November 25, 2020
- Status update by component/project
- Roadmap, and Release discussion
- Community Q&A
- In-Depth discussions
MTD/Device tree: It was evaluated that it would help to have some sort of a device tree in PX4
The structure of buses and devices a board has will get into a statically documented tree and it’s possible to register devices dynamically e.g. for external sensors. We put focus on reusing existing tooling and make it scalable. The current state of the pull request is just trying possibilities.
Security special interest group
There was the first instance of the Security special interest group one hour ago. It’s in the dronecode calendar https://www.dronecode.org/calendar/.
Goal: Render a specification for authentication of artifacts, protection against several attacks.
Summary: Initial discussion about how to get security certified.
[Link to meeting minutes :https://docs.google.com/document/d/1twgmiGZpYADzYOVYwAiPS-3an68WwUdV1M_0Spo7oHs/edit]
If you’re interested or have a Security background please join the next call.
OS / NuttX
The release candidate of NuttX was not cut because too many issues were found. We don’t update PX4 yet.
Lidar lite issue is still present on 1.11.1 and the fix got ported to the release branch. We’ll cut another point release soon with this and the RC fixes.
Settings like the sensor orientation should be runtime user-accessible. There’s a summarized issue for the general problem. Instances of sensors need to be clear. Board sensors should be identifiable from the device tree and external sensor with additional user-facing information like the plug on the board. The reference of device IDs needs to be shared with the ground station UI such that it can bring the devices into context.
User intent on the mode switch was misinterpreted in certain cases reported last week as the unexpected mode switch out of offboard.
Multi-EKF fix for excessive primary EKF switches.
Existing optical flow pr will likely be tested soon.
There is the possibility that delayed samples are used where they should not
Possibility to initialize the EKF without baro or mag on the way:
Magnetometer bias reset improvements ready for review. It gets things into place for magnetometer on the fly magnetometer switches.
Issue detected where acceleration stick mapping
- Rover CI SITL tests sometimes segfault. @Jaeyoung-Lim has a quick look.
Most use cases for PX4 rover are R&D. It’s experimental.
- UX with Missions and GPS
- Mandatory mission takeoff setpoint leads to rover getting stuck
- The altitude has to very accurate for waypoints to get accepted
In offboard mode it’s easier to use.
NXP uses rovers indoors.
@bperseghetti is working on an open design for rover hardware if anyone is interested. We’ll post links when it’s ready to be published.
Gazebo simulation overhaul. Please give feedback or reshuffle the content anyone who’s interested.
MAVROS / DDS / ROS2
Call yesterday of the Payload workgroup:
- @Cory_S from Freefly opened and presented the payload physical connection interface that will be available on the Freefly Astro. It was very well received and there was discussion about some pin details.
If you are interested join the next call.
We’ll cut a new 1.11 point release with the recent bugfixes soon.
The schedule for 1.12 is not in place yet. The goal is still to have a short well defined period for this next release. A quarterly release could help the commercial side people to plan updates.
Abraham: In QGC only linear straight line missions can be planned and that’s very common in the industry. Is there a limitation to not have curve or circular shaped plans?
- There is logic to have nice ~2m radius turns in the corners. @bresch is working on an even nicer corner cutting based on the acceptance radius. It’s coming soon.
- The avoidance interface allows to set bezier anchor points and can be used for smooth trajectory planning. Docs staring point: https://docs.px4.io/master/en/computer_vision/obstacle_avoidance.html
- RC switch, stick processing
- Hotfix for reevaluation https://github.com/PX4/PX4-Autopilot/pull/16264
- Debounce switches https://github.com/PX4/PX4-Autopilot/pull/16270
- RC frame processing, Filter for sticks Discussion
- RC protocol detection is dangerous, noise and unexpected packets can trigger decoders that have no CRC. We should log protocol switches.
Errata and Feedback
Let me know below if we failed to capture anything the right way, and if there are any updates (not present here), or you have feedback you would like to share with the dev-team.