Discipline: Technology and Engineering
Subcategory: Computer Science & Information Systems
James Stigall - Bowie State University
Co-Author(s): Sharad Sharma, Bowie State University, Bowie, MD
A Multi-Agent System (MAS) is a simulated environment featuring several agents that mimic real-life behavior. They are useful in a wide array of applications, especially in evacuation planning where they are used by psychologists, architects, and emergency personnel to study evacuation patterns. An effective MAS is one that incorporates specific emotional attributes, specifically stress and panic. We have developed a Java-based MAS where agents possess four different types of behaviors –goal following, follower, panic, and hostile. Agents following goal following behavior evacuate in calm, rational manner while those possessing the follower behavior follow the closest agent to the exit. The panic and hostile behaviors are the most interesting to us because evacuees with those behaviors can affect the evacuation flow of other evacuees the most as those agents are designed to evacuate in an irrational manner.
The panic and hostile agents were modified using Eclipse, a Java integrated development environment (IDE), so that their evacuation speeds are determined by fuzzy logic reasoning, which incorporates rules using factors such as the number of neighboring agents, and distance for neighboring agents, and level of panic.
Our hypothesis is that, in our proposed MAS, highly panicked agents will more likely be stuck and will have lesser evacuation times. The results of our previously-developed MAS accurately predicted evacuation times for agents. Future work includes generating more rules for panic and stress as well as testing them out in scenarios involving numerous obstacles.
Not SubmittedFunder Acknowledgement(s): The authors would like to thank the National Science Foundation for supporting the project. This work is funded by the Grant Award number HRD-1238784. The authors would also like to acknowledge the support of the Virtual Reality Laboratory at Bowie State.
Faculty Advisor: Sharad Sharma, ssharma@bowiestate.edu