UC Berkeley

This dissertation addresses various aspects of realizing a three-dimensional (3D) controlled flock of swimming micro-robots that operate in, and cooperatively influence, viscous fluid environments. A systematic approach is then presented to equip the agents with an adaptive decision-making intelligence, so as to enable flocks of these artificially intelligent swimming micro-robots to achieve various objectives in the presence of flow-mediated interactions.

In the first part of this dissertation, we introduce a versatile swimming robot with full 3D maneuverability in viscous environments. The experimental realization of this artificial low-Reynolds swimmer is then reported, and a hierarchical control strategy is implemented to perform various swimming maneuvers. The major challenge, which makes the swarm-control of swimming micro-robots substantially different from other well-studied swarms, is the presence of long-range flow-mediated (i.e. hydrodynamic) interactions. Therefore, the second part of this dissertation is devoted to the investigation of swarm hydrodynamics, including mutual interactions between these micro-swimmers, and their behavior in vicinity of solid boundaries. In particular, we unveil orbital topologies of interacting micro-swimmers, and report diverse families of attractors including dynamical equilibria, bound orbits, braids, and pursuit-evasion games. The third part of this dissertation is focused on optimal swarm-control strategies for swimming micro-robots to achieve various objectives in the presence of flow-mediated interactions. We show that micro-swimmers can form a concealed swarm through synergistic cooperation in suppressing one another's disturbing flows. Various control schemes are then demonstrated for the concealed swarming and stealthy maneuvers of swimming micro-robots. We also discuss how state-of-the-art reinforcement learning algorithms can be used to realize flocks of artificially intelligent swimming micro-robots. Specifically, a systematic approach is presented to equip the swimming micro-robots with an adaptive decision-making intelligence in response to non-linearly varying hydrodynamic loads. Flocks of these artificially intelligent micro-swimmers are then deployed to actively cloak swimming targets in a crowded environment. This study provides a road-map toward engineering cooperative flocks of smart micro-swimmers capable of accomplishing a new class of group-objectives. We, therefore, hope that it will spur further research on this field at the intersection of fluid mechanics, robotics and artificial intelligence.