DISTRIBUTED OPTIMIZATION BY WEIGHTED ONTOLOGIES IN MOBILE ROBOT SYSTEMS

Heterogeneous mobile robots are often required to cooperate in some optimal fashion following specific cost functions. A global cost function is the sum of all agents' cost functions. Under some assumptions it is expected that a provable convergent computation process gives the optimal global cost for the system. For agents that can exchange ontological information via a network, different variables in the global vector are relevant when ontology instances have been recognized and communicated among agents. It means the optimization depends on what is known to each agent at the current time. There are two ways to solve an optimization of this kind: (a) to weight agents according to the ontology instances or (b) to add ontology-defined optimization constraints. This paper illustrates the benefits of the weighted optimization method.