Signal processing architecture with bidimensional network topology for flexible sensor data integration system

The authors describe a new sensor data integration architecture that utilizes a physical network paradigm. The authors have realized a prototype system on a parallel processor network using Inmos's transputers. Simple demonstrations of 2-D shape recognition and target tracking are shown and the authors have demonstrated its ability to respond appropriately to varying environmental constraints on the structure of the signal flow. It is shown that one can gain higher reliability in the architecture using the flexibility of the signal flow topology.