Website: realgazebo.chungbuk.ac.kr Repository: github.com/SUV-Lab/RealGazebo
RealGazebo is an integrated simulator that bridges PX4-Gazebo physics simulation with Unreal Engine 5 photorealistic rendering. It enables large-scale, high-fidelity simulation of multi-heterogeneous unmanned vehicles (UAVs, UGVs, USVs) with real-time RTSP video streaming.
Developed by SUV Lab, Chungbuk National University as a next-generation platform for autonomous vehicle research and development.
The RealGazebo plugin consists of 4 runtime modules working together:
The central coordination module that provides unified API access to all subsystems. Acts as a lightweight master coordinator that registers manager actors and forwards API calls to appropriate child subsystems.
Key Role: System integration and subsystem orchestration
Receives UDP data from PX4-Gazebo simulation and manages vehicle lifecycle. Handles packet parsing, vehicle spawning via object pooling, and real-time state updates.
Key Role: UDP communication (port 5005) and vehicle lifecycle management
Provides camera control modes and vehicle UI widgets for enhanced visualization. Offers multiple camera perspectives (follow, orbit, free-fly) and heads-up display elements.
Key Role: Camera systems and user interface
Hardware-accelerated H.264 video streaming via RTSP. Uses NVENC/AMF encoders for zero-copy GPU texture encoding with ultra-low latency (< 100ms end-to-end).
Key Role: Real-time video streaming (port 8554) for multi-camera vehicle feeds
Plugins/RealGazebo/
├── Source/
│ ├── RealGazebo/ # Master orchestrator module
│ ├── RealGazeboBridge/ # Network & vehicle management
│ ├── RealGazeboUI/ # Visualization & camera control
│ └── RealGazeboStreaming/ # RTSP streaming module
├── Content/ # Plugin content assets
└── README.md # This file
- Multi-Vehicle Support - Up to 256 instances per vehicle type (0-255 range). Tested with 8+ simultaneous heterogeneous vehicles (quadcopters, rovers, boats, VTOL)
- PX4 Integration - Full compatibility with PX4 autopilot via Gazebo bridge
- RTSP Streaming - Hardware-accelerated H.264 streaming at 30/60 FPS with < 100ms latency
- Zero-Copy Pipeline - Direct GPU texture to NVENC encoding (no CPU memory copies)
- Dynamic Spawning - Runtime vehicle creation/destruction via UDP protocol
- Object Pooling - Efficient vehicle reuse for performance optimization
- Cross-Platform - Linux and Windows support with platform-specific optimizations
- Unreal Engine: 5.1 or later
- Operating System: Linux (Ubuntu 20.04/22.04) or Windows 10/11
- GPU: NVIDIA RTX 30/40/50 series (for NVENC) or AMD RX 6000/7000 (for AMF)
- RAM: 16 GB minimum, 32 GB recommended
- CPU: Multi-core processor (16+ threads recommended)
RealGazebo (Master - UWorldSubsystem)
├── RealGazeboBridge (UGameInstanceSubsystem)
│ ├── UDP Receiver (port 5005)
│ ├── Vehicle Pool Manager
│ └── Data Stream Processor
│
├── RealGazeboUI (UGameInstanceSubsystem)
│ ├── Camera Controllers
│ └── UI Widgets
│
└── RealGazeboStreaming (UGameInstanceSubsystem)
├── RTSP Server (port 8554)
├── Hardware Encoders (NVENC/AMF)
├── Frame Pool & Capture
└── Streaming Pipelines (per-camera isolation)
rtsp://<IP>:<PORT>/<vehicle_type>_<vehicle_num>/<camera_id>
Examples:
rtsp://localhost:8554/x500_1/front
rtsp://localhost:8554/lc_62_2/bottom
rtsp://192.168.1.100:8554/rover_0/rear
Port: 5005
Packet Types:
- MessageID 1: Pose data (position + quaternion)
- MessageID 2: Motor RPM data
- MessageID 3: Servo state data
- MessageID 4: Destroy vehicle command
- MessageID 5: Battery + navigation state
Packet Format:
Header (3 bytes):
- vehicle_num: uint8 (0-255)
- vehicle_code: uint8 (vehicle type)
- message_id: uint8 (1-5)
This project is licensed under the GNU General Public License v3.0. See LICENSE file for full license information.
Copyright (c) 2024-2025 SUV Lab, Chungbuk National University
Authors: Gonapinuwala Lahiru Sandaruwan, MinKyu Kim
Supervisor: Prof. SungTae Moon
SUV Lab (Smart Unmanned Vehicles Laboratory) at Chungbuk National University focuses on deep learning-aided smart unmanned vehicle systems, with expertise in:
- Multi-UAV swarming flight technology
- Flight control systems and Bayesian filtering
- Real-time object detection for UAV imagery
- Deep learning-enhanced autonomous navigation
Director: Prof. SungTae Moon
Location: Chungbuk National University, Department of Intelligent System & Robotics
Contact: [email protected] | +82-43-261-3256
- Homepage: realgazebo.chungbuk.ac.kr
- SUV Lab: sites.google.com/view/suvlab
- PX4 Integration: github.com/SUV-Lab/RealGazebo-PX4
- ROS2 Integration: github.com/SUV-Lab/RealGazebo-ROS2
- Installation Guide: Documentation
For issues, questions, or contributions, please visit the GitHub Issues page.
SUV Lab - Chungbuk National University Advancing autonomous vehicle simulation technology.