This book presents the astonishing potential of deep sensorimotor policies for agile vision-based quadrotor flight. Quadrotors are among the most agile and dynamic machines ever created. However, developing fully autonomous quadrotors that can approach or even outperform the agility of birds or human drone pilots with only onboard sensing and computing is challenging and still unsolved.
Deep sensorimotor policies, generally trained in simulation, enable autonomous quadrotors to fly faster and more agile than what was possible before. While humans and birds still have the advantage over drones, the author shows the current research gaps and discusses possible future solutions.
Author(s): Antonio Loquercio
Series: Springer Tracts in Advanced Robotics, 153
Publisher: Springer
Year: 2023
Language: English
Pages: 68
City: Cham
Series Editor’s Foreword
Acknowledgements
About This Book
Contributions
Journal Publications
Peer-Reviewed Conference Papers
Open-Source Software
Awards
Contents
1 Introduction
1.1 Motivation
1.1.1 Advantages
1.1.2 Challenges
1.2 Related Work
1.2.1 Autonomous Drone Navigation in Unknown Environments
1.2.2 Autonomous Drone Racing
1.2.3 Uncertainty Estimation for Safe Deep Learning
1.2.4 Unsupervised and Weakly-Supervised Learning Robot Vision
2 Contributions
2.1 Transfer Learning for Agile Drone Navigation
2.1.1 Dronet: Learning to Fly by Driving
2.1.2 Deep Drone Racing: From Simulation to Reality with Domain Randomization
2.1.3 Deep Drone Acrobatics
2.1.4 Agile Autonomy: Learning High-Speed Flight in the Wild
2.1.5 Limitations of Transfer Learning via Abstraction
2.2 Uncertainty Estimation for Safe Deep Learning
2.2.1 A General Framework for Uncertainty Estimation in Deep Learning
2.2.2 Limitations of Proposed Framework for Uncertainty Estimation
2.3 Unsupervised and Weakly-Supervised Learning of Robot Vision
2.3.1 Learning Depth with Very Sparse Supervision
2.3.2 Unsupervised Moving Object Detection via Contextual Information Separation
2.3.3 Limitations of Unsupervised or Weakly-Supervised Learning
2.4 Additional Contributions
2.4.1 Drone Racing Demonstrator at Switzerland Innovation Park
2.4.2 Unrelated Contributions
3 Future Directions
Appendix References