Sequential Task Allocation in Diverse Robotic Swarms
PhD candidate in Mechanical Engineering
Barry Trimmer, Biology
For decades, conventional robotics has focused on developing monolithic machines that are able
to perform a single task or a related group of tasks. The growing body of research on "swarm robotics"
uses an alternative paradigm: a large number of inexpensive and easily replaced robots working cooperatively
to accomplish a task. In swarm robotics, novel algorithms utilize input from all of the members of the swarm
to develop a picture of the environment in which the swarm exists, and to determine how the swarm behaves
within that environment.
Swarm robotics research is a natural fit for soft robotics, because one of the promising aspects of soft
robotics is the ability to cheaply and easily produce large numbers of individual units using casting or
3D printing. Just as soft robotics utilizes biological models to develop morphological and control structures,
swarm robotics examines the behavior of social species, such as ants, bees, and fish, to develop algorithms
that have useful applications in a robotic context.
I am particularly interested in how swarms allocate their members to execute serial and parallel tasks.
The initial inspiration for this line of research came from a paper that developed a simple algorithm for
partitioning a swarm between two tasks, based on observations of how leaf cutter ants allocate themselves to
"harvesting" and "storing". Harvesters climb a tree to the leaf canopy, where they cut leaves into small pieces
which flutter to the ground. Storers collect the fallen leaf bits and transport them to the colony's storage area.
Dividing the labor in this way makes the entire colony more efficient, because only the harvesters expend energy
to climb the tree, and a single trip up the tree generates many leaf cuttings.
My current line of research aims to expand this algorithm so that it can deal with more than two tasks. Future
directions involve the development of methods for handling heterogeneous swarms, and exploring the roles that
global knowledge and communication play in the ability of a swarm to allocate tasks efficiently.