The abstraction of agents acting in environments is a common method used to simplify and reduce complex systems into more manageable pieces which can in turn be broken down, recursively iterating until a sufficient understanding has been developed which can describe the original system. The subject of this thesis focuses on the development and use robotic platforms which abstract two unique systems of active granular matter.
The first system is inspired by biological observations made while studying colonies of the species Solenopsis invicta, commonly known as the fire ant. Colonies of fire ants build shelters by excavating extensive tunnel networks underground, though not all members of the colony participate in this nest construction equally. Hypotheses have been developed, and robotic implementations have been tested in the lab which have probed the effects of heterogeneous behavior distributions within a task-oriented ensemble, though the capabilities of the robots were found to be somewhat limited. A new robotic model with improved navigation and sensory capabilities was needed to further investigate the aspects of the fire ant ensembles which may inform and govern the individual agent behaviors in order to optimize the collective operation.
Granular media are often investigated in the standard context of spherical, convex particles, though atypical, nonconvex particles have been demonstrated to exhibit interesting entanglement dynamics which demonstrate a rich and diverse physics which may be exploited by biological systems such as the assemblages of social insect assemblies. The second system discussed pertains to the development and use of another confined active matter composed of particles which phase between convex and nonconvex geometries.