In recent years, both the civilian ad military sectors have actively pursued research on unmanned vehicles across land, sea, and air for a variety of purposes. In particular, as interest grows in technologies for operating multiple heterogeneous unmanned vehicles simultaneously, the demand for simulators capable of prior verification has also increased. Gazebo simulation, developed as open-source software, has been widely used in various fields as a pre-verification tool for unmanned systems. However, it lacks photorealism, making it insufficient for generating training data for deep learning-based vision tasks, and it is not well-suited for multi-heterogeneous operations.
We introduce RealGazebo, a system designed to build a digital environment with high fidelity to the real world and optimized for the use of multiple heterogeneous unmanned vehicles. RealGazebo integrates Gazebo with Unreal Engine on a flexible architecture and is built on a ROS2-based system, enabling interoperability among unmanned vehicles with different structures. Furthermore, it has been released as open source, allowing anyone to utilize it.
Main Features
Flexible Simulation Architecture
A Flexible Simulation Architecture with Separation of Rendering, Physics Engine, Control, and Mission Logic for Large-Scale Multi-Unmanned Vehicle Operations
ROS2-Based Operational Framework
ROS2-Based Operational Architecture to Achieve Interoperability of Heterogeneous Unmanned Vehicles
Digital Twin-Driven High-Fidelity Environment
Development of a Digital Twin-Based High-Fidelity Environment for Autonomous Vehicle Experiments
Refined Mobility Models for Physical AI
Refinement of Mobility Models for Autonomous Drone Flight toward Physical AI Learning
Simulation Scenarios
C-Track Overview
C-Track offers a variety of spaces designed for different types of experiments. These include environments like forests, unpaved roads, urban roadways, and the VILS test facility, among others.