AuraTracer智迹闻
中文

EVENT DOSSIER

Yushu Technology: First in the world to achieve real-time control of a fully autonomous humanoid robot in combat using a world model

2026-09-07 19:14 Models 🔥 44.2 heat score
1sources
1days unfolding
44.2heat score
2mentions
SummaryAI generated

On September 7, 2026, Yuzhi Technology announced the world’s first demonstration of full-autonomous humanoid robot combat driven by a world model in real time. This breakthrough relied on the UnifoLM-X2-1.0 large model, overcoming bottlenecks in instantaneous planning, decision-making, and dynamic interaction execution. The robot can predict future states based on its internal physical environment, rather than relying on pre-set programs or manual control. This achievement verifies the basic feasibility of the world-model approach for large-scale deployment, marking a significant leap in humanoid robots from passive execution to real-time prediction and autonomous decision-making capabilities.

Related eventsRELATED EVENTS
Key entitiesKEY ENTITIES
UnifoLM-X2-1.0Yushu Technology

Coverage · reports per dayLANGUAGE SPLIT

Entity relations
UnifoLM-X2-1.0 × Yushu …1

SignalsSIGNALS

Keyword heat
  • Yushu Technology1
  • UnifoLM-X2-1.01

All reports (1)SOURCES

I IT之家 zh 2026-09-07 19:14

Yushu Technology: The world’s first fully autonomous humanoid robot combat system driven by real-time world models

Unistar Technology announced today that it has successfully achieved real-time driving of a fully autonomous humanoid robot in combat for the first time worldwide. UnifoLM-X2-1.0 has broken through bottlenecks in instant planning, decision-making, and dynamic interaction execution based on large-scale action models, enabling high-dynamic, strong-interaction, real-time future prediction and planning. This has led to the realization of fully autonomous humanoid robot combat, verifying the basic feasibility of large-scale implementation of this approach. Previously, Unistar Technology demonstrated a combat performance that relied on preset programs or manual remote control at the China International Import Expo on November 5, 2025. The breakthrough here is that the robot no longer relies on preset programs or manual remote control; instead, it uses a world model to predict future states in real time and independently make decisions and perform dynamic interactions. The core concept of the “world model” is to enable intelligent agents to establish predictive representations of the physical world internally, and to deduce and plan for future states in an abstract space, rather than reconstructing the environment pixel by pixel. Industry representatives included Yann LeCun, Chief AI Scientist at Meta, who proposed the JEPA architecture series. Unistar emphasized UnifoLM-X2-…