IDENTIFICATION OF TARGET PRIMITIVES WITH MULTIPLE DECISION-MAKING SONARS USING EVIDENTIAL REASONING

In this study, physical models are used to model reflections from targ et primitives commonly encountered in a mobile robot's environment. Th ese targets are differentiated by employing a multi-transducer pulse/e cho system that relies on both time-of-flight data and amplitude in th e feature-fusion process, allowing more robust differentiation. Target features are generated as being evidentially tied to degrees of belie f which are subsequently fused by employing multiple logical sonars at geographically distinct sites. Feature data from multiple logical sen sors are fused with Dempster's rule of combination To improve the perf ormance of classification by reducing perception uncertainty. Using th ree sensing nodes, improvement in differentiation is between 10% and 3 5% without false decision, at the cost of additional computation. The method is verified by experiments with a real sonar system. The eviden tial approach employed here helps to overcome the vulnerability of the echo amplitude to noise, and enables the modeling of nonparametric un certainty in real time.