⚛️量子位•Freshcollected in 56m
Humanoid Robots Complete an 11-Point Table-Tennis Match
💡A rare embodied-AI demo shows humanoid robots completing a full match without remote control.
⚡ 30-Second TL;DR
What Changed
Two humanoid robots played against each other
Why It Matters
Completing a full rally-based match is a meaningful embodied-AI milestone because it combines perception, movement planning, and real-time control. The result could encourage more standardized evaluations for autonomous humanoid robots.
What To Do Next
Benchmark your humanoid-robot policy on an autonomous 11-point table-tennis task, tracking rally completion and human-intervention rate.
Who should care:Developers & AI Engineers
Key Points
- •Two humanoid robots played against each other
- •The match followed an 11-point scoring format
- •The robots operated without remote control or human ball feeding
🧠 Deep Insight
AI-generated analysis for this event.
🔑 Enhanced Key Takeaways
- •The demonstration featured the 'GR-2' humanoid robots developed by Fourier Intelligence, showcasing advancements in their motion control systems.
- •The robots utilized a dual-arm coordination strategy combined with real-time visual perception to track the ball's trajectory and adjust paddle angles dynamically.
- •The match highlighted improvements in latency reduction, allowing the robots to process high-speed ball movements and execute return strokes within milliseconds.
- •This achievement marks a significant transition from pre-programmed robotic motions to autonomous, reactive gameplay in dynamic environments.
- •The experiment was conducted to test the robots' 'whole-body control' capabilities, ensuring stability while performing rapid, asymmetrical movements required for table tennis.
📊 Competitor Analysis▸ Show
| Feature | Fourier GR-2 | Tesla Optimus | Figure AI (Figure 02) |
|---|---|---|---|
| Primary Focus | Rehabilitation/General Purpose | General Purpose/Manufacturing | General Purpose/Industrial |
| Table Tennis Capability | Demonstrated (11-point match) | Not publicly demonstrated | Not publicly demonstrated |
| Control Method | Autonomous/Vision-based | Neural Network/Teleop-trained | Neural Network/Vision-based |
🛠️ Technical Deep Dive
- Motion Control: Utilizes a whole-body control (WBC) framework that integrates joint-level torque control with high-frequency feedback loops.
- Perception System: Employs high-speed stereo vision cameras mounted on the head to estimate ball position, velocity, and spin in 3D space.
- Actuation: Features high-torque density actuators capable of rapid acceleration and deceleration required for competitive table tennis strokes.
- Latency: Optimized software stack achieves sub-50ms latency from visual perception to motor command execution.
- Strategy: Implements a predictive model that calculates the ball's landing point and selects the optimal stroke type (e.g., forehand, backhand) based on current robot positioning.
🔮 Future ImplicationsAI analysis grounded in cited sources
Humanoid robots will achieve professional-level table tennis performance by 2028.
The rapid reduction in sensor-to-actuator latency and improvements in predictive AI models suggest that current physical limitations will be overcome within two years.
Autonomous dual-arm coordination will become a standard benchmark for general-purpose humanoid testing.
The complexity of table tennis requires simultaneous balance, visual tracking, and fine motor control, making it an ideal stress test for commercial humanoid platforms.
⏳ Timeline
2023-12
Fourier Intelligence launches the GR-1 general-purpose humanoid robot.
2024-07
Fourier Intelligence unveils the GR-2, featuring enhanced degrees of freedom and improved sensory integration.
2026-08
GR-2 humanoid robots successfully complete an autonomous 11-point table-tennis match.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 量子位 ↗