Optimizing collective fieldtaxis of swarming agents through reinforcement learning

Swarming of animal groups enthralls scientists in fields ranging from biology to physics to engineering. Complex swarming patterns often arise from simple interactions between individuals to the benefit of the collective whole. The existence and success of swarming, however, nontrivially depend on microscopic parameters governing the interactions. Here we show that a machine-learning technique can be employed to tune these underlying parameters and optimize the resulting performance. As a concrete example, we take an active matter model inspired by schools of golden shiners, which collectively conduct phototaxis. The problem of optimizing the phototaxis capability is then mapped to that of maximizing benefits in a continuum-armed bandit game. The latter problem accepts a simple reinforcement-learning algorithm, which can tune the continuous parameters of the model. This result suggests the utility of machine-learning methodology in swarm-robotics applications.