A wheeled-arm robot integrating multimodal perception capabilities and embodied manipulation skills, suitable for patrol and inspection duties in office environments, factories, and industrial sites. Equipped with visible light recognition and thermal infrared identification capabilities, it enables customisable object recognition through
large-scale models at any time.

Utilising robotic arms and grippers, the system employs VLA's end-to-end large language model to automatically identify lifts. It autonomously navigates lift travel through operations such as button pressing and card swiping, requiring no equipment modifications. Coordinated with chassis movement, the robot independently performs common tasks encountered in daily operations, including transporting items, operating buttons, and opening/closing cabinet doors.