ROS2 is steadily becoming an essential foundation for robotics research and engineering development. Its more efficient communication mechanisms, more flexible modular architecture, and continuously expanding hardware and software ecosystem are opening up new possibilities for autonomous perception, navigation and planning, and intelligent control in drones.

Today, the Prometheus 450-ROS2 research drone is officially available.

Developed around the practical needs of university laboratories, research teams, and robotics developers, this new version introduces major upgrades in three areas: system architecture, onboard computing, and environmental perception. It also comes with multiple perception configurations and development resources to help developers conduct algorithm research, system integration, and live-flight validation more efficiently.

From simulation and debugging to real-world flight, and from validating individual algorithms to executing complete missions, the P450-ROS2 further connects the full flight-robotics R&D workflow—helping innovative ideas move from the screen into real flight scenarios faster.

01

Built for the ROS2 Ecosystem: A New-Generation Research Drone Platform

The Prometheus 450 is a flight-robotics platform designed for drone research, teaching, and custom development.

The complete system integrates the flight platform, flight-control system, onboard computing, communication links, and environmental perception devices. It supports research in localization and mapping, path planning, autonomous navigation, target recognition, intelligent control, and related areas.

While retaining the P450's mature flight platform and open development capabilities, the new ROS2 version further upgrades the software framework, computing resources, and perception configurations, making it easier for developers to work within the ROS2 ecosystem.

The platform covers the complete R&D workflow:

Developers no longer need to repeatedly select components, adapt hardware, and build the basic environment from scratch. More time can instead be focused on algorithm innovation and mission development.

02

System Upgrade: Full ROS2 Integration for a Clearer Development Workflow

The P450-ROS2 runs the Prometheus V3 autonomous UAV software system. Compared with the previous architecture, the new system reorganizes its communication and functional modules around ROS2 and introduces XRCE-DDS for bidirectional data transmission with the PX4 flight controller. This provides a stronger foundation for communication responsiveness, system scalability, and future feature integration in scenarios where multiple nodes and sensors operate concurrently.

The control module remains the bridge between software applications and UAV hardware execution. It supports seven control submodes: body-frame position, velocity, combined position-velocity, and trajectory control, together with inertial-frame position, velocity, and combined position-velocity control. Developers can write control logic through ROS2 interfaces, or use PrometheusGroundStation-Pro to view system status, operate the aircraft, and adjust parameters dynamically.

03

Computing Upgrade: 100 TOPS Standard, Up to 157 TOPS in Super Mode

The P450-ROS2 is equipped with the Allspark 2 mini edge computer, powered by an NVIDIA Jetson Orin NX. Standard mode provides 100 TOPS of AI performance, while Super mode reaches up to 157 TOPS. This gives the onboard system substantially more GPU acceleration headroom for real-time tasks such as target detection and tracking, 3D SLAM, depth perception, and path planning.

Usability matters as much as the headline performance figure. Allspark 2 comes with the development environment preinstalled and combines a compact footprint, low weight, and strong system portability. This reduces the preparation required for system flashing, dependency configuration, and functional validation. Together with the 410 mm wheelbase modular airframe, composite-material structure, semi-enclosed protection, and optimized wiring, the platform offers a practical balance between payload expansion and routine maintenance.

04

Perception Upgrade: Connecting Localization, Mapping, Obstacle Avoidance, and Tracking

The new perception chain combines the MID-360S 3D LiDAR with either a D435i depth camera or a G1 electro-optical gimbal. These options cover spatial mapping, near-field depth perception, and target tracking, giving the P450-ROS2 a clear configuration path for different research tasks.

MID-360S: Establishing a Stable Coordinate Frame for 3D Space

The MID-360S provides a 360° × 59° field of view, a point rate of 200,000 points per second, and a 0.1 m near-field blind zone. At 10% reflectivity, its detection range reaches 40 m. Its compact design and built-in IMU make it well suited to onboard deployment. Combined with FAST-LIO 3D LiDAR SLAM, the P450-ROS2 can perform 3D mapping and localization while continuously supplying environmental information to EGO-Swarm path planning.

D435i and G1: Two Clearly Defined Vision Options

The D435i configuration is designed for vision-based obstacle avoidance and depth perception. The camera integrates an Intel RealSense D4 vision processor, depth-sensing capability, and an IMU. It outputs real-time depth information both indoors and in illuminated outdoor environments, providing input for visual obstacle avoidance and spatial understanding.

The G1 configuration is designed for gimbal-based target tracking. Its three-axis stabilized gimbal uses an OV OS12D40 color image sensor, delivering image signals at up to 11.3 megapixels and 60 fps, with support for 4K video recording at 30 fps. Combined with SpireCV-Pro, the system can perform target detection, tracking and lock-on, gimbal control, still-image capture, and video recording.

05

How to Choose Between the ROS 1 and ROS 2 Versions

The P450-ROS1 and P450-ROS2 use the same airframe, flight controller, remote control, video and data link, and MID-360S 3D LiDAR. Their main differences are the onboard computing platform, software architecture, positioning configuration, and support for certain features. For outdoor flight—especially when positioning accuracy and hover stability are priorities—the P450-ROS1 is the recommended choice. It comes standard with a NEO-F9P RTK positioning antenna and can achieve centimeter-level outdoor positioning through network RTK corrections.

For existing ROS 1 projects, course instruction, reproduction of research-paper code, or projects requiring QR-code-guided landing, the ROS 1 version is easier to use directly. For new projects that place greater emphasis on 157 TOPS of onboard performance, XRCE-DDS communication, multi-UAV collaboration, system scalability, and long-term maintenance, the P450-ROS2 is the better fit.

06

Key Features

UAV Control

Working with the control submodule in Prometheus V3, the platform provides precise UAV control. This submodule acts as the bridge between the software application and the aircraft. It offers control states including hover at the initial position, hover at the current position, landing, and movement. Six control submodes cover position, velocity, and combined position-velocity control in both the inertial frame and body frame, with trajectory control forming the seventh mode.

Gimbal Control

With the G1 electro-optical gimbal installed, developers can read data from and control the gimbal through ROS2 interfaces, or use PrometheusGroundStation-Pro for gimbal control, still-image capture, and video recording. Onboard integration with the SpireCV-Pro intelligent perception library enables target detection and tracking, allowing the gimbal to maintain a real-time lock on selected targets.

Target Detection

The system loads a YOLO target-detection algorithm to identify objects in the video stream. Operating instructions are provided so users can quickly replace the supplied model with their own YOLO model and accelerate algorithm validation.

Click-to-Track

YOLO11 can load a designated training dataset or the general-purpose COCO dataset to detect targets, while LiteTrack locks onto and follows the selected object. Indoors, the gimbal can track a trained small-vehicle target; outdoors, it can use person detection from the general-purpose dataset for target tracking.

EGO-Swarm LiDAR/Visual Obstacle Avoidance

The platform supports EGO-Swarm, an upgrade to EGO-Planner, for path planning. The algorithm significantly improves the UAV's autonomy in dynamic environments by analyzing and flexibly avoiding moving obstacles in real time while rapidly adapting to and following dynamically updated goal points. This improves mission efficiency and enables reliable operation in unknown and complex scenarios. It is easy to get started and works with PrometheusGroundStation-Pro for autonomous 3D dynamic path planning, supporting obstacle-avoidance flight at speeds of up to 2.4 m/s.

07

Upgraded Toolchain

Prometheus V3 Software System

Prometheus is an open-source autonomous UAV software platform that provides a complete solution for intelligent and autonomous flight. Building on the capabilities of V2, the project team developed Prometheus V3 with stronger support for the new-generation ROS2 ecosystem and its advantages in real-time communication, system performance, and scalability. XRCE-DDS enables bidirectional data transmission with PX4 at lower resource overhead and higher communication efficiency. Prometheus V3 opens a new stage of UAV development for mission planning, multi-UAV collaboration, intelligent perception, and related applications. The platform will continue to evolve with support for additional hardware platforms and emerging technologies.

PrometheusSim

PrometheusSim is a high-fidelity simulation tool for UAV development, research, and teaching. Built on Unreal Engine, it renders detailed and realistic 3D environments and supports simulation of lighting, shadows, weather, and complex terrain, creating a more immersive experience for algorithm validation, feature debugging, and demonstrations.

Unlike a conventional Ubuntu + Gazebo setup, PrometheusSim uses a coordinated Windows 11 + WSL2 architecture. High-quality simulation scenes run in Windows, while the PX4 flight controller and ROS packages run in WSL2, with close data exchange between the two environments. This balances visual quality, development efficiency, and debugging convenience. A one-click installation script also reduces the time required for environment configuration, allowing users to build a complete UAV simulation environment and begin learning, development, and testing more quickly.

SpireCV-Pro Intelligent Perception Library

SpireCV-Pro is an edge real-time perception development platform built for intelligent unmanned systems and upgraded from SpireCV. It provides a broad set of foundational capabilities across x86 PCs, Jetson, Rockchip, and Huawei Ascend platforms. Key functions include lightweight large-model inference, image- and radar-based target detection and tracking, camera and gimbal control, video storage and streaming, and iterative edge-data management. Its goal is to provide mobile-robot developers with high-performance, reliable, easy-to-use, and feature-rich visual-perception capabilities.

PrometheusGroundStation-Pro

PrometheusGroundStation-Pro is the professional ground-control software designed specifically for the Prometheus UAV system. A single installation supports both ROS 1 and ROS 2. Its TCP/UDP-based communication mechanism greatly simplifies network configuration for multi-UAV collaboration and remote control, making it especially suitable for swarm operations. The interface is intuitive, runs across platforms, and supports both real-flight and simulated-flight environments, covering the complete workflow from algorithm validation to live testing. Developers can also integrate custom functions quickly. The software currently covers the Prometheus project's core capabilities, including basic control, swarm control, gimbal control, visual tracking, and path planning.