The Swarm Application Framework

Publications and Documents

• ``Predicting and Controlling System-Level Parameters of Multi-Agent Systems by Don Miner and Marie desJardins. AAAI Fall Symposium on Complex Adaptive Systems and the Threshold Effect, 2009
  Boid ßocking is a system in which several individual agents follow three simple rules to generate swarm-level ßocking behavior. To control this system, the user must adjust the agent program parameters, which indirectly modiÞes the ßocking behavior. This is unintuitive because the properties of the ßocking behavior are non-explicit in the agent program. In this paper, we discuss a domain-independent approach for detecting and controlling two emergent properties of boids: density and a qualitative threshold effect of swarming vs. ßocking. Also, we discuss the possibility of applying this approach to detecting and controlling trafÞc jams in trafÞc simulations.
• ``Learning Non-Explicit Control Parameters of Self-Organizing Systems" by Don Miner and Marie desJardins. Third IEEE International Conference on Self-Adaptive and Self-Organizing Systems (SASO 2009).
  Abstract-Controlling self-organizing systems with given control parameters is often unintuitive and inefficient for the user. Typically, there is some emergent behavior that the user would like to produce that is not directly controllable. Our goal is to develop an autonomous and domain-independent learning framework for adding non-explicit control parameters to self-organizing systems that provide users with more intuitive real-time control of the system. These additional controls are created by using regression to learn a mapping between the explicit control parameters and the non-explicit control parameters.
• ``Controlling Particle Swarm Optimization with Learned Parameters" by Kevin Winner, Don Miner and Marie desJardins. Third IEEE International Conference on Self-Adaptive and Self-Organizing Systems (SASO 2009).
  Controlling particle swarm optimization is typically an unintuitive task, involving a process of adjusting low-level parameters of the system that often do not have obvious correlations with the emergent properties of the optimization process. We propose a method for controlling particle swarm optimization with non-explicit control parameters: parameters that describe self-organizing systems at an abstract level. Effectively, this process converts intuitive control parameter values into explicit configurations that particle swarm optimization can directly apply. In this paper, we introduce the motivation, methodology, and implementation of our approach.
• ``Rule Abstraction: Understanding Emergent Behavior in Swarm Systems" ~ PhD Dissertation Proposal.
  In this dissertation proposal, I present a framework named, Rule Abstraction, for learning how swarm systems will behave, given their operating parameters. Frequently when studying swarm systems, one has an understanding of the low-level local behaviors as well as an understanding of the high-level emergent behaviors. What is not typically understood is how these levels relate to one another. Rule abstraction is an intuitive new tool that we propose for implementing and controlling swarm systems. The methods presented in this proposal encourage a new paradigm for designing swarm applications: engineers can interact with a swarm at the abstract (swarm) level instead of the individual (agent) level. This is made possible by modeling and learning how particular abstract properties arise from low-level agent behaviors.
• ``Understanding the Brain's Emergent Properties" (Position Statement) by Don Miner, Marc Pickett, Marie desJardins. AGI '09.
  In this paper, we discuss the possibility of applying rule abstraction (presented as a method designed to understand emergent systems) to the physiology of the brain.
• ``The Swarm Application Framework" by Don Miner, Marie desJardins, and Peter Hamilton. AAAI '08 Student Abstract.
  A short (2 pages) introduction on SAF, which discusses motivation and architecture.

Presentations

• Poster: ``Learning Non-Explicit Control Parameters of Self-Organizing Systems"
  Presented at the 2009 UMBC CSEE Research Review and ICML 2009 Student Scholarship Program.
• ``Rule Abstraction and Future Directions" 9/23/08
  Briefly gives an overview of the motivations of Rule Abstraction and then reviews some preliminary results in toy domains. Also gives a preview of my current, and more complicated, Boids domain in which I plan on receiving some interesting results.

Other Output

• Google Code project
• 3D Flocking boids Python script.
• Some interesting videos that pertain to this project.

Old Presentations

• Poster: ``Learning Abstract Rules for Swarm Systems" Spring 2008
  An overview of the motivation, applications, and initial experimental results of learning abstract rules for swarm systems. This discusses the structure of the rule abstraction learning problem and how we tested the mechanism with the simple form circle domain.
• ``Abstract Rules for Swarm Systems" 4/25/08
  Gives an short overview of rule abstraction, abstract rule learning, and the form circle experiment.
• ``The Swarm Application Framework, Rule Abstraction and Future Directions" 3/5/08
  Briefly talks about the fundamentals of SAF and Rule Abstraction and then provides a discussion of possible future directions.
• ``Swarm Intelligence and The Swarm Application Framework" 2/25/08
  Gives an introduction to Swarm Intelligence and gives an overview of SAF. A talk given to Sue Evans' freshman seminar class at UMBC.