Review of data-driven computational guidance for unmanned aerospace vehicles

Abstract

This paper explores the application of data-driven computational guidance in unmanned aerospace vehicles, emphasizing improving the optimality of guidance strategies through data-driven approaches. Unmanned aerospace vehicles are engineered to execute predetermined missions while adhering to a variety of physical and operational constraints. Both their design and operational strategies prioritize the efficient utilization of onboard resources. Data-driven methods can learn from data to develop well-trained neural networks that uncover underlying guidance patterns. These trained neural networks can rapidly generate optimal outputs in response to inputs with minimal computational cost. This characteristic of data-driven methods is particularly well-suited for guidance applications in scenarios with limited onboard computational resources. This paper reviews the state-of-the-art achievements in data-driven computational guidance. Simultaneously, we categorize these advancements based on the role of neural networks within the guidance system, referring to them as neural-end-to-end computational guidance and neural-assisted fixed-structure guidance, respectively. Additionally, the paper highlights several open problems and potential future research directions.