Active particles using reinforcement learning to navigate in complex motility landscapes

Abstract As the length scales of smallest technology continue to advance beyond micron scale it becomes increasingly important equip robotic components with means for intelligent and autonomous decision making limited information. With help a tabular Q-learning algorithm, we design model training microswimmer, navigate quickly through an environment given by various different scalar motility fields, while receiving amount local We compare performances defined via time first passage target, suitable reference cases. show that strategy obtained our reinforcement learning indeed represents efficient navigation strategy, outperforms By confronting swimmer variety unfamiliar environments after finalised training, generalises classes random fields.