Adaptive Terrain Memory Navigation: A Modular Fail-Safe for Unmanned Aerial Vehicles in Satellite-Navigation-Denied Environments

Abstract

Tactical unmanned aerial vehicles (UAVs) rely almost exclusively on GPS and a radio-control (RC) link for navigation and recovery. Either link can fail independently — through deliberate jamming or spoofing, terrain and weather blockage, or physical damage to the receiver hardware — and a UAV with no independent means of localization typically drifts, strikes terrain, and is destroyed in the resulting fall. This paper presents Adaptive Terrain Memory Navigation (ATMN), a fail-safe safety system rather than a primary navigation system. Its LiDAR and Simultaneous Localization and Mapping (SLAM) stack run continuously in the background throughout the sortie, building a terrain map and tracking the UAV's position, but this output does not participate in flight control while GPS and RC remain available. Navigation authority passes to ATMN the instant either GPS or RC signal is lost, without waiting for both to fail together, and reverts automatically once both are confirmed restored. On engagement, ATMN holds the aircraft within its self-built terrain map and retraces the last known safe corridor, preventing the uncontrolled descent that would otherwise destroy the airframe. It is built as a modular, platform-agnostic add-on connecting through the standard MAVLink payload interface, requiring no modification to the host airframe or flight-control software. We present the system architecture, an either-link trigger logic, a nine-item bill of materials totaling 462.6 g and 20.7 W at a cost of approximately ₹1.6 lakh per unit, detailed power and weight budget calculations against a publicly specified commercial hexacopter platform, and a comparative evaluation against alternative approaches.