Neural Network Based Pedestrian Dead Reckoning Navigation

This SBIR Phase I project will develop a prototype system for pedestrian dead reckoning navigation system. The proposed system is based on a novel deep neural network and does not require any pre-defined dynamic models. Hence, this system is significantly more robust to nonlinearities in the dynamic models than the traditional Kalman filter variants. The system is capable of handling uncertain and noisy data and, therefore, is suitable for low-cost MMES-IMU on mobile devices. The project will benefit from Novateur Team’s expertise in the areas of deep neural networks, pedestrian dead reckoning systems, localization in GPS-denied environments, nonlinear filtering, time-series analysis, and context-based model selection.