Flow theory for flocks of autonomous mobile physical agents

Abstract

This research is a study of fuzzy rules which could be used to model autonomous entities and the development of a theory to describe the traffic flow of autonomous mobile physical agents. Pedestrians, autonomous vehicles, or terrain-searching robots could be modeled as autonomous mobile physical agents. For this research, an object-oriented architecture was developed to implement simulations of agents for analyzing their emergent behavior. In an agent system, agents execute their own set of rules during each iteration. The architecture in this research is a framework that can be used for experimenting with variations of rule sets to assist in discovering a rule set that results in desirable system-level behavior. The architecture was used to simulate a flock of mobile agents which navigate in 2-dimensional space. The output of the simulation (effective speed and collision probability) was regressed to develop flow theory equations to calculate the expected output as a function of several variables that affect the agent system. The flow theory equations describe the emergent behavior of the system, or system-level behavior resulting from interactions at the level of the parts of the system. The fuzziest rule set consistently gave the best results in terms of collision avoidance for this study. The improved collision avoidance is gained at the cost of a lower effective speed. However, the fuzziest rule set also has more complex rules, which means it takes longer to execute. In systems with agents traveling in scattered patterns, there were potential collisions among agents even at low utilization rates. This may indicate that a “traffic control” policy may be desirable when congestion reaches a limit.