Time series classification of radio signal strength for qualitative estimate of UAV motion

Abstract

Many common navigation solutions fall short when an aircraft’s GPS signal is either jammed or spoofed. This is typically due to the iterative nature of the estimation process, which requires an acceptably accurate initial estimate, or due to the accumulated error of inertial sensors, which are unable to directly observe the position of an aircraft. A mechanism is presented in this paper which operates on qualitative information, allowing an aircraft to remain within a vicinity despite an absence of precision localization. A long-short-term-memory neural network was used for time series classification of radio signal strength data on a light weight fixed wing UAV. Simulation results show that the two class classifier is able to determine the motion of an aircraft with respect to a radio beacon with 97.73% accuracy. The classes used for classification represent motion as either towards, or away from a beacon. A simple high level controller was designed to use the classification output and converge on a beacon. Results from this paper indicate that this unique application of qualitative navigation by the application of time series classification offers a viable alternative to aircraft navigation in GPS denied environments.