Abstract
Multi-agent coordination has been researched in a diverse multitude of domains, but the complexities of the physical world necessitate further study to engineer autonomous systems capable of addressing challenges within it. Teams of submersible robots, for instance, have the additional constraint of a reduced communication range as radio waves do not travel as far underwater as they do on land. The need for communication is also further accentuated for heterogenous teams as agents must coordinate tasking based on their varying capabilities. In an environment that fits this description, agents must be reliably independent yet adaptively opportunistic in their task scheduling depending on the proximity of cohorts in their immediate area. To study multi-robot coordination with the physical world limitations, an integrated platform for developing collaborative virtual robots utilizing goal-oriented task scheduling in a simulated physical world environment is presented. The agent world-perceptive layer, a novel component, is introduced to model each robot’s special sensor and transmitter capability and constraints in the physical world. This grants the ability to examine the effects of worldly limitations on agents using traditional, non-lossy communication frameworks such as the one used in the Java Agent DEvelopment (JADE). Finally, a non-centralized approach to multi-agent coordination, where each agent makes its own decisions based on its local knowledge and information is presented. Low-level task scheduling is performed with respect to high-level goals using the Task Analysis, Environmental Modeling, and Simulation Framework (TAEMS). The combination of this goal-based task scheduling approach with adaptation for opportunistic collaboration within a communication-constrained environment represents a unique contribution to the field of multi-agent coordination.