On July 23, Alibaba Group's Amap announced that its ABot embodied system has completed a full-stack upgrade, and重磅ly launched five core models: ABot-N1, ABot-M0.5, ABot-ER, ABot-AgentOS, and ABot-C0. This upgrade achieved systematic integration of perception, decision-making, execution, and memory, significantly enhancing the robot's ability to autonomously perform general tasks, marking a new stage of highly collaborative and integrated embodied intelligence technology.

As the world's first full-stack embodied technology system, ABot deeply integrates world models, base models, and embodied Agent architecture, building a self-evolving closed-loop flywheel that is mutually reinforced by "simulation training, physical interaction, and memory scheduling."

The five upgraded models precisely map the core functional components of robots: ABot-N1, as a general navigation base, maps the "feet," using a dual-system architecture with fast and slow modes, balancing long-range logical reasoning and real-time control; ABot-M0.5, as a general operation base, maps the "hands," separating movement and operation into independent action streams and combining frame-level implicit actions to achieve real-time alignment of eyes, hands, and feet; ABot-ER, the embodied brain, and ABot-AgentOS, the execution center, together form the top-level architecture. The former deepens reasonable decision-making on the relationship between space and tasks, while the latter enables unified adaptation for humanoid, quadrupedal, and wheeled robots, as well as cross-task multimodal lifelong memory management. ABot-C0, the motion control system, acts as the "motor nerve," building a unified behavioral base for quadrupedal robots, ensuring precise conversion of decisions into physical actions.

Currently, embodied intelligence is at a critical turning point from single-technology breakthroughs to complex scenario deployment. Amap has overcome the previous bottleneck of fragmented robotic capabilities by establishing a full-stack collaborative architecture covering "hands, feet, brain, central nervous system, and motor nerves." This not only significantly enhances the generalization and self-evolution capabilities of mechanical bodies in real-world open environments but also provides a highly valuable systematic model for large-scale commercial deployment of general-purpose robots in industries and services.