Скачать или смотреть Digital Twins for Real-Virtual Collision Avoidance in Mobile Robotics using ROS2 and Unreal Engine

This work presents a Digital Twin framework for mobile robotic agents that tightly integrates a virtual robot with its physical counterpart through ROS2. The digital twin is implemented using the TurtleBot3-UE plugin based on the rclUE library, providing a kinematic differential-drive model driven by linear and angular velocity commands. The physical platform is an iRobot Create 3 robot, equipped with multiple onboard sensors, including a YD-LIDAR, IMU, wheel encoders, and proximity sensors.

Both the virtual and real robots are provided with LIDAR-based perception, enabling a direct comparison between synthetic and real sensor data. A collision avoidance controller is developed in ROS2, generating forward motion commands and reactive steering maneuvers in response to detected obstacles. The controller is evaluated in a realistic warehouse-like environment containing shelves, pallets, barrels, and boxes.

Experimental results demonstrate that synthetic LIDAR data generated by the digital twin can effectively guide the behavior of the physical robot, yielding comparable performance to real sensor data. This validates the proposed Digital Twin approach as a reliable tool for controller development, testing, and validation, while reducing risk and dependence on physical experimentation.