Researchers Create Innovative Computer Programs Based on Genetics, Insect Behavior A computer-science professor and two student researchers are looking to nature for inspiration on two high-tech research projects funded by a $10,000 Whitworth Science Faculty/Student Research Grant.
Assistant Professor of Math and Computer Science Susan Mabry and senior computer science major Steve Kollmansberger are breaking new ground as they develop a DNA multi-agents system which involves a new approach to computer programming that mimics genetic code and Darwinian biological processes.
"Steve has developed a system that generates and optimizes agent code that adapts to changing environmental demands," Mabry says. "For example, it provides the opportunity for a robot to learn and adapt to an environment, devising the best way to perform its assigned tasks."
Kollmansberger has developed an extensive language that mutates to continually improve the surviving agents. This system is programmed to execute in parallel across multiple computers. Test case applications of the team's research include a "herbivore/carnivore world" consisting of herbivore and carnivore agents which try to survive and travel among multiple environments containing water, poison herbs and safe herbs.
"The states are continually changing; herbivores search out only good herbs to eat, and to make things interesting, carnivores eat herbivores," Mabry says. "Both types of agents endeavor to satisfy their hunger and maintain good health."
The application demonstrates strengths such as a dynamic population, significant variability, interactive effects and multiple scoring criteria, all of which can be applied to challenging, real-world problems. Research results are being submitted as conference papers and computer-science students are in the process of embedding the system in robots.
Mabry and senior computer science major Aaron Kangas are tackling another area of research based on a behavior model of an ant colony, which displays a natural order of organizational management through foraging for food and coordinating movement by emitting chemical secretions known as pheromones to create trails.
"The ant colony comprises individual ants which display collective behaviors," Mabry says. "Aaron's work involves creating representative modeling in a computer program and implementing ant-colony movements in a parallel environment with accompanying graphics."
Upcoming Whitworth computer-science classes will use research results and robotic equipment from both projects to further their studies.
Math & Computer Science Home Page
Department Spotlight >