Artificial-Intelligence-Assisted Geomagnetic Navigation Framework

Over the past few decades, the paradigm of geomagnetic navigation has been under consideration as a potential alternative to GPS in situations where GPS signals are subject to interference or are inaccessible. However, navigation methodologies that rely on magnetic fields continue to face practical development challenges due to their complex characterization, which has so far resulted in modest localization performance compared to GPS. This article presents the design and development of an innovative intelligent geomagnetic-based framework for navigation within GPS-denied environments. Two artificial-intelligence-based concepts, i.e., generative adversarial networks and deep reinforcement learning, are integrated within the framework with a Rao–Blackwellized particle filter in an attempt to introduce the onboard intelligence necessary to perform navigation using earth's magnetic anomalies with an acceptable performance. This article includes a detailed formulation of the proposed framework, its components, and demonstration of selected key capabilities through numerical simulations. The results show the promising characteristics of the proposed navigation strategy and the potential to bridging the gap that separates geomagnetic-based navigation from practical applications and deployment.