This project constitutes a systematic investigation of
biologically inspired principles for prescribing and analyzing
capabilities, roles, and control strategies for different individuals in
teams of autonomous agents. In particular, the role of
heterogeneity is studied from a biologically inspired
vantage-point in order to enable teams to respond to environmental changes
in robust, effective, safe, and predictable ways.
We propose to shift focus from the traditional
coordination problem for networked, autonomous agents - focusing on how control and coordination strategies should be
designed for a particular set of objectives and tasks - to the question of what these objectives and
tasks should be? For example, we know how to do formation control, but what formations should we
use in the first place? Similarly, we know how to assign roles to different agents and incorporate
heterogeneity into teams. However, we are yet to study what heterogeneous roles are
beneficial given a target application? Under the banners of bio-inspired and bio-mimetic multi-agent
robotics, we propose to investigate if there is an argument to be made for having robots with different
capabilities and characteristics working together, such as slow robots and fast robots, in
real-world settings. We focus explicitly on this shift of vantage-point using biologically
inspired principles for designing and evolving networks of autonomous agents. In
particular, the following key, inter-connected research questions will be pursued:
Temporal Heterogeneity: Robots operating at different
time scales will be studied. In particular, the SlowBots will be introduced as a way of making explicit
the benefits of using slow robots in conjunction with fast ones. Tree sloths, lorises, and
natural ecosystems will be used as inspiration for this work.
General Heterogeneity Measures: Ultimately, a general
theory of heterogeneous teams must be developed - driven by biology but drawing inspiration from other
fields, such as economics and sociology. This will provide such a general framework in the setting on
teams of autonomous agents.
Heterogeneous
Multi-Robot Teams: A number of different aspects
of heterogeneous multi-robot teams will be investigated in order to cover, characterize, and
understand heterogeneity along different dimensions.
This
project is a collaboration between three Principle Investigators: two at the
Georgia Institute of Technology, and one at the University of Pennsylvania.
Under Dr. Ronald Arkin, the Mobile Robot Lab’s focus is on the temporal heterogeneity
component. Specifically, the methods, challenges, and benefits of
incorporating intentionally slow robots (SlowBots)
into a team. Two veins of research have been explored. Based on existing
ethological literature on sloths and slow lorises, ethograms highlighting
the various behaviors of sloths and slow lorises were drawn
and behaviors relevant to the design of a robotic controller were
identified. For each of the behaviors, different
behavioral coordination mechanisms have been explored, namely, Action
Selection vs. Behavioral Fusion or a combination thereof.
SlowBots, as slow and persistent agents in their environment,
experience significant changes in their environment. While a quadrotor running
for thirty minutes can safely ignore the daily solar cycle, it is a
significant factor for a SlowBot operating for
weeks in the field. To address this challenge for SlowBots,
an artificial circadian system has been designed to allow for a SlowBot to model the dynamics of its environment, predict
how it will continue to change in the future, and thus adapt its behavior
to the changes.
|