Add MultiWinchController

[ufechner7]

Add a controller that has as input set_torque and set_speed (one of them should be NaN or nothing) in addition to set_depower and set_steering, and as ouput set_torque_l, set_torque_r and set_torque_p. See:

This would allow to use the existing controllers for the ram air kite.

[baptistelabat]

Thanks Uwe.

I think we can start with a linearised version, no saturations at all, assuming lines long enough and point model. KISS.

However, let's keep in mind the difficulties which could arise in real world or with more evolved models: Optimal dynamic allocation might be needed at one point !

• Can we assume set_torque or set_speed is allocated only to Winch_P (not summing Winch_R and Winch_L torques) ?
• Need to compensate for the distance between attachment points on ground/boat when kite is moving in wind window and differential created in line length ? -> would require kite position relative to ground station and attachment points position.
• Need to take account of Pythagora when lines length l becomes short in front of distance between attachment points on kite which is related to kite span (when l>>b is not verified). ? -> would require attachment points position
• Need to take into account the maximal rotating speed (rate limiter) of each winch to saturate orders giving priorities for example to steering over depower, to depower over torque/length ? -> would require these rates
• Need to take into account change in drum diameter due to reeling lines already on winch ?
• Need to take into account position saturation of each winch to be able to apply priorities (on this point I think we assume multiturn winches in your mind as we are able to change line length, but it is not true for all setup) ?
• Need to take into account acceleration limit of each winch ?
• Need to take into account torque limit of each winch ?
• Do we want to be able to take into account torque feedback from winch ?
• Do we need feedback to the controller if setpoints can not be reached ?
• Do we need to be able to deal with failure of winch ?
• Ability to support other combinations (one motor for steering, one motor for depower, one for all the lines ?) -> the mapping is mechanically done with pulleys instead of a software controller

++
Edit:
-Do we need to compensate for different and varying line sag between power and steering lines ?
-Do we need to compensate for extra forces requires due to extra friction when line are twisted ?
-Do we need to compensate for variation in line length due to twist (when l>>b not verified) ?
-Do we need a feedback of kite position in term of angle of key (relative to lines direction), twist or bank angle ?

Most of these questions are related to the winches on ground (some would not arise in the same way with a pod), but independent of the choice of winch configuration (power, left, right)

[baptistelabat]

For the steering winches, I think it would make sense to use a length controller instead. I expect a speed feedforward would be needed as well when varying length with a realistic model. From an organizational and architecture point of view, I think this would give priority to the following issues:

• Support length control #9
• Use feed-forward in a winch controller #14

[baptistelabat]

As suggested on signal group, the following package should be used:

• https://github.com/baggepinnen/TrajectoryLimiters.jl
• https://github.com/JuliaControl/DiscretePIDs.jl

[ufechner7]

Thanks Uwe.

* Can we assume set_torque or set_speed is allocated only to Winch_P (not summing Winch_R and Winch_L torques) ?

I would say no. The set torque is the sum of the torques. For the set speed we might ignore the impact of the right and left winch speeds for now. It cancels out while steering. It does not cancel out when depowering the kite. But if we want to take this into account we need to know the geometry of the kite.

* Need to compensate for the distance between attachment points on ground/boat when kite is moving in wind window and differential created in line length ? -> would require kite position relative to ground station and attachment points position.

Not now.

* Need to take account of Pythagora when lines length l becomes short in front of distance between attachment points on kite which is related to kite span (when l>>b is not verified). ? -> would require attachment points position

Not now.

* Need to take into account the maximal rotating speed (rate limiter) of each winch to saturate orders giving priorities for example to steering over depower, to depower over torque/length ? -> would require these rates

Yes. But this is already defined in the section "winch" of system.yaml.

winch:
    winch_model: "TorqueControlledMachine" # or AsynchMachine
    max_force: 4000        # maximal (nominal) tether force; short overload allowed [N]
    v_ro_max:  8.0         # maximal reel-out speed                          [m/s]
    v_ro_min: -8.0         # minimal reel-out speed (=max reel-in speed)     [m/s]
    drum_radius: 0.110    # radius of the drum                              [m]
    max_acc: 4.0           # maximal acceleration of the winch               [m/s²]
    gear_ratio: 1.0        # gear ratio of the winch                         [-]   
    inertia_total: 0.024   # total inertia, as seen from the motor/generator [kgm²]
    f_coulomb: 122.0       # coulomb friction                                [N]
    c_vf: 30.6             # coefficient for the viscous friction            [Ns/m]
    p_speed: 1.0           # proportional gain of the winch speed controller [-]
    i_speed: 0.1           # integral gain of the winch speed controller     [-]

This assumes that the three winches are identical, which might be wrong in practice. But for now it is good enough.

* Need to take into account change in drum diameter due to reeling lines already on winch ?

No. For a good lifetime of the tether there should be only one layer.

* Need to take into account position saturation of each winch to be able to apply priorities (on this point I think we assume multiturn winches in your mind as we are able to change line length, but it is not true for all setup) ?

The length controller must take into account minimal and maximal line length. But that is a separate issue.

* Need to take into account acceleration limit of each winch ?

Yes, see settings above.

* Need to take into account torque limit of each winch ?

Yes, in the settings it is called force limit, but because the radius is defined you can convert it to a torque limit. I think the set torque relates to the motor, so you need to take the gear ratio into account.

* Do we want to be able to take into account torque feedback from winch ?

The TorqueControlledMachine works in torque control mode which is very fast, so set value and actually value can assumed to be the same (as long as the battery is not weak and the design limits are respected).

* Do we need feedback to the controller if setpoints can not be reached ?

Well, we have set points for depower and steering, and actual values for both values. So yes, the actual values should be reported back.

* Do we need to be able to deal with failure of winch ?

Not now.

* Ability to support other combinations (one motor for steering, one motor for depower, one for all the lines ?) -> the mapping is mechanically done with pulleys instead of a software controller

Not now.

++ Edit: -Do we need to compensate for different and varying line sag between power and steering lines ?
No.
-Do we need to compensate for extra forces requires due to extra friction when line are twisted ?
No. We assume they are not twisted.
-Do we need to compensate for variation in line length due to twist (when l>>b not verified) ?
No.

• Do we need a feedback of kite position in term of angle of key (relative to lines direction), twist or bank angle ?
  The winch controller does not need to know where the kite is. The flight path controller needs to know it.

Most of these questions are related to the winches on ground (some would not arise in the same way with a pod), but independent of the choice of winch configuration (power, left, right)
There is a controller for kite control pods, see: https://github.com/aenarete/KitePodModels.jl (well, this model includes a very simple P controller).