Program Trace Optimization with Constructive Heuristics for Combinatorial Problems

Abstract

Program Trace Optimisation (PTO), a highly general optimisation framework, is applied to a range of combinatorial optimisation (COP) problems. It effectively combines "smart" problem-specific constructive heuristics and problem-agnostic metaheuristic search, automatically and implicitly designing problem-appropriate search operators. A weakness is identified in PTO's operators when applied in conjunction with smart heuristics on COP problems, and an improved method is introduced to address this. To facilitate the comparison of this new method with the original, across problems, a common format for PTO heuristics (known as generators) is demonstrated, mimicking GRASP. This also facilitates comparison of the degree of greediness (the GRASP α parameter) in the heuristics. Experiments across problems show that the novel operators consistently outperform the original without any loss of generality or cost in CPU time; hill-climbing is a sufficient metaheuristic; and intermediate levels of greediness are usually best.

A review on developing optimization techniques in civil engineering

“…As a result, scholars have developed more advanced heuristic methods, particularly metaheuristic methods, to improve optimization results. Problem-agnostic metaheuristic techniques (McDermott and Moraglio, 2019) employ trade-off randomization for shuffling local in global inquiry. With this kind of optimization approach, optimization research has been significant in recent decades.…”