Traditional flight controllers mainly solve the problem of stable flight. FlyCore further complements perception, planning, and task-decision capabilities.

In recent years, the low-altitude economy has swept across industries, and UAVs are gradually evolving from "flying cameras" into intelligent terminals for autonomous operations. Yet in more complex industrial applications, one issue has become increasingly prominent: traditional flight controllers mainly handle low-level flight-control tasks, while their support for high-computing perception, AI recognition, and complex mission planning is limited.

As the number of sensors increases, AI algorithms iterate rapidly, and mission scenarios become more complex, the microcontroller-based architecture of traditional flight controllers has inherent limits when processing high-computing perception and intelligent missions. The FlyCore UAV fusion-control system recently launched by AMOV Lab is designed for this demand. By integrating perception, positioning, planning, and control capabilities, it helps UAVs move from "stable flight" toward "autonomous operation."

01

Flight Control 1.0 vs. Fusion Control 2.0

To understand the value of FlyCore, it is first necessary to understand a key distinction.

A traditional flight-control system is like the brainstem of a human body. It is responsible for basic actions such as attitude calculation, sensor fusion, and motor control, ensuring that the UAV flies steadily. But for image processing, AI model operation, complex environment understanding, and mission planning, traditional flight controllers usually need to rely on an additional onboard computing unit.

AMOV Lab regards UAVs with intelligent application capabilities as an important direction for the "2.0 era." In this new stage, UAVs need autonomous perception, environmental understanding, and mission-planning capabilities. Flight Control 2.0 is not a simple upgrade of Flight Control 1.0, but a shift in system-architecture thinking. At the core of this shift is the fusion-control system, which more tightly integrates intelligent computing with flight control.

02

Triple Heterogeneous Computing: A Layered Architecture with Clear Division of Labor

The core design of FlyCore lies in its hardware-level triple heterogeneous computing architecture. It breaks down the model in which a single processor in a traditional flight controller handles many types of tasks, and separates them into three onboard computing units with distinct responsibilities. Through layered computing, it meets different requirements for real-time response and computing power.

Perception Layer

The mapping and localization unit based on the Intel Ultra platform processes visual and LiDAR data and runs perception algorithms such as mapping, localization, and SLAM, giving the UAV more stable environmental-perception capabilities.

Decision Layer

The recognition and planning unit integrated with the RK3588 chip serves as FlyCore's "AI brain." It runs the SpireCV visual-recognition algorithm and the Prometheus planning-control algorithm, and is responsible for upper-level intelligent tasks such as target detection, path planning, and mission allocation.

Execution Layer

The flight-control unit built around a microcontroller focuses on the most essential flight-control tasks, ensuring the real-time performance of attitude calculation and motor control. It is the foundation for stable UAV flight.

The three layers are not simply stacked together; they work collaboratively. The perception layer handles mapping and localization, the decision layer handles recognition and planning, and the execution layer handles flight control, forming a relatively clear and stable layered architecture.

03

Computing Power, Perception, and Synchronization in One System

Behind this architecture are several engineering-oriented technical designs.

Decoupled Computing Power and Layered Deployment

Traditional flight-control systems face a practical contradiction: low-level flight control requires millisecond-level response, while AI vision tasks often consume more computing power and take longer to process. FlyCore decouples computing power by layer, keeping highly real-time tasks in the execution layer and deploying high-computing tasks in the perception and decision layers. This both safeguards flight safety and reserves more room for AI applications.

Deep Multi-Sensor Fusion

At the system level, FlyCore integrates a perception-localization computer, a recognition-planning computer, a flight-control computer, vision and LiDAR sensors, high-precision GNSS and inertial navigation systems, and video/data transmission links. These units cover the major hardware modules required for UAV recognition, perception, planning, control, and AI applications.

Sensor Clock Synchronization

This is an easily overlooked but critical technical detail. Clock synchronization among vision sensors, LiDAR, IMU, and GNSS aligns multi-sensor data as closely as possible to the same time base, thereby improving the stability of mapping and localization.

Imagine that the position seen by the vision sensor and the attitude measured by the IMU come from different moments in time. The fused localization result may then contain errors. FlyCore uses a hardware-level synchronization mechanism to reduce the impact of such issues on localization results.

04

Research Practicality: Shortening R&D and Integration Cycles

From the perspective of research and industrial applications, FlyCore also offers strong practical value.

Shortening the R&D and Integration Cycle

Traditionally, when OEM manufacturers develop an industrial UAV capable of GNSS-denied flight, intelligent obstacle avoidance, and autonomous operation, the process often takes a long time. FlyCore is designed to be integrated directly by manufacturers. Through pre-optimization in areas such as electromagnetic compatibility, flight-controller vibration damping, and ISP tuning, it can help manufacturers shorten prototype integration and early-stage validation to approximately one month when the necessary conditions are in place.

Providing a Relatively Complete Open-Source Toolchain

With the Prometheus autonomous UAV open-source project at its core, SpireCV visual recognition introduces node-based programming. Developers can split tasks such as recognition, streaming, saving, and control into independent nodes and quickly build vision workflows like assembling building blocks. The supporting lightweight messaging system SpireMS solves data distribution and integration challenges across systems, languages, and environments, making it easier to integrate vision modules into the main mission workflow.

This combination of an open-source ecosystem and standardized product platform enables AMOV Lab to form a technical closed loop covering perception, localization, planning, control, communication, and deployment. What it accumulates is a reproducible, engineerable overall solution, rather than a single algorithm or one-off demonstration.

Multi-Scenario Test Accumulation

Combined with AMOV Lab's technical accumulation in localization and mapping, the UAV can maintain localization and attitude-control capabilities in complex environments such as low light, occlusion, narrow spaces, dense forests, and night operations, and can complete inspection-type missions in GNSS-denied environments. This shows that FlyCore has a foundation for further application and validation in real-world scenarios.

05

Conclusion

Fusion Control Is an Important Direction for UAV Intelligence

If traditional flight controllers mainly solve the problem of making UAVs "fly steadily," then fusion-control systems solve the perception, planning, and mission-execution problems that come after stable flight. The core significance of FlyCore is not that the technical concept itself is entirely new, but that it attempts to answer a key question in UAV intelligence: how to balance controllable flight safety with autonomous mission intelligence.

As applications in the low-altitude economy continue to expand, UAVs are entering more industry scenarios. In the future, those who can more quickly complete the capability integration from low-level flight control to intelligent mission systems will be better positioned to build advantages in the intelligent transformation. AMOV Lab's FlyCore provides researchers, OEM manufacturers, and industry developers with a higher-starting-point infrastructure for intelligent UAV development.