V-DART: VR-Based Dual-Arm Robot Teleoperation with Mobile Base Control and Data Recording

We present V-DART (VR-based Dual-Arm Robot Teleoperation), an extended virtual reality framework for the teleoperation of mobile dual-arm robotic systems, building upon our previous work on stationary manipulator control. The enhanced system leverages Unreal Engine 5.6.1 (UE5) for real-time visualization and interaction, introducing critical advancements for complex environments: the control of the robot's base movement via both joystick input and direct mapping of the user's physical position and orientation within the tracked space. This allows users to navigate the robot intuitively by simply moving their body, while maintaining precise control over both arms using VR (Virtual Reality) controllers. Furthermore, we introduce a robust data logging layer that records expert movements during teleoperation tasks. By capturing timestamped streams of user pose, controller inputs, and robot states that can be synchronized downstream, this module facilitates the creation of high-quality datasets from expert demonstrations. These datasets can be leveraged for machine learning applications, such as behavior cloning, reducing the need for extensive manual annotation. The system maintains bidirectional communication between the VR interface and the physical robot, ensuring responsive control and real-time feedback. We describe the architectural enhancements, user interaction design, and performance evaluation, demonstrating the viability of this unified approach for both teleoperation and data-driven robotics development.