3DR SOLO with REACH

Our agenda right now is to release Pixhawk integration for real-time navigation. Next step would be the mapping workflow. The ETA on this is about 4 weeks.

I know that there is confusion when trying to separate Pixhawk integration from precise georeferencing, so I feel like there is a need to provide further explanation that these two are very different things. This post is more for future reference, than a direct reply.

###Pixhawk integration

includes a custom binary protocol which is used to transfer calculated coordinates from Reach to Pixhawk and RTCM3 corrections all the way from Mission Planner, through the telemetry link and Pixhawk to Reach.

Used for:

  • Precision navigation
  • Precision landing
  • Repeatable day to day missions

This method assumes:

  • constant communication between base and rover Reach
  • APM based autopilot with latest firmware

Camera timestamp integration

The main issue with precision georeferencing is that there is a random shutter delay and discretization of coordinates. Let’s assume that we have perfect RTK solution in flight and are logging camera positions in Pixhawk as usual. After you have triggered the picture you do not know when the camera actually took it and you know a new coordinate at 5Hz rate. Assume that we are flying at 10m/s. 10/5 = 2 meters, between coordinate updates. Combined these two effect result in meters of errors even with RTK onboard. And if communication was lost in flight, that is another source of errors.

What Emlid proposes based on vast mapping experience?

There is only one way to know when shutter was actually opened, it is when camera attempts to fire the flash. This is why a hotshoe adapter (or internal modification for cameras with integrated flash) is required. You will need to connect Reach to the pin on the hotshoe and it will store timestamps of each picture. These timestamps are then stored in RINEX file along with raw GNSS observations. After the flight you will postprocess logs from base and rover in modified RTKLIB and it will give you a list of coordinates along with accuracy estimate for each of them. Postprocessing gives better precision and is more robust compared to RTK.

This method assumes:

  • No need for communication between rover and base, they log the data.
  • Not autopilot dependent, can work on Solo, DJI etc.
  • No autopilot software modification.
  • Requires flash sync cable to be connected to your camera of choice.

Some companies are sweet talking their customers into mapping with GoPro and referencing pictures with normal GPS. That simply shows lack of knowledge on their side. That might work for ok looking 3D models, but for precise ones that you can trust for measurements of volumes, areas and coordinates never. It surprises me how some are actually selling this as a business tool.

Can I have both?

You can both have precision navigation in flight and precisely referenced pictures. Just implement both integrations.

Solo integration

Most Solo users would not want to change their Pixhawk2 firmware to one that supports Reach, so we will focus on camera only integration.

We recognize that providing a tutorial for Solo modification for precision mapping without GCP would be a big hit. Emlid will give a free Reach to someone who is ready to make a demo (after the software is released). Requirements: your Solo with a camera with hotshoe installed, pulse checked with scope and ready to be connected to Reach. You need to have some publication related to SOLO or drones on the web and basic electronics skills and drone building skills (Reach needs to be integrated in the Solo bay, connected to 5V power there, ground plane and antenna placed on top of the shell). In exchange you will also get private integration support. The result of this demo would be a blog post here and on DIYD. If someone is interested please PM me for details. (@Global_Saffer it could be you, of you are ready for some hacking :smile: )

7 Likes