Autonomous Target Search Using Unmanned Aerial Vehicles in the Presence of Signal Noise

A search vehicle routing algorithm is presented, inspired by thermalling in glider operations, for use in unmanned aerial vehicle (UAV) search routing problems where radio frequency signal noise affects target geolocation. We conduct ground testing experiments and observe that signal strength significantly deviates from its mean value — 10-20% variation in our experiments — at each distance epoch away from a target of interest. The thermalling algorithm uses straight-line segments and relatively simple guidance logic to generate routes that appear promising for fixed-wing, short endurance and autonomous UAV routing applications. We compare the simulated performance of the thermalling algorithm to a stochastic gradient descent algorithm using 1,000 randomly generated problem instances and show that, when considering signal readings and distance traveled as performance metrics, the thermalling algorithm performs better in 100% and 37% of problem instances respectively.