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

In a new arXiv paper, researchers Sibo Wang-Chen and Pavan Ramdya present a comprehensive framework for 'neuromechanical digital twins.' These are advanced computational models designed to simulate the complete loop of animal behavior by embedding artificial neural network controllers within highly realistic, biomechanically accurate body models operating in simulated environments. The core argument is that to truly understand behavioral control algorithms, one must move beyond studying the brain in isolation and instead model its continuous, dynamic interaction with a physical body and the external world. This approach directly challenges more abstract, disembodied models of neural function.

The 18-page review outlines how these digital twins serve as powerful tools for scientific discovery. They allow researchers to infer biophysical variables—like specific muscle forces or neural signals—that are extremely difficult or impossible to measure directly in living organisms. By running systematic 'in silico' experiments and perturbations on the twin, scientists can generate novel, experimentally testable hypotheses about how the nervous system controls complex actions. The authors highlight that this methodology creates a vital bridge between traditionally separate fields: neuroscience provides the biological questions, robotics offers the tools for embodied simulation, and machine learning contributes advanced artificial neural controllers and optimization techniques.

Looking forward, Wang-Chen and Ramdya envision these neuromechanical twins significantly accelerating progress in fundamental neuroscience. The paper also showcases promising applications in healthcare, such as modeling neurological disorders or designing more effective neuroprosthetics by simulating their interaction with a patient's digital twin. The ultimate goal is to establish a tight feedback loop where data from real-world experiments continuously improves the digital model, and simulations actively guide the next round of physical experiments, creating a virtuous cycle of discovery.