The embodied brain: Bridging the brain, body, and behavior with neuromechanical digital twins

TL;DR

It is envisioned that coupling studies on animals with active probing of their neuromechanical twins will greatly accelerate neuroscientific discovery and show how neuromechanical twins can advance healthcare.

Abstract

Animal behavior reflects interactions between the nervous system, body, and environment. Therefore, biomechanics and environmental context must be considered to dissect algorithms for behavioral control. This is enabled by leveraging neuromechanical digital twins: computational models that embed artificial neural controllers within realistic body models in simulated environments. Here we review advances in the creation and use of neuromechanical digital twins while also highlighting emerging opportunities for the future. First, we illustrate how neuromechanical models allow researchers to infer hidden biophysical variables that may be difficult to measure experimentally. Additionally, by perturbing these models, one can generate new experimentally testable hypotheses. Next, we explore how neuromechanical twins have been used to foster a deeper exchange between neuroscience, robotics, and machine learning. Finally, we show how neuromechanical twins can advance healthcare. We envision that coupling studies on animals with active probing of their neuromechanical twins will greatly accelerate neuroscientific discovery.

Figures (4)

• Figure 1: Parallels between real animals and their neuromechanical digital twins. How physiological, mechanical, and environmental variables can be accessed and manipulated in real animals (upper half), or in digital twins (lower half). Rows on the lower half summarize the full promise of twins, the current status of the approach, and the challenges that need to be overcome to accomplish its full potential. Real datasets that can be used to link these systems are indicated.
• Figure 2: Prior knowledge of biological systems helps infer structure, behavior, and function from limited experimental data.(Top row) As an analogy, consider the task of reconstructing a vase based on a partial sketch. Naïvely, we cannot accurately model the shape of the vase using only partial observations. However, if we apply the prior that the vase should be radially symmetric, we can fill in missing regions and consolidate redundant measurements. (Middle and bottom rows) Similarly, embedding biologically plausible features (e.g., priors concerning body morphology, musculoskeletal organization, and neural network topology) into the digital twin constrains the solution space and improves reconstruction quality from sparse or noisy experimental data. Since simulated dynamics are fully accessible, this can also greatly extend the amount of information that can be inferred from limited experimental data.
• Figure 3: Two directions for creating a dialogue between animal experiments and neuromechanical simulations. (a) In reverse simulations, one can replay experimental behavioral data in silico and use this replay to infer hidden variables like neural and muscle activities that are otherwise difficult to measure. (b) Starting from a hypothesized neural circuit, one can roll out behavior in silico and compare it with experimental observations.
• Figure : Image attributions: human model: MyoSuite vittorio_myosuite_2022; C. elegans model: BAAIWorm zhao_baaiworm_2024; rodent model: MIMIC aldarondo_virtual_2024; zebrafish model: simZFish liu_artificial_2025; walking Drosophila model: NeuroMechFly lobato-rios_neuromechfly_2022wang-chen_neuromechfly_2024; flying Drosophila model: FlyBody vaxenburg_whole-body_2025.