Flapping Wing Aerial Vehicles (FWAVs) hold immense potential for applications such as search-and-rescue missions in complex terrains, environmental monitoring in hazardous areas, and exploration in confined spaces. However, their adoption is hindered by the challenges of autonomous navigation in unknown environments, exacerbated by their limited onboard computational resources and demanding flight dynamics. This work addresses these challenges by presenting a lightweight, vision-based autonomous navigation system weighing 26.0 g, enabling FWAVs to achieve obstacle-avoidance flight at a speed of 9.0 m/s. Central to this system is a novel end-to-end Bi-level Cooperative Policy (BCP) that significantly improves flight efficiency and safety. BCP employs lightweight neural networks for real-time performance and leverages Hierarchical Reinforcement Learning (HRL) for robust and efficient training. Quantitative evaluations show that BCP achieves up to 6.5% shorter path lengths, 11.2% faster task completion time, and improved explainability compared to state-of-the-art reinforcement learning algorithms. Additionally, BCP demonstrates 35.7% more efficient and stable training, reducing computational overhead while maintaining high performance. The system design incorporates optimized lightweight components, including a 4.0 g customized stereo camera, a 6.0 g 3D-printed camera mount, and a 16.0 g onboard computer, all tailored to FWAV applications. Real-flight experiments validate the sim-to-real transferability of the proposed navigation system, demonstrating its readiness for real-world deployment in challenging scenarios. This research advances the practicality of FWAVs, paving the way for their broader adoption in critical missions where compact, agile aerial robots are indispensable.

This paper establishes and analyzes a high-fidelity nonlinear time-periodic dynamic model and the corresponding state observer for flapping vibration suppression of a novel tailless Flapping Wing Micro Air Vehicle (FWMAV), named NPU-Tinybird. Firstly, a complete modeling of NPU-Tinybird is determined, including the aerodynamic model based on the quasi-steady method, the kinematic and dynamic model about the mechanism of flapping and attitude control, combined with the single rigid body dynamic model. Based on this, a linearized longitudinal pitch dynamic cycle-averaged model is obtained and analyzed through the methods of neural network fitting and system identification, preparing for the design of flapping vibration suppression observer. Flapping vibration is an inherent property of the tailless FWMAV, which arises from the influence of time-periodic aerodynamic forces and moments. It can be captured by attitude and position sensors on the plane, which impairs the flight performance and efficiency of flight controller and actuators. To deal with this problem, a novel state observer for flapping vibration suppression is designed. A robust optimal controller based on the linear quadratic theory is also designed to stabilize the closed-loop system. Simulation results are given to verify the performance of the observer, including the closed loop responses combined with robust optimal controller, the comparison of different parameters of observer and the comparison with several classic methods, such as Kalman filter, H-infinity filter and low-pass filter, which prove that the novel observer owns a fairly good suppression effect on flapping vibration and benefits for the improvement of flight performance and control efficiency.