Sequential and parallel models in noncooperative games for decentralized decision making in vision systems

An examination is made of two models, one parallel and the other sequential, of decentralized decision-making in modular systems aimed at vision or potentially robotic tasks. The theoretical analysis and the experimental results from a two-module image analysis system reveal the critical dependence of the system behavior on the exact decision model. The experiments with the two-module vision system indicate that the parallel model leads to more robust performance for a certain class of multiple objective systems