Robust and Efficient State Estimation for Micro Aerial Vehicles

Abstract

Autonomous robots provide excellent tools for information gathering in a wide variety of domains, from environmental management to infrastructure inspection and search and rescue. Micro aerial vehicles, in particular, offer a high degree of mobility that can further their effectiveness in such environments. Deployment of aerial robots deployment in remote environments can render human operation unfeasible, necessitating resilient autonomous systems. At the core of any autonomous mobile robot is the capability to produce a consistent belief of the robot's location with respect to its environment. The efficacy and efficiency of the state estimator affects the performance of nearly all other robotic systems, from high-level motion planners to feedback controllers. This thesis examines the challenges associated with providing an accurate state estimate for MAVs operating in diverse environments.

Operation in cluttered, indoor environments precludes the use of GPS and requires laser- and vision-based odometry methods for state estimation. Additionally, the added mobility of aerial platforms comes at the expense of size, weight, and power constraints that preclude more information-dense and computationally expensive observation modalities. This thesis addresses these challenges through the development and evaluation of two modern state estimation methodologies representing both primary sensing modalities. The approach to the methodologies are driven by computational efficiency and extensibility. A novel multi-modal framework is then formulated by combining both observational models to produce a consistent and accurate state estimator that is robust to environmental diversity. All three state estimation methods are implemented on an experimental platform and evaluated through a series of flights. Quantitative analysis is provided through flights in a motion capture arena while qualitative evaluation is provided by traversals through challenging indoor environments. These evaluations demonstrate the ability to provide consistent and accurate state estimation in real-time on constrained aerial platforms operating in diverse environments.