On the evening of August 20, Gaode Momentum brought its robotic dog, Tutu, to the 2026 World Robot Conference, unveiling five major scenario capabilities: guide assistance for the visually impaired, immersive tour companionship, smart shopping guidance with autonomous errands, multi-space intelligent inspection, and last-mile logistics delivery. As the world's first four-legged robot dog with fully autonomous action capabilities in open environments, Tutu has evolved from its initial launch in April to now covering five practical use cases, marking a shift in embodied intelligence from technical demonstrations to real-world applications in open spaces and daily life services.

Open environments represent the toughest test for embodied robots: tactile paving may be obstructed, pedestrian flows shift unpredictably, and tasks are often delivered as nothing more than a vague sentence. What Tutu aims to achieve is converting a single instruction into a complete task chain—comprehending intent, sensing the environment, planning routes, adapting to changes, and docking safely—all while navigating diverse and complex scenarios. The five use cases, spanning barrier-free guidance, cultural tourism commentary, shopping assistance, inspection and security, and last-mile delivery, are not merely a simple stack of functions but a concentrated demonstration of autonomous task execution in open settings: Tutu can independently interpret needs, plan actions, and continuously adjust to changing conditions until the task is completed.

The key to Tutu's ability to autonomously handle multi-scenario tasks in open environments lies in Gaode's development of the ABot full-stack embodied technology system, the world's first built for AGI. This system is structured into three layers: the DATA layer, the MODEL layer, and the AGENT layer. Unlike fragmented single-model research, Gaode's ABot system establishes a self-evolving closed-loop flywheel—simulation training, physical interaction, and memory scheduling reinforce one another, enabling the system to improve continuously as it operates in the real, open world. To date, the ABot system has achieved SOTA (state-of-the-art) results across 17 authoritative global benchmarks.

At this exhibition, Gaode Momentum also showcased the ABot-World Studio, a general world model workshop, alongside real-life task demonstrations by a wheeled dual-arm robot. Visitors can generate an interactive AI world in real time by inputting text or an image, while the wheeled dual-arm robot demonstrates visual perception, task planning, and dynamic force control through arithmetic tasks performed with building blocks.