A Phased Research Roadmap for Embodied Intelligence and Whole-Body Digital Twins: From Biological Measurement to Closed-Loop Computational Embodiment

The Human Embodied Intelligence Research Program proposes a staged research program connecting the empirical study of organism-level embodied cognition with the longer-term development of personalized, dynamically updated whole-body computational models. The program deliberately separates what can be tested now from what depends on unresolved biological findings or future technological capabilities. This paper translates the preceding theoretical, mechanistic, experimental, and technological manuscripts into a dependency-driven research roadmap. Rather than assuming that full-body intelligence constitutes a validated latent construct, that whole-body digital twins can presently be constructed, or that computational fidelity implies continuity of consciousness or personal identity, the roadmap treats each proposition as contingent on explicit empirical milestones. Three broad phases are proposed. Phase I establishes the biological and measurement foundation: validation of candidate dimensions of full-body intelligence, characterization of within-person and between-person variability, multimodal physiological acquisition, standardized perturbation paradigms, and initial personalized computational models. Phase II develops and tests closed-loop computational embodiment: multimodal integration, subject-specific state estimation, internal-state regulation, sensorimotor interaction, modality ablation, functional-equivalence testing, and increasingly rigorous digital-twin verification, validation, and uncertainty quantification. Phase III addresses longitudinal computational embodiment: adaptive updating, model drift, sustained subject-specific prediction, progressively richer environments, human–AI interaction, governance, and long-duration tests of computational continuity. Progression is governed by explicit go/no-go criteria. Failure to validate a unified full-body-intelligence construct does not terminate the program; it redirects modeling toward independently validated component mechanisms. Failure of anticipatory physiological activity does not undermine the mainstream embodied-cognition foundation. Failure of increasingly detailed bodily modeling to improve held-out prediction or adaptive behavior would count against strong versions of the embodied-digital hypothesis. Previously proposed 1–3-, 3–7-, and 7–15-year horizons are retained only as illustrative development scenarios, not forecasts. Scientific progress is instead defined by prerequisite completion and evidential thresholds. The resulting roadmap is designed to permit success, revision, branching, or termination at every stage. Its objective is not to guarantee eventual digital reproduction of a human being, but to determine empirically how far validated biological dynamics can be measured, modeled, reconstructed, coupled, and sustained computationally. This description follows the manuscript's abstract closely and preserves its central distinction between scenario-planning horizons and evidence-gated scientific progression.