Toward the Cognitive--Physical Limits of Embodied Intelligence through a World-Model-Centric Autonomous Racing Agent

2026-08-11

AI Summary

This work addresses the challenge of jointly modeling cognitive and physical limits in existing embodied intelligence systems. It proposes a boundary-aware embodied intelligence framework centered on a world model to develop an autonomous racing agent capable of co-optimizing perception, reasoning, and control near performance limits through closed-loop iterative learning. The approach leverages, for the first time, a world model trained on real high-speed racing data, integrating high-frequency localization, future-state reasoning, and limit-handling control, with continuous refinement via closed-loop interaction to enhance the joint representation of cognitive and physical boundaries. Real-world experiments achieved speeds of 256.3 km/h and lateral accelerations of 26.8 m/s², while simulations demonstrated an 88.3% success rate across diverse challenging scenarios, significantly improving极限 performance, failure recovery, and cross-scenario generalization.

Abstract

Embodied artificial intelligence aims to develop agents that perceive, reason, and act through continuous interaction with the physical world. However, most embodied systems are still evaluated within conservative safety margins or moderate interaction regimes, leaving their capability boundaries under extreme conditions insufficiently understood. Autonomous racing provides a stringent testbed by combining high-frequency localization and perception, adversarial interaction, near-saturated vehicle dynamics, and strict safety constraints. Existing systems push high-speed performance but rarely model and refine cognitive and physical limits jointly. Here we show that a world-model-centric autonomous racing agent provides a concrete step toward exploring these coupled limits. The framework learns predictive world models from near-limit successes and failures to capture interaction evolution, ego dynamics, and feasible-motion boundaries, coupling world-state construction, future-aware reasoning, and near-limit control in a closed-loop refinement process. Training data were collected from real-vehicle autonomous racing, where the onboard system maintained robust localization and perception at speeds up to 256.3 km/h and peak lateral acceleration of 26.8 m/s$^2$. In full-scale simulated racing, the well trained world-model-centric agent achieves an 88.3% interaction success rate across various challenging simulated racing scenarios. Closed-loop refinement of the world model and policy further improved utilization of cognitive-physical limits, recovery from failure modes, and generalization across varying conditions and unseen circuits. These results suggest a boundary-aware methodology in which world models help embodied agents represent, predict, and continually refine their capability boundaries for safer real-world deployment.

Problem

• embodied intelligence
• cognitive-physical limits
• autonomous racing
• world model
• capability boundaries

Innovation

• world model
• embodied intelligence
• autonomous racing
• cognitive-physical limits
• closed-loop refinement