Development and Hardware

Take a deep dive into the hardware and software that makes the commercial drone revolution possible. Whether you operate a mutlirotor or fixed-wing drone, these 60-minute sessions and panels will cover hardware components such as sensors and payloads as well as mastering VTOL (vertical take-off and landing), pre-flight planning and autopilots.


Wednesday, September 4

Getting Started with the Dronecode SDK

Class

1:00 PM – 2:00 PM

Track: Development and Hardware

Speakers: Julian Oes, Jonas Vautherin

The Dronecode SDK is an open source SDK allowing to control PX4, or other MAVLink powered drones, remotely or from a companion computer running onboard the drone. It is available in C++, Swift, Java, Python, Javascript and runs on Windows, macOS, Linux, Android, and iOS. It has been successfully deployed in production, e.g. in 3DR SiteScan, for controlling the Yuneec H520.

For many of today’s use cases of drones, it has become fundamental that flights can be automated. The SDK enables drone flights for different applications and aims to support drone operators to varying automation levels. Surveyors, first responders, or even deliveries have different requirements to sensors and user interfaces.

The SDK has a permissive open source license and is designed from the ground up to be easy to use from different platforms and in different languages. Moreover, it is flexible enough to allow drone/sensor manufacturers to extend it to their needs.

It is an important part of the PX4 ecosystem, allowing drone manufacturers to share the costs of developing and maintaining an SDK; and at the same time it allows companies to write a custom app once which is then compatible with all PX4-powered drones, which means over a million drones. Being the successor of DroneKit, it is the only alternative to the DJI SDK (with the difference that other drone manufacturers cannot benefit from the DJI SDK).

This class is targeting anyone who would want to write software controlling PX4 drones, as well as any drone manufacturer who may want to provide an SDK to its developer community. This presentation is based upon a presentation which was held at Interdrone 2018 (https://interdrone.com/conference/sessions/#developwiththedronecodeplatform) but heavily improved and updated to the latest state of the SDK..

Students will be introduced to the SDK and its philosophy, review its features (actions, telemetry, missions, camera, gimbal, offboard) and the lectureres will explain how to use it in different programming languages and on different platforms. Finally, demos will be run on simulated environments (software in the loop, hardware in the loop) using the material presented during the session.


Managing Drone Data: Choosing the Right Data Service Provider

Class

2:15 PM – 3:15 PM

Track: All

Speaker: Colin Snow

Choosing a drone data service means choosing a trusted business partner. With so many choices, it can be hard to know which one you should use. What advantage does one have over another? Unfortunately, the answers are not simple. A lot depends on things like your business model, your target market, what functions you need, and quite frankly how much you are willing to spend. The good news is the current wave of development and innovation in online drone data services is focused on mapping and analytic solutions that drone business service providers and enterprises can use to help solve real world problems – problems like infrastructure decay, crop yields, stockpile inaccuracies, improper construction sequencing, mining site logistics, etc. With that in mind, in this session I’ll outline the important criteria to help you evaluate the various web-based mapping and imaging services for drones.

In this session, you’ll learn:

• Who are the major providers and what’s their focus?

• How to determine the value of the services they provide?

• What questions should you be asking when comparing offerings?

• What’s the best way to integrating a service into your existing workflow so that it simplifies your operations and save costs?


Drone Hardware Reference Design

Class

3:30 PM – 4:30 PM

Track: Development and Hardware

Speaker: Michael Schäuble

Every established industry has a minimum level of standardization – consumer devices have USB, car tires fit more than one brand. The drone industry is maturing rapidly and has a need for standardization. Pixhawk as open source hardware is ubiquitous in the commercial drone industry and this talk will begin with shedding some light what this semi-standard has achieved for the industry already.

Needs for standardization in flight control are very similar across use cases, and main building blocks are shared amongst diverse hardware designs. The variety of use cases and the attempt to use COTS components with the goal of reducing cost can easily conflict when platforms get smaller or hardware requirements get more demanding, e.g. for certification and mass production. As a result, many manufacturers create custom-made PCB designs with a long list of custom components, pinouts, interfaces, and placements which lead to cost and time intensive evaluation and manufacturing processes that require significant effort to keep up with technological advancements. Knowledge and expertise are rarely shared between vendors and new platforms are engineered from scratch.

In his talk, I will explain why using hardware reference designs is the key to scale and rapid development for a drone manufacturer.

Hardware reference designs guarantee compatibility with upstream open source software and other peripherals. Moreover, sharing hardware design principles ensures compliance with requirements on placement and integration and improves the quality, reliability, and robustness of these components. Hardware reference designs are an excellent choice to create industry-wide standards when faced with a wide variety of products and use cases, reduce the room for error and mistakes when building complex products like drones, and provide a solid foundation when customization and flexibility prevent a one-size-fits-all solution.


Thursday, September 5

Improving Reliability, Verifying Integrity and Certifying Unmanned Aerial Vehicles Using Advanced Cloud and Deep Learning-Based Analytics of Flight Logs

Class

3:00 PM – 4:00 PM

Track: Development and Hardware

Speaker: Johannes Brand

Trials for air traffic management of unmanned aerial systems (UAS) are currently ongoing in the United States of America as well as in Europe [1, 2]. While air traffic management will enable UASs to operate more safely beyond visual line of sight (BVLOS) by implementing important regulatory control, it will also raise the requirements for the integrity of the underlying UAS control software which needs to reliably execute BVLOS operations under external supervision. Manual flight testing and analysis are common approaches to ensure the reliability and integrity of an UAS. However, manual approaches are slow and costly and thus limit the amount of testing and analysis that can be performed. Approaches for automatic flight data analysis on the other hand have the potential to increase reliability of the UAS control software by accelerating development cycles through extended testing and constant monitoring of the UAS in operation. The extended test data collection and analysis also directly facilitates the certification process of an UAV by providing comprehensive automatically generated reports of safety statistics for regulators.

In this session, I will present how cloud-based tools for automatically uploading, monitoring, and analyzing UAV flight data can be employed to increase the reliability of UASs and to verify the integrity of UAS software. I will address the challenges of developing such tools for large varieties of UAV types and models as well as UAS objectives. I will present how deep learning based automatic flight data analysis has improved the reliability of the open source autopilot software PX4 and elaborate on how it has enabled certification processes with the FAA and equivalent European regulators. Finally, I will show how valuable information for regulators can be extracted from the big amount of cloud data uploaded by users and present analysis results for PX4 from this process.

Prerequisite Skills or Knowledge:

This session intends to benefit a wide audience from the UAV industry and targets manufacturers and operators who are interested in improving the reliability of their UAV and certifying their UAV for BVLOS operations. The session requires some basic understanding of the technical problems faced by UAVs and of the certification process. However, it does not require a profound technical experience and can therefore benefit product managers and business leaders as well as engineers.


Friday, September 6

Adding AI and AR to Your Drone Mobile App

Class

9:00 AM – 10:00 AM

Track: Development and Hardware

Speaker: Godfrey Nolan

This talk is a hands on guide to adding AI (TensorFlow) and AR (ARCore) to your drone’s mobile app. After a brief overview of machine learning and how to use TensorFlow in Android apps we’ll show how to process the drone video so we can use the AI to identify and count objects. We’ll also look at how to use ARCore to provide context sensitive information such as roads and third party information to help guide your mobile app users to make better decisions.


Drone Hacking for the Newcoming Researcher: OS Attack Vectors

Class

10:30 AM – 11:30 AM

Track: Development and Hardware

Speaker: Ron Munitz

In this software oriented technical talk, Ron will survey common drone Hardware/Firmware/Software stacks, compare them to other embedded, IoT or mobile industries, and highlight interesting places to research, should one want to identify vulnerabilities that may result in a malfunction, or a drone takeover.

Ron will discuss possible mitigation, and give expert advice in applying Secure Development Lifecycle and hardening techniques when possible, or wisely designing the next product when things are inherently beyond repair.