Biomechanically Realistic Virtual Rodent Created

Neuroscientists from Harvard University collaborated with scientists at Google's DeepMind AI lab to develop a virtual rat with an artificial brain that can mimic the agile movements of real rodents. This innovative approach aims to unravel the mysteries of how the brain controls movement.

Leveraging high-resolution data from real rats, the team trained an artificial neural network—the virtual rat's "brain"—to control a biomechanically realistic digital model within a physics simulator. The results, published in Nature, revealed that the activations in the virtual control network accurately predicted neural activity measured from the brains of real rats performing the same behaviors.

"The collaboration was 'fantastic,'" said Bence Ölveczky, the lead researcher. "DeepMind had developed a pipeline to train biomechanical agents to move around complex environments. We simply didn't have the resources to run simulations like those, to train these networks."

Matthew Botvinick, Senior Director of Research at Google DeepMind, echoed the sentiment, stating, "We've learned a huge amount from the challenge of building embodied agents: AI systems that not only have to think intelligently, but also have to translate that thinking into physical action in a complex environment."

The virtual rat's artificial neural network was trained to implement inverse dynamics models, which scientists believe our brains use to guide movement. Graduate student Diego Aldarondo worked closely with DeepMind researchers to feed the network reference trajectories of desired movements, allowing the virtual rat to imitate a diverse range of behaviors, even those it hadn't been explicitly trained on.

These simulations may pave the way for a new field of virtual neuroscience, where AI-simulated animals, trained to behave like real ones, provide transparent models for studying neural circuits and how they are affected by disease. While Ölveczky's lab is interested in fundamental questions about brain function, the platform could also be used to engineer better robotic control systems. 

Looking ahead, the researchers aim to give the virtual rats autonomy to solve tasks akin to those encountered by real rats. "From our experiments, we have a lot of ideas about how such tasks are solved, and how the learning algorithms that underlie the acquisition of skilled behaviors are implemented," Ölveczky said. "We want to start using the virtual rats to test these ideas and help advance our understanding of how real brains generate complex behavior."