←News detail

Shanghai ObserverUpdated Oct 30 22:44

Agibot KuTuo Quadruped Robot's Perception-Based Motion Control and Autonomous Navigation Both Iterated

AI summary

At the 2026 Agibot Innovation Day, Agibot KuTuo released the latest iterations of two core technologies for its quadruped robots, "Lingyue" and "Lingtu." The "Lingyue" all-terrain intelligent motion engine integrates reinforcement-learning motion control with perception-based motion control, achieving an "automatic transmission" capability of terrain recognition and automatic gait switching; the "Lingtu" all-domain autonomous navigation engine demonstrated VLN semantic navigation, enabling the robot to understand targets and spatial relationships in natural language. Currently, the relevant quadruped robots have been deployed in security patrol and industrial inspection scenarios.

Body · AI translation

Following the first release of two major technologies at the World Artificial Intelligence Conference (WAIC) held in July this year, at yesterday's 2026 AgiBot Innovation Day, AgiBot Cuto brought the latest iterative results of its two core technologies: the quadruped robots "Lingyue" and "Lingtu." From technology debut to capability advancement, "Lingyue" continues to break through the challenge of traversal efficiency in complex terrain. When faced with diverse terrain such as stairs, high platforms, obstacles, and narrow passages, traditional quadruped robots often need to switch movement modes manually or by preset, increasing operation and development costs. The "Lingyue" all-terrain intelligent motion engine demonstrates a new "automatic transmission" capability. By integrating reinforcement learning motion control with perception-based motion control, the robot can autonomously adjust its movement strategy according to terrain changes, achieving terrain recognition and automatic gait switching, reducing the need to repeatedly switch modes between different terrains, improving traversal efficiency in complex environments, and lowering the threshold for secondary development. From coordinate navigation to semantic understanding, "Lingtu" continues to expand the boundaries of autonomous navigation capabilities. Addressing the problem of complex navigation task configuration in large industrial environments, the "Lingtu" all-domain autonomous navigation engine demonstrates VLN (vision-language navigation) semantic navigation capability. This technology combines a visual large semantic model with an embodied navigation large model, enabling the robot to understand goals, directions, and spatial relationships in natural language, and to actively seek targets by combining visual perception and map information, taking navigation from heading to specified coordinates further toward understanding business goals, and simplifying task configuration and development processes. From the technology debut at WAIC to the capability implementation and verification at this Innovation Day, AgiBot Cuto continues to focus on iteration in two core directions: perception-based motion control and autonomous navigation. Currently, AgiBot Cuto's quadruped robots have been deployed in security patrol and industrial inspection scenarios. In the future, AgiBot Cuto will continue to be guided by real operational needs, persistently tackle pain points in complex environment applications, promote the transformation of technological capabilities into scenario value, and help quadruped robots achieve large-scale commercial use in more industries.

Source

Shanghai Observer