On August 19, at the 2026 World Robot Conference (WRC 2026), Guangxiang Technology’s industrial-grade, self-evolving embodied AI robot, Phi-Bot X1, demonstrated its "multi-capability" performance in a live setting, continuously executing tasks ranging from welding-related material handling to mobile quality inspection. This marks the industry's first instance of a single embodied AI robot autonomously cycling through operations across heterogeneous workstations involving different types of tasks, thereby validating the technical feasibility of scaling "physics-native intelligence" from single-task execution to multi-task generalization.
In a continuous dual-station operation, the Phi-Bot X1 first uses its right-hand gripper to pick up a workpiece from the rack at the welding loading station; it then moves the part to a sliding platform, positioning it with millimeter-level alignment accuracy and an orientation precision of within 0.3° to complete the loading process. Next, coordinating with the sliding platform, it controls the platform's movement before grasping the workpiece and placing it onto the rack to complete the unloading process. Upon finishing the loading and unloading tasks, the Phi-Bot X1 autonomously moves to the quality inspection station, where it performs a dynamic, comprehensive inspection of the vehicle body surface using a vision module paired with a lighting array mounted on its left-hand end-effector. After the inspection, the Phi-Bot X1 returns to the welding loading station to begin the next cycle. The loading/unloading and mobile inspection tasks are handled separately by the left and right arms, with task switching and station transitions performed autonomously by the robot. Beyond the demonstrated tasks, the Phi-Bot X1 supports expansion to additional stations through the flexible swapping of end-effectors, truly realizing the concept of "one machine, multiple capabilities."
