Deadline-Budgeted Embodied Reasoning via Non-Verbal Spatial Tokens
Build an embodied reasoning model using spatial tokens and anytime speculative planning to meet strict physical latency deadlines without increasing collisions.
Source paper
Grid Robotics, Qingfeng Lan, Duo Li, Siyao Li, Yiheng Lyu, Shijian Lu, Zhenghao Qi, Shizun Wang, Xinchao Wang, Ming Yang, Zhiqin Yang, Zuhao Yang, Tao Yu
Generated by AI. The paper’s authors did not write these.
Build an embodied reasoning model using spatial tokens and anytime speculative planning to meet strict physical latency deadlines without increasing collisions.
Construct an independent multi-view verification network using split conformal prediction to prevent premature task completion bias in visual-language agents.
Deploy a pre-execution safety gate that projects low-level robot commands onto a digital twin using GPU-accelerated collision checkers to bridge the text-to-action refusal gap.
This paper presents a conceptual framework and research roadmap for 'agentic robotics', which uses foundation models within closed-loop embodied harnesses (termed Embodied System 2) to orchestrate skills, tools, and real-time control (Systems 1 and 0). The authors distinguish digital from physical agents, define a comprehensive nine-dimension evaluation protocol, and articulate a four-level ladder of delegable responsibility from basic primitives to standing autonomous roles. They argue that physical constraints—such as irreversible actions, dynamic deadlines, and unobservable contact forces—require physical agents to rebuild fundamental assumptions about verification, scaling, and safety.
The paper outlines a functional tripartite architecture comprising System 2 (foundation model and embodied harness for high-level reasoning, planning, memory, and outcome verification), System 1 (callable perception, motion planning, and learned VLA/WAM skills), and System 0 (high-rate real-time joint and body control). It combines this reference architecture with a 9-dimension, 53-subdimension evaluation taxonomy and the 'Harness Card' reporting protocol to evaluate closed-loop agentic performance on live world clocks.