XPeng IRON upgrade: second-generation VLA deployed on-device, re-infers every 0.2 seconds and predicts actions for the next 6 seconds
XPeng Motors announced the latest progress on its humanoid robot IRON. The upgraded second-generation VLA large model has been deployed on the robot's device side, enabling it to re-infer every 0.2 seconds based on real-time visuals and predict action changes within the next 6 seconds. The demonstration covered scenarios such as handshakes, fist bumps, and high-fives; when the interacting person temporarily changes gestures or position, IRON can adjust its actions and switch hands to coordinate. In terms of hardware, IRON is equipped with three self-developed Turing AI chips, with effective computing power up to 2250TOPS, a dexterous hand with 22 degrees of freedom per hand, and a model architecture combining VLT, VLA, and VLM.
Why read this
AI score 58
The original text provides the timing parameters and hardware configuration of IRON's on-device inference, which readers can use to assess the progress of real-time interaction in humanoid robots.
Topics
Body · AI translation
On October 22, XPeng Motors announced the latest progress on its humanoid robot IRON. The upgraded second-generation VLA large model has been deployed on the robot's device side, enhancing its temporal understanding and ability to predict future actions. According to the company, IRON can re-reason every 0.2 seconds based on real-time images and predict action changes within the next 6 seconds, continuously adjusting its interactive movements.
This demonstration covered scenarios such as handshakes, fist bumps, and high-fives. When the interactor temporarily changes an open palm into a fist, or changes the position of a high-five, IRON can adjust its movements accordingly and switch hands to coordinate. These capabilities rely on XPeng's physical AI foundation, updating actions through continuous observation and reasoning, enabling the robot to respond to changes during interactions and reducing its dependence on humans coordinating to a fixed rhythm.
On the hardware side, IRON is equipped with three self-developed Turing AI chips, with effective computing power of up to 2250 TOPS, and features dexterous hands with 22 degrees of freedom per hand. XPeng adopts a model architecture combining VLT, VLA, and VLM, which respectively handle autonomous decision-making, full-body control, and perception functions, providing support for the robot's dialogue, walking, and interaction.
On the industrialization side, XPeng announced in September this year that its humanoid robot production line had been put into operation, with IRON walking off the line autonomously after completing production. According to the plan previously announced by the company, its goal is to start large-scale mass production by the end of 2026, begin market launch and delivery in 2027, and take the lead in exploring commercial applications in scenarios such as XPeng stores.
Submissions/Tips: tougao@youxituoluo.com
Article/Business Cooperation: Lin Nan (WeChat 19250561593)
Join the industry exchange group: Lin Nan (WeChat 19250561593)
Metaverse Digital Industry Service Platform
Download the "Tuoluo Technology" APP to get cutting-edge, in-depth metaverse information