Abstract

Distributed assembly tasks, in which large numbers of agents collaborate to produce composite objects out of component parts, require careful algorithm design to ensure behavior that scales well with the numbers of agents and parts. Yet al-gorithm evaluation, through which design is guided, is com-plicated by the combinatorial nature of system states over the course of execution. This leads to a situation in which the algorithm design space is often severely cramped by the inefficiency of available analysis techniques. We review sev-eral available analysis strategies, and present two techniques for designing distributed algorithms that lend themselves to continuous differential analysis while avoiding catastrophic deviation between discrete and continuous system models. This methodology aims to allow optimization at the macro continuous level to inform parameter choice for discrete, real world systems. Categories and Subject Descriptors I.2.11 [Distributed Artificial Intelligence]: Multiagent systems