Drones Demystified!

K. Alexis, C. Papachristos, Autonomous Robots Lab, University of Nevada, Reno

A. Tzes, Autonomous Robots & Intelligent Systems Lab, NYU Abu Dhabi
Drones Demystified!
Topic: Introduction
Course Goals

 Be approachable and accessible for all.

 Develop a broad understanding on how aerial
robots fly and operate.
 Develop the capacity to design navigation and
other automation and autonomy robot
functionalities.
 Combine theory with intuition and practice.
 Be able to implement new ideas in practice.
 Provide the knowledge background and the
tools for your further steps in aerial (and not only)
robotics.
Course Approach

 Brief Overview and Focus Sections: Each topic is

overviewed and selected subtopics are
thoroughly described.
 Any Entry Point: Feel free to start the course from
any subsection of interest. Try to go through all of
them to understand the synergies.
 Coding Examples: Each focus section is
accompanied with coding examples using
Python, MATLAB and C++
Course Contents

 Introduction: Get a broad understanding about aerial robotics, what they

are, what they look like, what they can do and how they do so.
 Modeling: Understand the physical principles behind the flight dynamics of
aerial robotics and develop the capacity to derive their mathematical
model.
 State Estimation & Perception: Learn how on-board estimation of the
vehicle full pose (position and orientation) takes place. Understand
onboard vision and perception in general.
 Flight Controls: Learn how to design high-performance flight controllers.
 Motion Planning: Learn how to develop algorithms for autonomous motion
planning for aerial robotics.
Course Material

 Online Textbook: a brief textbook covering the topics of the course. To be

continuously updated every year.
 Lecture Slides: Used for the classroom presentations and also as a way for
notes keeping and direct reference to the course contents.
 Coding Examples: small programs written in Python, MATLAB, and C++
covering the focus sections of the course.
 Open-Source Aerial Robots Simulator: a complete simulation environment
for advanced designs.
 Historical Background of Aerial Robots

 Unmanned Aerial Vehicles lie on the

foundations set in aircraft design &
automatic flight control.
 First UAVs were based on modifications of
manned rocket and aircraft designs.
 Their evolution process has constantly
challenged the limits of technology
towards miniaturization, flight control
agility, perceptual skills, navigational
autonomy, flight endurance or even
unconventional skills such as aerial
manipulation.
 Current Trends in Aerial Robotics

System Miniaturization

Prolonged Endurance
AscTec Firefly with VI-Sensor,
UNR Agile Flight Control
VI-Sensor, ETH Zurich
Perception & 3D Mapping

Autonomous Motion Planning

Multi-Robot Systems

Augmented Human-Robot
AtlantikSolar UAV, ETH Zurich Oculus Rift, Oculus
 Current Trends in Aerial Robotics

Retaining Stability during Propeller Failure Convertible Unmanned Aerial Vehicle

 Current Trends in Aerial Robotics

Endurance World Record (81.5 Solar-powered Flight) Aerial Mapping

 Current Trends in Aerial Robotics

Information Gain-based Efficient Exploration and Localizability-aware Autonomous Exploration and

3D mapping of Unknown Environments Mapping
 Current Trends in Aerial Robotics

Multi-Modal Localization and Mapping GPS-denied navigation

 Current Trends in Aerial Robotics

Physical Interaction Control Inspection through Contact

 Impact in Society
Google trends “Drones”  Unmanned Aerial Vehicles
(UAV) Market worth 14.9
Billion USD by 2020

 Indicative Applications:
 Infrastructure Inspection
and Maintenance
 Humanitarian Assistance
 Precision Agriculture
 Climate Control
 Security
 …and much more!
 What are the challenges ahead?
No sensible decision can be made any longer
without taking into account not only
the world as it is, but the world as it will be. I.A.
 a robot may not injure a human being or, through
inaction, allow a human being to come to harm
 a robot must obey orders given it by human beings
except where such orders would conflict with the
first law
 Can we operate aerial robots without  a robot must protect its own existence as long as
such protection does not conflict with the first or
having special skills? second law

 Can aerial robots be something more than

a flying camera?
 Can we assign complex tasks to
autonomous robots?
 Can we ensure collision avoidance?
 Can we trust an aerial robot flying in the
urban landscape? In the controlled
national airspace?
Isaac Asimov
The Aerial Robot Loop

Block diagram of the main loops running at every aerial robot

The Aerial Robot Loop
 Real-life Aerial Robot
expressing its dynamic
behavior in response to
the control inputs and
external disturbances
The Aerial Robot Loop
 Real-life Aerial Robot
expressing its dynamic
behavior in response to
the control inputs and
external disturbances

 Inertial Navigation
System
 Localization & Mapping
 Sensor Fusion
The Aerial Robot Loop
 Real-life Aerial Robot
expressing its dynamic
behavior in response to
the control inputs and
external disturbances

 Flight control system  Inertial Navigation

responsible for ensuring System
flight stability and
reference trajectory  Localization & Mapping
tracking as well as
disturbance rejection.  Sensor Fusion
The Aerial Robot Loop
 Path planning in order to  Real-life Aerial Robot
compute the path the expressing its dynamic
robot should follow to
ensure safe navigation and behavior in response to
execute the desired the control inputs and
mission. external disturbances

 Flight control system  Inertial Navigation

responsible for ensuring System
flight stability and
reference trajectory  Localization & Mapping
tracking as well as
disturbance rejection.  Sensor Fusion
 The Aerial Robot Loop

A hard real-time system with relatively limited computational resources!

The Robot Platform

 Different aerial vehicle configurations satisfy

different needs in terms of precision flight,
endurance, operational requirements,
ETH Zurich
robustness, payload capacity and more.
 Rotorcraft Systems
 Fixed-Wing Systems ETH Zurich

 Hybrid Systems
 Biomimetic Systems
WingTra, ETH Zurich
 Lighter-than-Air Systems

ETH Zurich UT
The Robot Platform

 Derivation of the Vehicle Dynamics Model: A

complex mathematic process that heavily
depends on the good understanding of the
following topics:
 Frame Rotations, Representations and Coordinate
Systems
 Rigid Body Dynamics Modeling
 Derivation of Aerodynamic Forces
 Modeling is a critical step that enables state
estimation, flight control and prediction of the
flight behavior of the vehicle.
 As part of this course we will derive a first-level
model of a multirotor Micro Aerial Vehicle and a
Fixed-Wing UAV
Perception & State Estimation

 The process of using data coming from different

sensors on-board in order to estimate (in real-time)
the state of the vehicle (position, orientation and
possibly aerodynamic and other high-order states).
Typical sensors employed are:
 Accelerometers
 Gyroscopes
 Magnetometer
 GPS
 Camera systems
 LiDAR
 Structured-light sensors
 …and more
 It corresponds to fundamental loop for the robot
operation and critical to advance its autonomy.
Perception & State Estimation

 The sensor data fused to estimate the full state in

a bias-free and accurate way in high-update
rates. This process depends on methods of the
state estimation theory. Within this course, we will
investigate:
 Inertial sensing
 The concepts of Kalman Filtering
 The design of an Inertial Navigation System
Flight Control System

 The Flight Control System is responsible to ensure

the closed-loop stability of the vehicle and
guarantee high performance trajectory tracking
despite possible uncertainty in the modeled
dynamics and the effects of external
disturbances.
 Flight Control System design relies on concepts of
control theory and a precise enough but simple
model of the vehicle dynamics that allows
model-based control.
 Flight control theory is a huge field and the
community has proposed numerous alternative
approaches towards providing aerial robots with
the desired flying qualities.
Flight Control System

 Within the framework of this course we will

examine some basic and a few advanced
methods on flight control synthesis. More
specifically, we will cover the following topics:
 Proportional Integral Derivative control
 Linear Quadratic Regulator control
 Linear Model Predictive Control
 While, we will then investigate an autopilot
design example and apply the relevant control
strategies.
Path Planning

 Path planning is the process of deriving the path

that the robot should follow in order to ensure
safe navigation and successfully execute its
mission. The problems addressed through the
relevant methods include the following:
 Collision-free Navigation
 Inspection Path Planning
 Target Follow
 Coordinated Motion
 Collaborative Aerial Manipulation
 It corresponds to a critical topic to advance the
aerial robotics autonomy.
Path Planning

 Path planning is a rich field and several different

approaches have been proposed such as
sampling-based methods or algorithms relying on
convex optimization strategies. Within the
framework of this course we will examine
Sampling-based algorithms and their
applications in the problems of:
 Collision-free Navigation
 Structural Inspection Path Planning
 Autonomous Exploration
The Aerial Robot Loop

These Loops Run Simultaneously!

Thank you!
Please ask your question!