Learning protocols for fast and efficient state-to-state transformations in active matter

We show that it is possible to learn protocols that effect fast and efficient state-to-state transformations in simulation models of active Brownian particles. By encoding the protocol in the form a neural network, we can use evolutionary methods to identify protocols that bring a collection of active Brownian particles from a passive phase to an active one, as quickly as possible or with as little work done as possible. The learning scheme can in principle be used in experiment, suggesting a way of designing protocols for the efficient manipulation of active matter in the laboratory.