A NOVELTY-BASED APPROACH TO CAPTURING UNIQUE SOLUTIONS IN CONTINUOUS MULTIOBJECTIVE PROBLEMS

Novelty search is used in this study to reduce the half of selection space in U-NSGA-III algorithm. We modify the selection mechanism to consider the novelty of solutions during selection. And in this way, we reduce time and lose important unique solutions because their uniqueness is only exhibited in the changing space (not the objective space) which is not taken into consideration by any of these earlier algorithms. We use U-NSGA-III to implement our algorithm after some modifications to it. The most important modification is introducing additional constraints restricting the feasible region of the problem. The new constraints deem half the originally feasible objective space infeasible, except for a narrow region that - due to the new setup - has become separated from the rest of the feasible space.