Experimental System Design of an Active Fault-Tolerant Quadrotor

TL;DR

The paper tackles rotor-failure resilience in quadrotors by developing an active fault-tolerant framework that surrenders yaw control and relies on reduced-attitude metrics to maintain pointing while avoiding full attitude reorientation. It combines an L1-adaptation fault-detection mechanism with an autonomous transition to a fault-tolerant controller and couples this with a modular drag augmentation design to cap yaw rates within gyro limits. Experimental validation in a controlled arena demonstrates safe operation under single and dual rotor failures, defines a practical yaw-damping range ($k_z \in [0.05,0.35]$) and a transition latency of $\sim 100$–$150$ ms, and provides platform-agnostic guidelines for robust fault tolerance. The results offer a sensor-light, computation-light solution with broad applicability to diverse quadrotor geometries and nominal gyro constraints.

Abstract

Quadrotors have gained popularity over the last decade, aiding humans in complex tasks such as search and rescue, mapping and exploration. Despite their mechanical simplicity and versatility compared to other types of aerial vehicles, they remain vulnerable to rotor failures. In this paper, we propose an algorithmic and mechanical approach to addressing the quadrotor fault-tolerant problem in case of rotor failures. First, we present a fault-tolerant detection and control scheme that includes various attitude error metrics. The scheme transitions to a fault-tolerant control mode by surrendering the yaw control. Subsequently, to ensure compatibility with platform sensing constraints, we investigate the relationship between variations in robot rotational drag, achieved through a modular mechanical design appendage, resulting in yaw rates within sensor limits. This analysis offers a platform-agnostic framework for designing more reliable and robust quadrotors in the event of rotor failures. Extensive experimental results validate the proposed approach providing insights into successfully designing a cost-effective quadrotor capable of fault-tolerant control. The overall design enhances safety in scenarios of faulty rotors, without the need for additional sensors or computational resources.

Figures (9)

• Figure 1: Overview of our active fault-tolerant control scheme with state prediction, adaptive, and fault detection modules. The L1 augmentation calculates a damage estimate and is passed to the fault detection threshold which dictates the switch between a standard and fault-tolerant controller.
• Figure 2: Quadrotor model with inertial (blue) and body (green) frame definitions. Both frames are right-hand-frames.
• Figure 3: 3D printed landing gears with added surface area to introduce rotational drag to the system.
• Figure 4: Quadrotor in fault-tolerant mode with variables related to drag coefficient estimation. The added surface area on the drone are highlighted with white hash marks.
• Figure 5: Showcase of three different trajectories and their tracking performance with different system attributes such as single vs. dual rotor failure cases, slow vs. fast trajectories, and different error metrics.