Researchers from Harvard University in collaboration with the artificial intelligence laboratory Google Deep Mind have created a virtual model of a rat with an artificial brain, capable of imitating movements similar to movements of a living animal. The model was created to better understand how the brain controls movement, writes Interesting Engineering. Despite all the achievements, modern robotics has not yet been able to reproduce the natural movement of animals and humans, as challenges exist both in hardware and software.
Together with Professor Bens Helvecki and other researchers from Harvard and Google DeepMind, scientist Diego Aldarondo developed a biomechanically realistic digital model of a rat. The team used the MuJoCo physics simulator and the Motor Imitation and Control (MIMIC) system to train artificial neural networks on rat behavior using high-resolution data recorded from real animals.
Using artificial neural networks, researchers were able to create models of inverse dynamics, which the human brain uses to control body movements and allows the transition from the current state of the body to the desired one. Data from real rats helped the virtual model learn the forces needed to achieve the desired movement, even if it hadn't been trained on them.
When scientists measured neural activity in both real rats and virtual model, the researchers found that the virtual model accurately predicted the neural activity of real rats. «This opens up a whole new world of virtual neuroscience, where animals simulated with the help of artificial intelligence can be used to study neural circuits and understand , how they are damaged during diseases», — Harvard scientists believe.