Development and Simulation of a Novel Guidance System for Quadrotors Flying in a Contested Environment

Abstract

Due to their accessibility, quadrotors are used as testbeds for guidance, navigation, and control systems and have been used for a wide variety of applications. These applications range from commercial aerial photography to military surveillance. In the latter case, the vehicle can be exposed to danger if opposing agents are present. The vehicle is incentivized to take cover among obstacles in order to protect it during surveillance operations. However, current research seeks only to avoid obstacles. This thesis develops a guidance system capable of generating tactical flight of a quadrotor. This guidance system consists of three complementary subsystems. The first is fast model predictive control (FMPC). This algorithm uses the dynamics of the system to plan several time steps into the future. Using this plan, the algorithm communicates the optimal action to take and repeats the process. FMPC is advantageous due to its fast computation subject to the system dynamics and customizable cost functions that can be used to enforce tactical behavior. However, it is not designed to work in non-convex environments and is vulnerable to locally optimal points that are not the goal position. The convexity issue is mitigated by using quadratic discrimination to find a locally convex region. Local optima are avoided by employing a global pathfinding algorithm built off of the motion primitive library. Flight test results demonstrating the capabilities of the resultant guidance system are presented.