Jisan
Research
Institute

" A pioneering research community "

Swarm research pushes forward at JRI with a new paper

on

Dexin Jin (17) from La Cañada High School in La Cañada-Flintridge, CA and Mike Peiyao Li from San Marino High School in San Marino participated in a research project that developed a novel way of developing swarms of autonomous agents. During the project, the students and their Research Mentor developed a new method of designing swarms of autonomous agents. The largely theoretical project involved the development of a theoretical framework for swarms based on work originally completed in 2006. The group compiled their results in a scientific paper, submitted it to the IEEE-sponsored Sixth International Conference on Swarm Intelligence (ICSI 2015), and were invited to Beijing, China to present their research results. Dexin delivered the presentation in Beijing.

Swarms are well known phenomena in nature and can be found nearly everywhere where there is animal life, and some places where there isn't. Bees, ants, termites, fish, sharks, zebra, and geese all form swarms in one way or another when they come together. The swarms are, generally speaking, groups of individuals that in some way interact with one another in an ongoing manner, affecting one-another through their interactions. What makes swarms interesting to many scientists and engineers is the ability of the group to exhibit properties and capabilities that the individuals in the swarm do not have independently. This means that the swarms can be made to do things that are complex, even if the agents that make them up are relatively simple.

Dexin Jin giving a presentation at the 6th International Conference on Swarm Intelligence in Beijing, China.

 

Despite an initially intense interest in intelligence exhibited by swarms and an ongoing interest in using swarms for military applications, most researchers cannot yet design a swarm using a set of theoretical tools that can guarantee that the swarm will do what it's supposed to do once it's been designed. Rather, researchers and engineers are forced to use a “guess-and-check” method in which they create a swarm based on a best guess and gradually modify the agent behaviors so as to eventually generate the desired global behavior. This means that if a swarm is going to build a wall, the engineer will have to go through several design cycles in order to build a swarm that constructs exactly the right wall with precisely the right dimensions and properties.

Jin and Li and their Research Mentor's research is aimed at solving this problem. The method creates a theoretical construct called a phase space that represents what is happening in the real world. The problem is then translated to points in this space. For instance, if the phase space is represented on a two dimensional map, the task might be represented by two points on the map. One of the points would indicate the state of the system and the other point would indicate the desired state of the system. Any pathway on the map between the points would be a solution to the problem. What makes this method so potentially valuable is that a research or development team would have mathematical proof that the swarm they constructed, which generated movement along the path, would accomplish the task. The method was theoretically applied to a problem called the “stick pulling problem”.

The paper is entitled "Utilizing abstract phase spaces in swarm design and validation" and is available here.

Both Jin and Li are continuing to work on this research with their Research Mentor. The duo has added members to their group and the combined group, now including students from six different high schools in Southern California, is continuing to improve their methodology. The group is working on a paper expected to provide theoretical backing to another area of swarm innovation known as physicomimetics in a research effort that incorporates a collaboration with a scientist from Indiana University.

The Jisan Research Institute · A California 501(c)3 Nonprofit Corporation · © Copyright 2009-2014 · All rights reserved