Models and simulations of collective motion in biomimetic robots and bacteria

Abstract

In nature, one finds many examples of collective motion, from flocking birds to swarming bees. Any one organism makes its decisions based solely on local information; either it can sense what its close neighbors are doing, or in the case of a single-celled organism, it can sense some local property of its environment. Yet complex global behaviors arise from these local interactions, and these large-scale patterns have neither a leader nor any other centralized control system. In this thesis, two specific cases of collective motion are studied: fish schooling and bacteria swimming across a surface. When fish swim in schools, they swim in the same direction as each other at approximately the same speed. Previous studies of fish have discovered three primary behaviors that, together, lead to large-scale coordination and schooling in the animals. This thesis demonstrates that the same algorithms can be applied to a group of identical underwater robots. If the robots need to coordinate with each other, they can use biomimetic control laws and adopt the interaction algorithms used by fish. A series of simulations are run to see what possible group behaviors can come from these control laws. At a smaller scale, prior experiments have revealed that bacteria and other small organisms also show collective motion.

(cont.) Unlike fish, bacteria cannot see their neighbors; the individual can only sense the bulk contribution of its neighbors to the flow at its location. The single-celled organisms are small and swim slowly, so they have very small Reynolds numbers. They are modeled in this work in a Stokes flow regime; the model is built bottom-up starting from the hydrodynamic field created by one organism and then superimposing these fields on top of each other. Different possible control policies are tested where each organism has an instantaneous desired direction based on some local property of the flow. While simulations of the current model do not yield results that fully emulate real bacteria, they have some similarities and provide insight into the complex hydrodynamic interactions between low Reynolds number swimmers.