This repository delves into the world of UAV routing considering multiple objectives by proposing a fuzzy extension to regular a priori preference settings that converges towards the initially defined preference setting. the problem characetristics specificly complicating this problem is that the decision maker cannot be implemented in-operation. This means the preferences need to be implemented prior to the operational execution - that is, preferences have to be implemented a priori and how do we do this in the best possible way?
We do this by investigating the trade-offs one can obtain from implementing preferences a priori through a regular elitist GA and from a posterior through NSGA-II.
One would have thought that for large scale NP-hard MOO the knowledge of preferences prior to the solution search would ensure that the a priori at least converges faster and towards a better solution relatively to the a posterior NSGA-II approach that identifies a large set of solutions with a large spread around the entire objective space.
But, we find that despite the need for exploring the entire pareto front, the NSGA-II benefits from its inherent diversity in the populations.
Consequently, allowing the fuzzy a priori preferences to initially yield the search of a fuzzy pareto front will allow for a faster convergence on average than the two approaches.
Here is a problem scenario:
and here is two pareto fronts where the a priori and a posterior are compared:
