36Kr Exclusive: Shanghai Jiao Tong University PhD Student Secures Three Rounds of Funding in Three Months to Enable Embodied Intelligent Bionic Flapping-Wing Robots to Understand and Master Fluids 36氪首发 | 三个月融三轮，上交大00后博士让具身智能仿生扑翼机器人理解并驾驭流体

TL;DR

• Eagle Eye Wings completed its third funding round in three months, raising tens of millions of RMB led by Yuanhe Puhua to accelerate commercialization and R&D.
• The company distinguishes itself through a proprietary fluid simulation engine, Vortrix, which enables reinforcement learning-based control for complex aerodynamic modeling.
• Product strategy splits into a consumer-grade gliding drone ("Eagle X") launching on Kickstarter and an industrial-grade 15-DOF biomimetic robot capable of active flight control.
• The core technological breakthrough lies in zero-shot transfer from virtual simulation to physical prototypes, bypassing costly and slow traditional wind tunnel testing.
• Investors view the sector as a high-ceiling blue ocean compared to crowded ground robotics, leveraging low noise and safety advantages for near-human scenarios.

Why It Matters

This development highlights a critical shift in embodied AI from passive mechanical execution to active environmental interaction, specifically in non-linear aerodynamic domains. For researchers and practitioners, it demonstrates the viability of using high-fidelity simulation engines combined with reinforcement learning to solve complex physical control problems that were previously intractable without extensive real-world trial-and-error.

Technical Details

• Vortrix Fluid Simulation Engine: A proprietary engine using particle methods calibrated with real wind tunnel data to model non-steady, non-linear aerodynamics, enabling rapid reinforcement learning training.
• Reinforcement Learning & Zero-Shot Transfer: The team trains agents in simulation for millions of iterations, allowing direct deployment to physical hardware without further tuning, significantly reducing development cycles.
• High-Degree-of-Freedom Mechanism: The industrial prototype features approximately 15 degrees of freedom, allowing independent wing adjustment (angle, twist, fold) to actively manipulate airflow rather than just passively gliding.
• Data Flywheel: Real-flight data collected by deployed units is fed back into the simulation engine to continuously correct and align the virtual model with physical reality, creating a self-improving loop.

Industry Insight

• Market Differentiation: With ground-based embodied AI becoming saturated, aerial biomimetic robots offer a distinct competitive advantage in noise-sensitive and safety-critical "near-human" environments.
• Platform Potential: The consumer product's open architecture aims to build a developer ecosystem early, similar to Tesla’s approach, turning hardware into a platform for community-driven innovation and data collection.
• Strategic Roadmap: The transition from a closed-loop proprietary engine to a potential open-source or subscription-based base for the broader low-altitude economy suggests a long-term play on becoming infrastructure for aerial robotics.