A robust variance information fusion technique for real-time autonomous navigation systems

Real-time navigation systems rely on multi-object environment information for productive routing and assisted movements of vehicles. The multi-object environment includes infrastructure, neighbours, smart building, and traffic management information for providing assisted driving for the vehicles. The dynamic vehicle environment and sensor failures impact the navigation assistance for the driving users. A robust variance-based information fusion (VIF) technique is proposed in this paper for addressing this issue. The proposed technique makes use of swappable sensor information for fusion for providing reliable navigation assistance. This means the fusion process is performed based on the active sensor information in a decisive manner. For effective fusion with available information, this technique makes use of classification learning. Through this learning process, the error causing information due to sensor failures is identified and mitigated from the fusion. Multi-sensor information classification is performed for performing errorless assistance and decisions. This process is linear throughout the information sensing time intervals for reducing the errors in navigation assistance. The proposed technique's performance is verified using experiments, and the metrics processing time, error, accuracy, and swapping instances are verified.