Advancing Drone Autonomy: A Comprehensive Framework for Agile Controller Design in Quadcopters

The complexity of controller design for achieving high agility and speed in maneuvers arises from the non-convex nature of the objective function, reliance on noisy parameter evaluations during tuning, and the need for dynamically changing parameters across different trajectory segments. Building upon these challenges, this paper presents a comprehensive framework for the controller design process, addressing the complexities inherent in achieving high agility and speed in maneuvering quadcopters. The proposed design philosophy is built upon a range of research conducted over the last 15 years, consolidating insights and advancements from diverse studies in the field. It begins by discussing early demonstrations of vision-based autonomous flight and progresses to recent investigations into human vs. algorithmic piloting abilities. Key challenges such as perception-to-action dependence, domain shifts, and sensor limitations are highlighted, along with proposed solutions including hybrid simulation-real world approaches. In addition to the comprehensive control system design philosophy, it also evaluates potential drone platforms for research and development purposes. Through this thorough examination, the paper provides a foundation for further studies in agile controller design for quadcopters, highlighting the importance of robust methods and innovative approaches in advancing drone autonomy. This framework provides a basis for further exploration and development in the field, with potential applications across various controller design methodologies.