Guidance, Navigation, Control and Mission Logic for Quadrotor Full-cycle Autonomy


Loading...

Author / Producer

Date

2018-03-08

Publication Type

Master Thesis

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

Autonomous multirotor Micro Aerial Vehicles (MAVs) are currently being actively developed for personal, commercial and military applications. Three of the most rapidly growing use cases are precision farming, environmental surveying and surveillance. These markets require full-cycle autonomous MAVs i.e. robots that can not only perform single-flight autonomous data collection but that can also automati- cally recharge so as to sustain long term operation with no human in the loop. No such suitable commercial MAV exists today. Motivated by this market gap, this thesis pioneers a full-cycle autonomous MAV system using an AscTec Pelican quadrotor. Designed for a data collection mission, the MAV takes off, flies a waypoint trajectory, lands on a contact-based charging pad, recharges and then repeats the cycle all without human presence. This thesis develops the entire guidance, visual landing navigation, control and autonomy engine subsystems from the ground up. After describing the hardware and software architectures, a visual landing navigation pipeline is developed based on Recursive Least Squares AprilTag bundle pose estimation. Then, guidance is explained which consists of polynomial trajectory generators, a real time trajectory sequencer and a trajectory tracker. A full cascaded control loop is designed next and this report explains its complete development starting from a flight dynamics model, then system identification, then controller theoretical design followed by implementation peculiarities, parameter tuning and finally an enlightening performance and robustness analysis. With the guidance, navigation and control system available, a state machine based autonomy engine is developed to carry out a full-cycle autonomous data collection mission. We finally present how these subsystems perform individually and as a complete, working system. Data is taken from a Software In The Loop high fidelity simulation as well as indoor and outdoor real life flight tests. The highlights of this body of work are: • An AprilTag 2 based visual landing navigation pipeline, an evaluation of bundle pose measurement accuracy and a definition of ideal bundle geometry; • A guidance trajectory sequencer that implements a well-defined set of real time trajectory sequencing mechanics; • A complete three stage cascaded control loop with novel quaternion-based yaw control; • A hierarchical autonomy engine based on self developed, Unified Modeling Language compatible state machines; • A modified RotorS simulator that can run a Software In The Loop simulation of the entire system presented herein.

Publication status

published

External links

Editor

Contributors

Examiner : Brockers, Roland
Examiner : Stastny, Thomas
Examiner : Hinzmann, Timo

Book title

Journal / series

Volume

Pages / Article No.

Publisher

ETH Zurich

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Guidance navigation and control; Autonomy; Takeoff; Vision-based landing; State machines; System identification; Quadrotor dynamics; Quadrotor control

Organisational unit

02620 - Inst. f. Robotik u. Intelligente Systeme / Inst. Robotics and Intelligent Systems

Notes

Funding

Related publications and datasets