 
6th World Congress on Global Optimization
Digital Object Identifer (DOI): 
10.1007/978-3-030-21803-4_62
Publication Date: 
14/03/2019
Conference Location: 
France
Institution: 
National University of Ireland, Cork (UCC)
Open access repository: 
Yes
Publication document: