Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Toward Embodied AGI: A Review of Embodied AI and the Road Ahead

Yequan Wang, Aixin Sun

TL;DR

The paper tackles the gap between current embodied AI and true Embodied AGI by proposing a five-level taxonomy (L1–L5) and four capability axes to benchmark progress. It analyzes current end-to-end and plan-and-act paradigms, identifies bottlenecks hindering advances to L3+, and presents a conceptual L3+ framework comprising an omnimodal model architecture and lifelong, physically grounded training. The work contributions include a structured roadmap, critical capability analysis, and a concrete framework intended to inspire future research toward open-ended, real-time, humanoid-capable embodied agents. Its significance lies in providing a principled direction for integrating multimodal perception, human-like cognition, and robust generalization with hardware evolution and safety considerations to advance toward practical Embodied AGI.

Abstract

Artificial General Intelligence (AGI) is often envisioned as inherently embodied. With recent advances in robotics and foundational AI models, we stand at the threshold of a new era-one marked by increasingly generalized embodied AI systems. This paper contributes to the discourse by introducing a systematic taxonomy of Embodied AGI spanning five levels (L1-L5). We review existing research and challenges at the foundational stages (L1-L2) and outline the key components required to achieve higher-level capabilities (L3-L5). Building on these insights and existing technologies, we propose a conceptual framework for an L3+ robotic brain, offering both a technical outlook and a foundation for future exploration.

Toward Embodied AGI: A Review of Embodied AI and the Road Ahead

TL;DR

The paper tackles the gap between current embodied AI and true Embodied AGI by proposing a five-level taxonomy (L1–L5) and four capability axes to benchmark progress. It analyzes current end-to-end and plan-and-act paradigms, identifies bottlenecks hindering advances to L3+, and presents a conceptual L3+ framework comprising an omnimodal model architecture and lifelong, physically grounded training. The work contributions include a structured roadmap, critical capability analysis, and a concrete framework intended to inspire future research toward open-ended, real-time, humanoid-capable embodied agents. Its significance lies in providing a principled direction for integrating multimodal perception, human-like cognition, and robust generalization with hardware evolution and safety considerations to advance toward practical Embodied AGI.

Abstract

Artificial General Intelligence (AGI) is often envisioned as inherently embodied. With recent advances in robotics and foundational AI models, we stand at the threshold of a new era-one marked by increasingly generalized embodied AI systems. This paper contributes to the discourse by introducing a systematic taxonomy of Embodied AGI spanning five levels (L1-L5). We review existing research and challenges at the foundational stages (L1-L2) and outline the key components required to achieve higher-level capabilities (L3-L5). Building on these insights and existing technologies, we propose a conceptual framework for an L3+ robotic brain, offering both a technical outlook and a foundation for future exploration.

Paper Structure

This paper contains 8 sections, 1 equation, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. L1 $\sim$ L5: Roadmap to Embodied AGI
  3. L1 $\sim$ L2: Status Quo and Challenges
  4. L3 $\sim$ L5: Key Constituents
  5. A Conceptual Framework for L3+ Robots
  6. Model Structure
  7. Training Paradigm
  8. Conclusion and Future Challenges

Figures (4)

  • Figure 1: Roadmap of the five levels of Embodied AGI, inspired by the established levels of autonomous driving.
  • Figure 2: Illustration of four basic constituents of Embodied AGI. For a detailed discussion of the four subdomains of humanoid cognition, please see Section \ref{['sec:l3l5']}.
  • Figure 3: Omnimodal Capabilities and Illustrative Model Structure.
  • Figure 4: Conceptual Framework: Illustrative Training Paradigms.

Theorems & Definitions (1)

  • Definition 1: Embodied AGI