Dronecode SDK WG

Dronecode SDK WG

Overview

The Dronecode SDK WG is tasked to develop a Drone Application Development SDK. This SDK should allow developers to create drone “applications” that can run on mobile devices (tablets, phones) and on-board companion computers, and that can also seamlessly integrate drones with cloud-based services. The SDK should ideally support Python, C++ and Android Developer APIs, including options like ROS, Cloud etc., and also integration with any simulation environment.

Currently we are evaluating existing developer APIs.

Some requirements include:

Robust, SDK solution for commercial deployments of PX4, and other compliant flight controllers
First Pass should include: Dronecode SDK with : basic onboard / basic mobile / basic cloud (RESTful) APIs implemented

Monthly WG Meetings Schedule

Bi-weekly meeting: Tuesday 8 AM PST.

Join the call: www.uberconference.com/lf-dronecode Dial in number: 415-891-1494 No PIN needed

Comparative Analysis of existing SDKs

Note: Essentially Mobile SDK is same as onBoard SDK with an additional wrapper layer and remote connection support

Mobile SDKs:

Type	Features	DJI	Parrot	FlytOS	DroneKit
Flight	High and Low Level Flight Control	Yes	Yes	Yes	Yes
	Aircraft state through telemetry and sensor data	Yes	Yes	Yes	Yes
	Obstacle Avoidance	Yes	No	Yes	No
Camera	Camera and gimbal control	Yes	Yes	Yes	Yes
	Live Video Feed	Yes	Yes	Yes	No
	Remote access to media stored on Camera	Yes	Yes	Yes	Yes
Missions	Define custom missions	Yes	Yes	Yes	Yes
	Predefined Missions: Waypoint, HotPoint, FollowMe	Yes	No	Yes	Yes
Subsystems	State information and control of Battery and Remote Control	Yes	Yes	Yes	Yes
	VR Support	No	Yes	No	No
	Cloud API Support	No	Yes	Yes	No

OnBoard SDK:

Type	Features	DJI	Parrot	FlytOS	DroneKit	AirLib
Flight	High and Low Level Flight Control	Yes	NA	Yes	Yes	Yes
	Aircraft state through telemetry and sensor data	Yes	NA	Yes	Yes	Yes
	Obstacle Avoidance	Yes	NA	Yes	No	Yes(minimal)
Camera	Camera and gimbal control	Yes	NA	Yes	Yes	No
	Live Video Feed	Yes	NA	Yes	No	No
Missions	Define custom missions	Yes	NA	No	Yes	No
	Predefined Missions: Waypoint, HotPoint, FollowMe	Yes	NA	Yes	No	No
Subsystems	State information and control of Battery and RemoteControl	Yes	NA	Yes	Yes	No
	Cloud API Support	No	No	No	No	No

SDK WG Bi-Weekly Meeting Minutes

SDK WG - June 20th

→ Comparative analysis of existing SDKs (listed above) was discussed in brief

→ Mark:

Licensing information to be added
API/Interfaces available as part parrot SDK could be considered as reference
Need to consider both Commercial and industrial applications/use cases

→ AirLib → SDK from Microsoft

MIT License
It has both CPP and Python Interfaces
Cross flight stack friendly and not only MAVLink tied (it has provided flexibility to add non mavlink protocols too)
Supports UDP, UART, TCP connection.
No support for Android and iOS
Supports only clang compiler
Mavlink library is ported to CPP. With this object oriented approach - makes easy to have pack/unpack function for each MAVlink message
Allows apps to configure drone connection parameters; so apps can choose which drone it wanna talk to.

→ DroneLink - > CPP Library provided by PX4 (Julian)

Only CPP support for now
BSD 3 License
Tightly coupled with PX4 and Mavlink like there are px4 custom modes support in source code.
Supports only UDP connection for now, TCP and Serial connection not implemented for now
REST Interface support - on going - to be shared
Wrapper for iOS and Android - To be implemented and shared (However, build infrastructure for Android and iOS is present)
Auto detects the tool chain on the current platform
Uses discovery model so that apps binds to a well-known port used by MAVROS (14540).(Author mentioned this as part of code that either this SDK or mavros can be used)

Quick Observations/Issues DroneLink:

Suggested Documentation changes related to App development - resolved by Julian.
Raised an issue regarding Example app build error - resolved by Julian.

Plan/Next Steps:

Offline discussion with other relevant stakeholders who could not attend the call (on Slack maybe) and capture the inputs/suggestions and update on WG Wiki
Julian to share his road map for DroneLink implementation
Progress/updates could be shared on wiki
To evaluate efforts on supporting Android and iOS on AirLib
To figure out on collecting inputs/suggestion by the application developers

Observations on Dronesmith SDK:

Dronesmith SDK is not opensource SDK
It is specific to Luci hardware, though they support MAVLink
Luci hardware is combination of Edison and STM32F427 Cortex which acts as Flight controller.
Luci runs a customized fork of PX4 Firmware to handle the flight control.
Dronesmith Engine is been provided as release binaries which runs on Edison board and host webserver.
Support for Cloud APIs and local REST APIs(Which onBoard apps can use) is been provided.

Observations from developers:

	AirLib	DroneLink	Dronekit
Setup/Infrastructure	Easy to setup. Just follow wiki instructions and the HelloDrone application will work.	Minor changes had to be done to setup (fixed now)	Easy setup
Ease of use	Easy to use, as user needs to include RPC client and rest will happen in background. PythonClient provides RPC facilities without app's knowledge. Its free for Python apps!	APIs are well categorized and easy to use	Easy to use
Extensibility	Easy to extend, could add mission planning feature easily	Uses compile-time plugs. Need to get more understanding
Feature Delta
Language bindings	Cpp/Python	Cpp/Java/Objective-C	Python/Java
Supported platforms	Windows/inux	Linux/Mac/Windows/iOS/Android	Linux/Mac OS/Windows
Code Structure	Excellent	Simple and Good	Good
Communication compatibility	Cross-protocol (as of now MavLink, ways to extend to other protocols)	MAVLink	MAVLink
AutoPilot Compatibility	All AutoPilots that support MAVLink	PX4	All AutoPilots that support MAVLink. Most compatible with ArduPilot, less to others.
Cloud Support	No	No	Partial (Store, Share and access vehicle and flight log info)
Features	More Navigation APIs	Comparatively less	Good
Mission management	No	Yes	Yes

Building AirLib SDK for Android

Built RpcLib, AirLib and MavlinkCom individually. There were linker errors (-lpthread) while building apps.
Now, using Android Ninja build system for the same - In progress.
Android has slightly different implementation of pthreads. So, we may have to make changes in AirLib sdk source code for Android under a flag.
Posted queries in GitHub::Airsim on SDK architecture which happens to be in their TODO list.
Shared HelloDrone example source for both AirLib and DroneLink sdks in Slack #sdk group. Examples are tested on jMAVsim.

Current Status:

DroneLink is chosen as Dronecode SDK and is renamed as DroneCore with the necessary architectural changes to suit the purpose.
Brainstorming the Intuitive, On-demand and compelling applications for the user - Ongoing and WIP

Drone application features (spread across above categories) available in iOS/Android:

Providing detailed wind and weather information to help user decide if it’s safe to fly.
Feed the app with parameters of your drone and indicate whether or not today is the day to fly. App should make sure whether the weather favours flying.
Hourly forecast on a range of piloting info is provided, including wind speed and direction, wind chill and temperature, cloud cover, visibility and more are available for a week.
Provides mapping with visible no-fly zones.
Provides news feed from various drone websites.
Lets user notify nearby airports that he is flying a drone.
Should alert user on any unmanned aircraft flying nearby.
Lets user plan the flight and do the survey autonomously.
Lets user capture images at a specific time say sunrise or sunset or with different shades of light.
User can view augmented reality projections of the sun and moon paths, pinpoint the time the sun or moon will be at a particular location in the sky, predict and plan for the golden hour, blue hour and every sunrise, sunset, moonrise and moonset and prepare night photography shots.
Users get to know the drone Pilots nearby and socialize during fly.
Should help user find who is flying in neighbourhood by letting them know of any user-submitted (other user) flying spots nearby, and the profiles of the people who fly there.
Allows user to post a their profile and share any inspiring photos of popular spots with other fliers.
Enable user to keep the frame fixed on a tracking object.
Should keep an eye on local air traffic, warn user if any manned aircraft are flying nearby.
Should allow user to blur object out of the point of interest area thereby promoting the object in focus.
Should offer intuitive and powerful waypoint mission engine with seamless mission syncing across all personal devices of the user.
Provide features like VR, panorama, Orbit, Follow Me modes.
Provide fancy modes to user like mimic, intercept, etc. In mimic mode, when you tilt your device in any direction, the aircraft will mimic the tilting motion of the device to the directed pitch and roll. In Intercept mode, when you Point your device in any direction as if it were a laser pointer, the aircraft will intersect the imaginary laser created by your device.
Provide you with real-time information about thousands of aircraft around the world to avoid close encounters. It lets user know before falling close to a manned/ unmanned aircraft.
Provides advanced flight planning and autonomous control for your drone. Select a path of coordinates and watch your drone fly the route, have the app control camera exposure on the go.
Capture enough data while the drone is flying, which is useful if user has a technical problem and want to know where the problem lies.
Controlling the direction that the camera is pointing and controlling when video is recorded.
Provide option to save pre-planned flights and reload them at any time, which is useful if user wants to take seasonal or construction videos.
Provide 2D/3D mapping feature that automatically creates a flight path for your drone; then uploads the video and flight data to remote servers. User can then use these servers or your desktop PC to crunch through this data, turning the images and location of the drone into a 3D model of whatever user were flying over.
Allow a second user to view the wireless video from the camera that's attached to the drone. The second user should also be able to control the camera, panning and tilting it around as the pilot controls the drone itself. In effect, that person should be able to work as the camera operator while the pilot keeps the drone flying.
Reproducing the feel and experience of flying a drone. Should provide multiple viewpoints including the ground — plus a normal camera on the drone and a stabilized gimbal.

Table of Contents