Development of a distributed model predictive control framework for autonomous unmanned aerial vehicle swarms

Abstract

Unmanned Aerial Vehicle (UAV) technology has witnessed remarkable advancements, permeating various industries. Quadcopters, a specialized type of UAV with four rotors, are at the forefront of these innovations with recent trends and future projections indicating that the utilization of multiple quadcopters, forming what it is referred to as a swarm of quadcopters, is poised for substantial growth. However, the inherently nonlinear and highly complex behavior of quadcopters introduces serious challenges in terms of efficiently controlling them, necessitating the need of developing advanced control methods and intricate frameworks, particularly when addressing the collective behavior of a swarm. In response to these challenges, this diploma thesis introduces a comprehensive Distributed Model Predictive Control (DMPC) framework, designed to enable precise trajectory tracking for a swarm of quadcopters. Building upon the initial introduction of a Model Predictive Control (MPC) scheme for addressing the trajectory tracking task of a single quadcopter, which employs MPCs and Proportional-Integral-Derivative (PID) controllers in a cascaded design, the proposed framework seamlessly scales its capabilities. Through the integration of an inter-agent communication strategy, it achieves the transition to effective, autonomous and distributed control over quadcopter swarms, even in complex scenarios such as trajectory tracking with collision avoidance and obstacle evasion. The distributed nature of the proposed approach eliminates the necessity for a central controller, thus enhancing the system’s resilience in the event of individual failures, or communication disruptions, and offers easy adaptation to various swarm sizes. Additionally, the proposed framework diverges from conventional methods by liberating the agents of the swarm from following predetermined trajectories and predefined formation strategies. Instead, it empowers agents to dynamically tailor their behavior, enabling them to closely follow real-time desired trajectories with responsive agility to unpredicted scenarios. To evaluate the efficiency and practicality of the proposed framework, simulations have been conducted for trajectory tracking, including scenarios involving suddenly appearing obstacles. The results endorse its suitability for quadcopter swarm control and illuminate its potential application across a broad spectrum of domains, where the swarm’s precision, adaptability and robustness are paramount.