一种基于神经网络的仿生无人机系统设计

A Bio-inspired Drone Design Based on Neural Network

无人机自主安全着陆是国际无人机导航领域的研究热点与难点。基于卷积神经网络的视觉环境感知建模技术以及仿生着陆装置,能突破传统的无人机依赖平整地面标识的着陆技术,因此设计了一种新型仿生着陆机械系统相结合的爪式着陆系统,使得无人机能通过借助机载视觉传感器构建着陆环境并依靠仿生缓冲与夹持系统实现稳定着陆。通过实际测试验证了该系统设计能够辅助无人机在非结构化地面环境中着陆。

Hazardous terrain landing of unmanned aerial vehicles is desirable, and is a hot topic in the field of intelligent navigation. A crucial point is the landing gears autonomous adaptation to the rough surface, which is especially difficult in unknown and constrained environment. To enable this capability, a convolutional neural network model including the visual landing controller is designed to reject uncertainty disturbances as well as implement a bio-inspired guidance strategy for soft landing in hazardous terrain. The outcome can improve the competitive ability of the automation control system of unmanned aerial vehicles.