mavlink HIL with normal autostart scripts

[DrTon]

Work on seamless HIL switching. User will not need to set special SYS_AUTOSTART, but only set HIL mode with mavlink command. mavlink app will start sending HIL_CONTROLS stream and listening HIL_SENSORS/HIL_GPS/HIL_etc streams.

mavlink app already done, need to implement HIL mode support in gps and maybe other apps.

[LorenzMeier]

@DrTon There is a publication suppression method implemented in commander and the device nodes for HIL mode. This just needs to be implemented in each device so it doesn't publish during HIL. Of course the actuators should silently ignore this, but all sensor drivers should stop publishing. I've tested this partially and it works fine for sensors, GPS and the individual sensor drivers including airspeed.

[julianoes]

I'm getting timeouts about every 5s when trying this with flightgear.

[DrTon]

Which SYS_AUTOSTART?

[julianoes]

I'm using SYS_AUTOSTART 3032

[DrTon]

Connected via USB, right? Try to start mavlink at lower message rates, with -d 5000 or even -d 3000. Maybe data rate is too high for qgroundcontrol (I had such issue: qgroundcontrol was not working with high data rate, while jMAVSim connected directly to FMU worked perfectly, without any lost messages).

[julianoes]

-r 3000 worked, -r 5000 didn't

The GPS position is initialized at 0,0 and 0 alt, that's still wrong. Also RC input doesn't work and if you press the safety switch the plane suddenly jumps and the estimator goes nuts.

[LorenzMeier]

@julianoes I think that QGroundControl indeed is a real issue here. I recommend to pipe it through MAVProxy - I need to fix the serial port. Not sure why it shows up now.

[DrTon]

Gps driver still publishes when HIL enabled, so this issue is expected. See first two comments.

[LorenzMeier]

@julianoes I think I fixed your comments in beta_mavlink2. I recommend to merge this, as these changes here do not fundamentally change the beta_mavlink2 operation.

Note the issue with the QGC serial port improved after the last set of beta_mavlink2 changes, as the effective data rate dropped. I still need to fix QGC though.

[LorenzMeier]

@julianoes @DrTon Tested successfully with fixed wing HIL.

[DrTon]

What about normal gps driver publishing? Doeas publishing blocking ioctl works now?

[LorenzMeier]

@DrTon I forgot - need to check.

[DrTon]

Now we have this, on enabling HIL. Not good for merging, I think.

commander: Current state: 0, requested state: 1
Disabling accel: ERROR
Disabling accel1: ERROR
Disabling adc0: ERROR
Disabling baro: ERROR
Disabling console: ERROR
Disabling gps: ERROR
Disabling gyro: ERROR
Disabling gyro1: ERROR
Disabling hmc5883: ERROR
Disabling l3gd20: ERROR
Disabling led: ERROR
Disabling lsm303d_accel: ERROR
Disabling lsm303d_mag: ERROR
Disabling mag: ERROR
Disabling mag1: ERROR
Disabling mavlink: ERROR
Disabling mmcsd0: ERROR
Disabling mpu6000_accel: ERROR
Disabling mpu6000_gyro: ERROR
Disabling ms5611: ERROR
Disabling mtdblock0: ERROR
Disabling mtdblock1: ERROR
Disabling null: ERROR
Disabling pwm_output: ERROR
Disabling px4io: ERROR
Disabling ram0: ERROR
Disabling rgbled0: ERROR
Disabling tone_alarm: ERROR
Disabling ttyACM0: ERROR
Disabling ttyS0: ERROR
Disabling ttyS1: ERROR
Disabling ttyS2: ERROR
Disabling ttyS3: ERROR
Disabling ttyS4: ERROR
Disabling ttyS5: ERROR
Disabling ttyS6: ERROR
commander: directory listing ok (FS mounted and readable)

[LorenzMeier]

@DrTon I've fixed this now I think:

Disabling /dev/adc0: ERROR
Disabling /dev/baro: OK
Disabling /dev/gps: OK
Disabling /dev/gyro: OK
Disabling /dev/gyro1: OK
Disabling /dev/l3gd20: OK
Disabling /dev/led: OK
Disabling /dev/lsm303d_accel: OK
Disabling /dev/lsm303d_mag: OK
Disabling /dev/mag: OK
Disabling /dev/mpu6000_accel: OK
Disabling /dev/mpu6000_gyro: OK
Disabling /dev/ms5611: OK
<hilservo> ioctl 0x0100 0x00000001
Disabling /dev/pwm_output: OK
Disabling /dev/px4io: ERROR
Disabling /dev/rgbled0: ERROR
Disabling /dev/tone_alarm: OK

[LorenzMeier]

@DrTon @julianoes I think this is good to merge now - including all bells and whistles. It does not yet fully enable HIL on arbitrary airframes, but the 'old' HIL approach still works and it could work now with any autostart script modulo smaller adjustments.

[LorenzMeier]

@DrTon I found a few more corner cases - now everything is nicely disabled, but this involves RC control 8). I need to re-enable a few core topics.