You see, most artificial microswimmers can only do simple moves that are hard to control. But Pak and the other researchers theorized that, like humans, the microrobots could learn to swim better through reinforcement—basically by telling them “good job.” They combined this reinforcement learning with artificial neural networks that mimic how the human brain operates. Through the machine learning approach, they successfully taught a simple microrobot how to swim and navigate in any possible direction.
“Our research demonstrates machine learning as an alternative avenue towards designing the next generation of smart microswimmers that can navigate complex environments for future biomedical applications,” Pak says.
In Santa Clara style, one of the study’s lead authors was a student. “It has been a truly transformative experience working with Professor Pak and our collaborators to merge knowledge and techniques across different disciplines in this work,” says Zonghao Zou M.S. ’22. “This valuable research experience in my master’s studies is an important stepping stone in pursuing my academic career.”