Rotorcraft low-noise trajectories design: black-box optimization using surrogates

This paper addresses the noise-minimal trajectory optimization problem for a specific type of aircraft: rotorcraft. It relies on realistic noise footprint computation software provided by industry that is black-box. Locally optimal trajectories are computed through tailored solution approach based Mesh-Adaptive Direct Search algorithm. We propose multiple surrogates defined according to our knowledge problem, including surrogate relying physics (approximating rotorcraft model), and another machine learning (neural network) method. The proposed further enhanced an appropriate starting guess path planning algorithm reduction variable space domain. performance methodology both in terms quality solutions (trajectories exhibiting significant compared those currently flown practice) computing time illustrated numerical experiments real-world case studies.