Iris: update iris_with_gimbal model to 3d gimbal

[srmainwaring]

Updates #86 which was on a main branch preventing maintainer edits.

• Update the gimbal model used by the Iris to 3d and add yaw control channel.
• Use the iris_with_gimbal model as default in example worlds.
• Reduce position controller PID gains to reduce potential oscillation.
• Update channels to match servo gimbal wiki doc: https://ardupilot.org/copter/docs/common-camera-gimbal.html
• Update servo control direction

Action	Channel	RC Low	RC High	Min (deg)	Max (deg)
Roll	RC6	Roll Left	Roll Right	-30	+30
Pitch	RC7	Pitch Dn	Pitch Up	-135	+45
Yaw	RC8	Yaw Left	Yaw Right	-160	160

Copter params

# Iris is X frame
FRAME_CLASS	     1
FRAME_TYPE	     1

# Match servo out for motors
MOT_PWM_MIN      1100
MOT_PWM_MAX      1900

# Gimbal on mount 1 has 3 DOF
MNT1_TYPE        1           # Servo
MNT1_PITCH_MAX   45
MNT1_PITCH_MIN   -135
MNT1_ROLL_MAX    30
MNT1_ROLL_MIN    -30
MNT1_YAW_MAX     160
MNT1_YAW_MIN     -160

# Gimbal RC in
RC6_MAX          1900
RC6_MIN          1100
RC6_OPTION       212         # Mount1 Roll
RC7_MAX          1900
RC7_MIN          1100
RC7_OPTION       213         # Mount1 Pitch
RC8_MAX          1900
RC8_MIN          1100
RC8_OPTION       214         # Mount1 Yaw
RC8_REVERSED     0           # Normal
RC9_MAX          1900
RC9_MIN          1100
RC9_OPTION       0           # Do Nothing

# Gimbal servo out
SERVO9_FUNCTION  8           # Mount1Roll
SERVO10_FUNCTION 7           # Mount1Pitch
SERVO11_FUNCTION 6           # Mount1Yaw

Figure: gimbal tracking ROI located at the home position while Iris circles

The GstCameraPlugin is also enabled by default. To view the video in QGC edit Application Settings > Video Settings. Select UDP h.264 Video Stream and use the default port 5600.

Testing

Tested on macOS Sonoma 14.5 and Ubuntu 22.04 (VM). Both running ArduPilot master with Gazebo Harmonic.

[rmackay9]

@srmainwaring,

Thanks so much for this. This is quite important for the high-altitude-non-GPS GSoC project!

[rmackay9]

Hi @srmainwaring,

What should MNT1_TYPE be set to? I think it will be either 1 (Servo) or 7 (BrushlessPWM?)

[srmainwaring]

What should MNT1_TYPE be set to? I think it will be either 1 (Servo) or 7 (BrushlessPWM?)

Forgot to add that: yes MNT1_TYPE should be 1 for servo. Updated PR notes at top.

[rmackay9]

Hi @srmainwaring,

Sanket, Tridge and I were playing around with this today and noticed a couple of issues:

1. the servos seem slightly laggy which causes the gimbal to move when the vehicle moves. In our high altitude project we expect that users will use brushless gimbals that will be very stable so we'd like to get close to perfect stabilization
2. The gimbal's pitch servo seems to be limited to 90 degrees downwards so when the vehicle pitches backwards the gimbal physically can't point downwards

Anyway, thanks again for this though, it is on the critical path for both the HighAlt project and the Visual follow-me project!

[srmainwaring]

1. the servos seem slightly laggy which causes the gimbal to move when the vehicle moves

There's a couple of possible causes for the lag which I'll look into. The most likely is the PID gains on the servo joint controllers in the simulation.

2. The gimbal's pitch servo seems to be limited to 90 degrees downwards so when the vehicle pitches backwards the gimbal physically can't point downwards

I'll adjust the joint limits and update the AP params to match the specs for the SIYI A8

• Controllable Pitch Angle: -135° ~ +45°
• Controllable Yaw Angle: -160° ~ +160°
• Controllable Roll Angle: -30° ~ +30°

Finally, the iris model could also do with a tune. IIRC from some other analysis the default params for gazebo-iris are quite close to the limit where roll oscillations become noticeable. There could be some cross-coupling with the gimbal movement that is also contributing to stability.

[rmackay9]

Hi @srmainwaring,

Thanks very much! By the way, in the instructions at the top of the PR (and maybe eventually on the wiki or in the README) we should specify what these MNT1_xxx parameters should be set to:

• MNT1_ROLL_MIN/MAX -90 and +90?
• MNT1_PITCH_MIN/MAX -90 and +90?
• MNT1_YAW_MIN/MAX -90 and +90

and I wonder if we also need to set these values? They seem to be ok that the defaults 1100 ~ 1900 but just in case

• SERVO9_MIN/MAX (used for roll, 1100~1900 is OK?)
• SERVO10_MIN/MAX (used for pitch, 1100~1900 is OK?)
• SERVO11_MIN/MAX (used for yaw, 1100~1900 is OK?)

[srmainwaring]

@rmackay9 and @snktshrma - should be much snappier now. The limits now match the SIYI A8. Main change to joint control is reducing the damping (equivalent to increasing the PID gains and max force exerted). The original physics settings were carried over from the model from Gazebo Classic - and correspond to a very stiff joint, not that appropriate for a small gimbal.

Figure: gimbal tracking ROI located at home position (location of axis), response is faster with new settings.

[rmackay9]

Thanks again @srmainwaring for this!

@snktshrma has told me that the latest changes greatly improve the attitude control of the gimbal which resolves the biggest blocker he was facing in the high-altitude-non-gps project