AGIBOT Launches AGILE 2.0, Advancing Perception-Driven Locomotion for Humanoid Robots

【SHANGHAI, China, September 9, 2026】AGIBOT today announced the launch of AGILE 2.0 (AGIBOT Generative Intelligent Locomotion Engine), its second-generation framework for perception-driven locomotion and whole-body control.


AGILE 2.0 integrates visual perception, environmental understanding, whole-body locomotion and manipulation within a unified real-time closed loop, enabling perception to continuously drive how humanoid robots move, manipulate and interact with their surroundings.


To demonstrate the framework in action, AGIBOT released “Acrobat BOT,” a short film featuring AGIBOT X2 humanoid robots performing a series of highly challenging dynamic tasks. From balancing on a rolling ball and Chinese diabolo to three-robot rope jumping, cooperative platform climbing and collaborative box handling, the demonstrations put AGILE 2.0 to the test across dynamic motion, unstable contact conditions, continuous locomotion and manipulation, and multi-robot coordination.


Acrobat BOT: Putting Perception-Driven Locomotion to the Test

“Acrobat BOT” opens with AGIBOT X2 jumping through a ring of fire, followed by a range of individual and multi-robot challenges including Chinese diabolo, balancing on a rolling ball, three-robot rope jumping, two-robot leapfrog, cooperative platform climbing, three-ball juggling and collaborative box handling.


Across these tasks, the robots must continuously perceive changing conditions, coordinate whole-body movement and perform manipulation without interrupting motion or losing stability. The demonstrations provide a visible test of AGILE 2.0 under highly dynamic conditions, where perception, locomotion and manipulation must operate together in real time.


1789383116215472.gif

AGIBOT X2 jumping through a ring of fire


The Core Breakthrough: From Perception to Whole-Body Action

AGILE 1.0 established the foundation by connecting perception with motion and enabling locomotion and manipulation to operate concurrently. Building on this foundation, AGILE 2.0 places visual perception directly within the locomotion-control loop, continuously driving localization, posture adaptation, contact decisions and motion replanning as the robot moves.

 

AGILE 2.0 also introduces a Unified Whole-Body Motion Representation, bringing locomotion and manipulation into a shared motion framework. This enables whole-body loco-manipulation, allowing robots to perform fine-grained manipulation during continuous movement while maintaining whole-body stability and motion continuity. Combined with real-time visual closed-loop reasoning and dynamic constraint optimization, the framework can adjust posture and contact strategies and replan motion paths in real time as conditions change.

 

One of the clearest demonstrations is AGIBOT X2 autonomously balancing on and propelling a large rolling ball. The task combines highly dynamic motion with an unstable contact surface. As the ball moves beneath its feet, the robot continuously adjusts its posture and center of gravity to maintain dynamic balance while coordinating whole-body movement.


1789383188595951.gif

AGIBOT X2 balancing on and propelling a large rolling ball


AGILE 2.0 extends these capabilities to multi-robot coordination. In the three-robot rope-jumping task, two AGIBOT X2 robots swing the rope while a third tracks its trajectory and determines when to enter, jump and exit. The framework supports cross-robot state perception and coordination, enabling robots to respond to their partners’ motion states, physical interactions and action timing. In collaborative box handling, two robots coordinate their movements to transfer and precisely stack a box.

 

1789383260852050.gif

AGIBOT X2 robots performing three-robot rope jumping

 

From Acrobatics to Real-World Deployment

Beyond the demonstrations, AGILE 2.0 is designed to operate in more dynamic real-world environments. The framework supports human-aware risk perception and adaptive motion responses, enabling robots to adjust movement trajectories, contact forces and motion speed in environments shared with people. It also supports autonomous recovery from unexpected disturbances and posture deviations, allowing robots to recover while maintaining the overall task logic.

 

These capabilities are designed to reduce reliance on scenario-specific programming and customized development, providing a more scalable motion-intelligence foundation for broader real-world deployment of embodied AI.


For more information, please visit AGIBOT.com and follow AGIBOT on:
X: https://x.com/AGIBOTofficial
LinkedIn: https://www.linkedin.com/company/agibot/
YouTube: https://www.youtube.com/@AGIBOTofficial
Instagram: https://www.instagram.com/agibotofficial/
TikTok: https://www.tiktok.com/@agibotofficial
Facebook: https://www.facebook.com/AGIBOTofficial/

 

 

About AGIBOT
AGIBOT is a pioneer in the global general-purpose AI robotics industry, with its core technology focused on embodied AI. AGIBOT's "Three Intelligences in One" architecture integrates Locomotion Intelligence, Interaction Intelligence, and Manipulation Intelligence into a unified embodied system. Its portfolio spans humanoid robots, quadrupeds, dexterous systems, and commercial cleaning solutions. In June 2026, AGIBOT announced that its 15,000th robot had rolled off the production line.