Favorite Goal in Agent Based Modeling and Simulation

摘要

Unlike robots, humans have a preconceived notion of the path they want to follow. They are more likely to go towards a goal that they have seen or they are familiar with instead of looking for the shortest path. Our approach models both individual behavior and group behavior in a multi-agent system. Individuals constantly adjust their behavior according to obstacles present in the environment. We hypothesize that humans are more likely to go towards an exit they have seen instead of looking for the shortest path to an exit. If there is an exit they have not seen but it is very close to them, they are unlikely to go towards it. This paper presents a favorite goal algorithm for simulating emergency evacuation scenarios in a multi-agent system. Our approach addresses the issues of agent characteristics such as collaboration, cooperativeness, learning ability, and level of panic.