Multi-condition multi-objective optimization using deep reinforcement learning

A multi-condition multi-objective optimization method that can find Pareto front over a defined condition space is developed for the first time using deep reinforcement learning. Unlike conventional methods which perform at single condition, present learns correlations between conditions and optimal solutions. The exclusive capability of examined in solutions novel modified Kursawe benchmark problem an airfoil shape include nonlinear characteristics are difficult to resolve methods. with high resolution successfully determined each problem. Compared multiple operations single-condition conditions, based on learning shows greatly accelerated search by reducing number required function evaluations. An analysis aerodynamics performance airfoils optimally designed shapes confirms indispensable avoid significant degradation target varying flow conditions.