Intelligent path planning and optimisation for real-time cooperative flight

Abstract

Unmanned aerial vehicles (UAVs) have seen an increase in military and commercial use over the past two decades. Whilst the majority of UAVs remain remotely operated, there is a desire amongst many for highly-autonomous aircraft capable of performing solo or cooperative missions with little human interaction. This work presents a novel path planning framework using an evolutionary algorithm to train a deep neural network for generating dynamically-achievable, continuously-flyable, and manoeuvreable paths for a large-scale fixed-wing military autonomous combat aircraft during three-dimensional cooperative flight missions. This autonomous flight control system has been demonstrated to provide robust and reliable path generation during solo and cooperative flight missions with online path planning. The robustness of the solution is demonstrated on a computational model representative of the F-16 Fighting Falcon during both static and dynamic way-point flight simulations, with online path re-planning. The computational model utilises a novel flight control system governs rotational motion of the aircraft, utilising multiple hyper-surfaces of optimised controller gains to ensure optimal dynamic performance and closed-loop stability throughout the flight envelope.