Low-Noise Multi-Agent Intelligent Navigation for Unmanned Aircraft Systems

Unmanned aircraft systems (UAS) have promising potential for on-demand air transportation modes. Integration of UAS into existing airspace operations faces challenges, one of these is the increased noise level caused by multiple vehicles operating within a limited airspace. Finding solutions to mitigate environmental noise is crucial for the public acceptance and sustainable implementation of UAS operations in urban areas. In this paper, we introduce a decentralized method to guide UAS flights, aiming to minimize the noise impact near the intersection points of multiple flight trajectories. The problem is formulated using a multi-agent Markov decision process framework, and it is solved using the Monte Carlo tree search algorithm. To achieve real-time noise prediction during the navigation, we employ a fast noise prediction method. Additionally, for noise result validation, we use an accurate Gaussian beam tracing method. Both approaches incorporate a high-fidelity sound source model of a generic UAS and a representative urban area. Virtual flight simulations validate the effectiveness of the approach, demonstrating noise reduction compared to the benchmark method without noise consideration. This study provides a solution for low-noise multi-agent intelligent navigation, applicable to various types of UAS operations.