Adaptive knowledge-based system for personal navigation in GPS-denied environments

This paper presents a design, prototype implementation and the performance analysis of a personal navigator based on multi-sensor integration, augmented by the human locomotion model that supports navigation during GPS signal blockage. The project is sponsored by the National Geospatial-Intelligence Agency (NGA), and aims at the development of a prototype of a personal navigator supporting navigation and tracking of ground military and rescue personnel. The accuracy requirement is considered at 3-5 m CEP (circular error probable) level. At the current stage of the research, the algorithmic concept of the GPS-based, Micro-electro-mechanical inertial measurement unit (MEMS IMU)-augmented personal navigator system with an open-ended architecture has been implemented. This paper presents the design architecture of the integrated system and the performance analysis, with a special emphasis on the navigation during the loss of GPS signals. The system architecture, supported by GPS, MEMS IMU, digital compass and a barometer is currently designed to incorporate the dynamic model of human locomotion. The system is trained during the GPS signal reception using Radial Basis Function (RBF) neural network with up to six input parameters, and is subsequently used to support navigation in dead reckoning mode when GPS signals are blocked. The calibrated, operator-dependant model of the stride length, provided by the neural network-based adaptive knowledge-based system, and heading information from the compass/LMU offer dead reckoning navigation that facilitates bridging of the GPS gaps. In this paper, the process of calibration/personalization of the human locomotion model using the current hardware prototype is presented. The calibration results are subsequently tested in the dead-reckoning mode to assess the quality and reliability of the adaptive knowledge-based system.