Development of a Novel Pursuit-Evasion Path-Planning Algorithm and Generatively Designed Propeller Guards to Avoid Aerial Capture

Abstract

Quadrotor unmanned aerial vehicle (UAV) technology continues to advance at incredible rates, to the point where governments have begun to consider how to combat “rogue” drones - those that might be driven by malicious operators. To combat particularly dangerous situations, the US government has invested in DroneHunters, which are drones that track, pursue, and capture “rogue” drones using a launched net. This project focuses on building a custom drone that is meant to evade a DroneHunter while still performing reconnaissance (or some other task) in an area protected by DroneHunters. We hope to expose limitations in the current DroneHunter technology to improve the Dronehunter's efficacy and improve public safety. This project breaks into two main design components. First, custom propeller guards were designed using a generative design program, they were 3D printed, and then installed on the drone so that the drone can withstand being caught in a net. Without the net getting caught in the propellers, the drone can still fly and escape the DroneHunter. Second, the RRTABC algorithm, a novel pursuit-evasion path planning algorithm based off of rapidly-exploring random trees, is introduced. This algorithm allows the drone to evade the DroneHunter without having to completely abort its path or change its destination. The propeller guards are shown to successfully protect against a net; however, further design iteration is necessary to allow for stable and robust flight while they are installed. The RRTABC algorithm is validated first through simulation and later through integration on a drone that is able to theoretically evade a pursuer drone. Further directions for both propeller guard design and algorithm improvement are also discussed.