Enhanced Performance, Stability, and Practicability of Attitude and Position Estimators for Robotic Vehicles

点击翻译按钮获取中文摘要

The research objective of this proposal is to further the state-of-the-art in attitude and pose estimation for robotic vehicles. This work is motivated by the proliferation of aerial robotic vehicles that currently or are envisioned to autonomously inspect infrastructure, monitor construction and mining operations, deliver goods in urban areas and medical aid in remote regions, and monitor agriculture and wildlife, often in close proximity to humans. These tasks are important to Canada's infrastructure maintenance and replacement, economic growth, wildlife conservation, and support of Northern regions. Typical aerial robotic vehicles have limited computational resources, and their on-board sensors provide imperfect data. For reliable, effective, and safe use of aerial robotic systems, either individually or in teams, the attitude or attitude and position (i.e., pose) of the vehicle must be estimated by an algorithm that is computationally simple and immune to bias and noise corrupting sensor data. Direction cosine matrix (DCM) estimators that estimate the DCM describing a vehicle's attitude directly have gained popularity because they are computationally simple and are provably asymptotically stable, unlike Kalman-like filters such as the extended and unscented Kalman filters. Moreover, by estimating the DCM directly, which is a global and unique representation of attitude, deficiencies of DCM parameterizations such as singularities are avoided. However, state-of-the-art DCM estimators, as well as similar pose estimators that estimate both attitude and position, do not actively filter bias and noise that corrupts interoceptive and exteroceptive measurement data, such as rate gyros and magnetometers, respectively. As a result, attitude and pose estimates are poor which, in turn, negatively impacts the precise and accurate operation of robotic vehicles. The overarching goal of the proposed research, and anticipated outcome, is realizing exceptional attitude and position estimates of robotic vehicles rotating and translating in three-space by negating the detrimental impact of measurement bias and noise. This will be achieved by integrating a disturbance estimator to estimate bias and noise corrupting interoceptive measurements, using a specialized linear time-invariant system to filter exteroceptive measurements, and using a different estimation error term to improve estimator convergence, all while guaranteeing asymptotic stability of DCM and pose estimators. Four PhD students, three MEng students, and five undergraduate students will be intimately involved in the proposed research.