The ArduPilot development team is proud to announce the release of version 3.4.0 of APM:Plane. This is a major release with a lot of changes so please read the notes carefully!
First release with EKF by default
This is the also the first release that enables the EKF (Extended Kalman Filter) for attitude and position estimation by default. This has been in development for a long time, and significantly improves flight performance. You can still disable the EKF if you want to using the AHRS_EKF_USE parameter, but it is strongly recommended that you use the EKF. Note that if an issue is discovered with the EKF in flight it will automatically be disabled and the older DCM system will be used instead. That should be very rare.
In order to use the EKF we need to be a bit more careful about the setup of the aircraft. That is why in the last release we enabled arming and pre-arm checks by default. Please don't disable the arming checks, they are there for very good reasons.
Last release with APM1/APM2 support
This will be the last major release that supports the old APM1/APM2 AVR based boards. We have finally run out of flash space and memory. In the last few releases we spent quite a bit of time trying to squeeze more and more into the small flash space of the APM1/APM2, but it had to end someday if ArduPilot is to continue to develop. I am open to the idea of someone else volunteering to keep doing development of APM1/APM2 so if you have the skills and inclination do please get in touch. Otherwise I will only do small point release changes for major bugs.
Even to get this release onto the APM1/APM2 we had to make sacrifices in terms of functionality. The APM1/APM2 release is missing quite a few features that are on the Pixhawk and other boards. For example:
- no rangefinder support for landing
- no terrain following
- no EKF support
- no camera control
- no CLI support
- no advanced failsafe support
- no HIL support (sorry!)
- support for far fewer GPS types
that is just the most obvious major features that are missing on APM1/APM2. There are also numerous other smaller things where we need to take shortcuts on the APM1/APM2. Some of these features were
available on older APM1/APM2 releases but needed to be removed to allow us to squeeze the new release onto the board. So if you are happy with a previous release on your APM2 and want a feature that is in that older release and not in this one then perhaps you shouldn't upgrade.
PID Tuning
While most people are happy with autotune to tune the PIDs for their planes, it is nice also to be able to do fine tuning by hand. This release includes new dataflash and mavlink messages to help with that
tuning. You can now see the individual contributions of the P, I and D components of each PID in the logs, allowing you to get a much better picture of the performance.
A simple application of this new tuning is you can easily see if your trim is off. If the Pitch I term is constantly contributing a signifcant positive factor then you know that ArduPilot is having to
constantly apply up elevator, which means your plane is nose heavy. The same goes for roll, and can also be used to help tune your ground steering.
Vibration Logging
This release includes a lot more options for diagnosing vibration issues. You will notice new VIBRATION messages in MAVLink and VIBE messages in the dataflash logs. Those give you a good idea of your
(unfiltered) vibration levels. For really detailed analysis you can setup your LOG_BITMASK to include raw logging, which gives you every accel and gyro sample on your Pixhawk. You can then do a FFT on the
result and plot the distribution of vibration level with frequency. That is great for finding the cause of vibration issues. Note that you need a very fast microSD card for that to work!
Rudder Disarm
This is the first release that allows you to disarm using the rudder if you want to. It isn't enabled by default (due to the slight risk of accidentially disarming while doing aerobatics). You can enable it
with the ARMING_RUDDER parameter by setting it to 2. It will only allow you to disarm if the autopilot thinks you are not flying at the time (thanks to the "is_flying" heuristics from Tom Pittenger).
More Sensors
This release includes support for a bunch more sensors. It now supports 3 different interfaces for the LightWare range of Lidars (serial, I2C and analog), and also supports the very nice Septentrio RTK
dual-frequency GPS (the first dual-frequency GPS we have support for). It also supports the new "blue label" Lidar from Pulsed Light (both on I2C and PWM).
For the uBlox GPS, we now have a lot more configurability of the driver, with the ability to set the GNSS mode for different constellations. Also in the uBlox driver we support logging of the raw carrier phase and pseudo range data, which allows for post-flight RTK analysis with raw-capable receivers for really accurate photo missions.
Better Linux support
This release includes a lot of improvements to the Linux based autopilot boards, including the NavIO+, the PXF and ERLE boards and the BBBMini and the new RasPilot board. If you like the idea of flying
with Linux then please try it out!
On-board compass calibrator
We also have a new on-board compass calibrator, which also adds calibration for soft iron effects, allowing for much more accurate compass calibration. Support for starting the compass calibration in the
various ground stations is still under development, but it looks like this will be a big improvement to compass calibration.
Lots of other changes!
The above list is just a taste of the changes that have gone into this release. Thousands of small changes have gone into this release with dozens of people contributing. Many thanks to everyone who helped!
Other key changes include:
- fixed return point on geofence breach
- enable messages for MAVLink gimbal support
- use 64 bit timestamps in dataflash logs
- added realtime PID tuning messages and PID logging
- fixed a failure case for the px4 failsafe mixer
- added DSM binding support on Pixhawk
- added ALTITUDE_WAIT mission command
- added vibration level logging
- ignore low voltage failsafe while disarmed
- added delta velocity and delta angle logging
- fix LOITER_TO_ALT to verify headings towards waypoints within the loiter radius
- allow rudder disarm based on ARMING_RUDDER parameter
- fix default behaviour of flaps
- prevent mode switch changes changing WP tracking
- make TRAINING mode obey stall prevention roll limits
- disable TRIM_RC_AT_START by default
- fixed parameter documentation spelling errors
- send MISSION_ITEM_REACHED messages on waypoint completion
- fixed airspeed handling in SITL simulators
- enable EKF by default on plane
- Improve gyro bias learning rate for plane and rover
- Allow switching primary GPS instance with 1 sat difference
- added NSH over MAVLink support
- added support for mpu9250 on pixhawk and pixhawk2
- Add support for logging ublox RXM-RAWX messages
- lots of updates to improve support for Linux based boards
- added ORGN message in dataflash
- added support for new "blue label" Lidar
- switched to real hdop in uBlox driver
- improved auto-config of uBlox
- raise accel discrepancy arming threshold to 0.75
- improved support for tcp and udp connections on Linux
- switched to delta-velocity and delta-angles in DCM
- improved detection of which accel to use in EKF
- improved auto-detections of flow control on pixhawk UARTs
- Failsafe actions are not executed if already on final approach or land.
- Option to trigger GCS failsafe only in AUTO mode.
- added climb/descend parameter to CONTINUE_AND_CHANGE_ALT
- added HDOP to uavcan GPS driver
- improved sending of autopilot version
- prevent motor startup with bad throttle trim on reboot
- log zero rangefinder distance when unhealthy
- added PRU firmware files for BeagleBoneBlack port
- fix for recent STORM32 gimbal support
- changed sending of STATUSTEXT severity to use correct values
- added new RSSI library with PWM input support
- fixed MAVLink heading report for UAVCAN GPS
- support LightWare I2C rangefinder on Linux
- improved staging of parameters and formats on startup to dataflash
- added new on-board compass calibrator
- improved RCOutput code for NavIO port
- added support for Septentrio GPS receiver
- support DO_MOUNT_CONTROl via command-long interface
- added CAM_RELAY_ON parameter
- moved SKIP_GYRO_CAL functionality to INS_GYR_CAL
- added detection of bad lidar settings for landing
Note that the documentation hasn't yet caught up with all the changes in this release. We are still working on that, but meanwhile if you see a feature that interests you and it isn't documented yet then please ask.
Comments
Tridge, I thought Tom's landing abort code made it into this release, but I don't see it in the release notes?
LAND_ABORT_THR PR#2759
Any chance we can get parachute landing?
Great work Tridge and team. I look forward to testing it.
+1 on a VTOL 4.0 version! Maybe a late Christmas pressie? ;-)
With the approach and landing sequence now ignoring fail-safes, specifically what failsafes are ignored and in what modes? Will it ignore GCS heartbeat interruption (aka comms loss to GCS), loss of GPS and geofence breach once on approach or landing? Is there alternatively a way to set which failsafes are applied, as for the medical express the geofence will likely need to be maintained regardless, and should always lead to flight termination?
Another possible idea would be the ability to pre-select safe landing zones, or alternatively to avoid certain higher risk landing zones as the case may be, and store these on the PXH, similar to the terrain feature. These could be set as alternative geofences towards which the aircraft would steer in case of a propulsion failure etc as to avoid roads or populated areas when it needs to land unpowered BLOS. I suppose this could be partially achieved in mission planing, like when navigating to comms hold waypoint on comms failure, but triggering landing at a certain WP instead. It could work like a "soft-flight termination" which increases the chance of recovering the aircraft and reduces ground impact or injury, normally resulting from a full flight termination. I think it would improve the overall safety case for operations at range should both comms and propulsion fail, as these areas could be preset on the PXH prior to flight and operate without user intervention. It could potentially be the the extension of exclusion areas, like those for around airports, but that are user specified according to local or real-time knowledge. An online database of these might also help reducing potentially dangerous and unintentional incursions. Doing so I think would significantly improve the outcome of a combined comms and propulsion failure (ie due to an electrical fault) and generally add another layer of safety to geofence use even outside of the competition.
Regards JB
I have been using the beta version of this for a couple weeks now and it is flying great! Definetly notice improvements with the EKF.
Thank you for all the hard work!
Good job Andrew and everyone involved on this update,...
Thank for take in consideration APM users too :) . Pixhawks looks much better to progress and we are migrating to there but we still enjoy new features in that process.
Hi Andrew,
Yes,,, now installs perfectly.
Great work and many thanks,
Nick.
@Nick, the problem turns out to be that our build server produces a very slightly larger binary for APM2 than my development system does. The slight difference takes it past the limit of the APM2 and the verify fails. I have replaced the binary with one from my development system. Can you try again now?
@Nick, I've reproduced it and I'm working on a fix. I hadn't noticed it as I do all my testing with command line upload on Linux.
Thanks Team for your continued work, much appreciated.
I have a problem with my apm 2.5. Installing the recent beta worked fine but when I have just installed the firmware I get a verification failure. Tried on my apm 2.6, same issue, beta fine, rc verification failure.
Any ideas,
Thanks,
Nick
-
4
-
5
-
6
-
7
-
8
of 8 Next