Handshake, fist bump, high five: XPENG's humanoid IRON switches gestures mid-motion
Chinese manufacturer XPENG shows in a lab test how its IRON robot responds to changing hand gestures without voice commands. The company credits its VLA 2.0 AI system.
Photo: Matti Blume / Wikimedia Commons, CC BY-SA 4.0 (cropped and resized)
Video: YouTube · XPENG · youtube.com
XPENG has released a video in which its humanoid robot IRON responds to a tester's hand: the tester first offers a handshake, then makes a fist and finally holds up an open palm for a high five. IRON adapts its movement each time without a word being spoken. When the high five moves out of reach of one arm, the robot switches to its other hand.
XPENG attributes this to its vision-language-action model VLA 2.0, which continuously analyses camera images and updates the planned motion instead of finishing a sequence once started. According to a presenter in the video, IRON reassesses the scene every 0.2 seconds – a claimed re-planning interval, not a measured response time. The model is said to run on board the robot and to predict behaviour better than its predecessor.
This is a lab demonstration; there are no independent tests. XPENG is best known as an electric-car maker.