Vision-based cooperative estimation via multi-agent optimization

In this paper, we investigate a cooperative estimation problem for visual sensor networks based on multi-agent optimization techniques. A passivity-based visual motion observer is employed as a tool to meet the objective. We first give an interpretation of the visual motion observer from the viewpoint of optimization and present new inputs motivated by the optimization techniques on manifolds. Based on the investigations, we formulate a novel cooperative estimation problem to be tackled. Then, a cooperative estimation algorithm is presented based on multi-agent optimization techniques. Finally, the effectiveness of the present algorithm is demonstrated through experiments.