Dynamic reconfiguration of parallel sensor integration system by model estimation

Abstract

Model-based sensor integration is a practical way of using multiple sensor information. The approach, however, cannot be applied when the model is unavailable. In this paper, we propose a dynamic reconfiguration mechanism of parallel sensor fusion network by model estimation. Our estimation algorithm recognizes the target mechanism from the state transition rules of relative motion between rigid bodies. Hierarchical grouping of sensor data realizes this idea. The approach is applied to a 2D-kinematic planar system, and the result is described.<>