System Engineering: Biological Swarm Based Distributed Systems Design

Future NASA missions will be multi-functional in nature, utilize large networks of smaller spacecrafts and sensors, and operate under non-stationary, dynamic, and energy-constrained environments. As a result they will be increasingly complex to manage and operate. The complexity will appear at the frontiers of communication, control, navigation, and operation. Such large-scale distributed systems, deployed far from Earth, and with little or no direct human intervention in situ, cannot and will not be efficiently configured and run by a team of operations personnel, or even a complex set of programming rules. The traditional approach of having a team continuously monitor the system, or update the rules manually whenever an unforeseen change occurs is simply not cost-effective, beyond human capabilities, and impossible in high-latency scenarios when split-second decisions need to be made.

The critical issues for future NASA missions will involvereconfiguration, adaptation, and dynamic resource allocation for remote, autonomous operation in unknown environments. Examples arecollective and cooperative mobile robotic systems, autonomous distributed spacecraft missions, and ad-hoc sensor networks for planetary exploration.

The distributed systems solution to these challenges lie in the area of biological swarms (distributed biosynthesis, and collective and cooperative behavior). This involves dynamic and adaptive configuration and operation of a system via distributed intelligence (``the system is the intelligence''). By developing and applying new biosynthesis techniques, and bio-inspired algorithms tailor-made to space operations and limitations, we can synthesize emergent intelligence of the type seen only in biological organisms, and dynamically adapt various parameters of a distributed system, to meet listed requirements. Such a fully autonomous system, capable of intelligent behavior, self-diagnosis, and repair, will result in a robust and efficient operation with high reliability.

Within this area, research challenges involve two inter-related and complementary problems:

1. The Reverse-Engineering (Micro) Problem: Development of a model that accurately characterizes the synthesis properties and behavior of a class of bio-organisms (such as flies or bees). This will include the mechanical implementation of a fly-like robot. Specifically this model must encompass algorithms that involve social two-way communications and interactions between and among two organisms, including courtship, mating, swarming, collective searching, etc. The goal here is to determine the organism's inner ``micro'' workings and information processing mechanisms based on its observed and stimulated behavior.

2. The Forward-Engineering (Macro) Problem: Development of distributed, dynamic resource allocation mechanisms based on the above model, and observed processes in single organisms, and mobile biological swarm colonies. The goal here is to look at the ``macro'' behavior of biological swarms, and consider numerous manifestations of their emergent intelligence.

Over the past three years we have made significant inroads in both of these areas in collaboration with Caltech, the University of Washington, and Western New England College. Our research and development goals have been and will continue to be

• Learning how control, communication, and navigation systems actually co-evolve in living systems, and give rise to highly robust and complex systems with many interacting parts.

• Predicting the emergence of controlled actions, behavior, and communication from such a complex system.

• Creating robust, reliable systems out of many simple components, leveraging understandings of bacterial ecosystems, and collective behavior of insects and biological swarms.

We have been focusing both on fundamental theory, and implementation and experimentation. A mobile distributed system testbed is currently under development both at the UW and WNEC. This is in addition to the Dickinson Lab at Caltech. We plan on hosting a workshop on Jan. 30-31, 2004 to develop a roadmap for NASA's future investments in this area as well, with participation from our collaborators and other researchers in the field, drawn from NASA and academia