Add generic test for inverted-Heli flying

[peterbarker]

Summary

Iterates the Heli frames, trying to fly each inverted

Classification & Testing (check all that apply and add your own)

• Checked by a human programmer
• Non-functional change
• No-binary change
• Infrastructure change (e.g. unit tests, helper scripts)
• Automated test(s) verify changes (e.g. unit test, autotest)
• Tested manually, description below (e.g. SITL)
• Tested on hardware
• Logs attached
• Logs available on request

Description

Follow-up I promised in #33521 (and currently has patches from that PR in here). It's actually only a single patch on top! I rebased

Iterates the Heli frames and changes the tuning a little to allow for inverted flight

  Per-frame inverted-flight performance

  ┌───────────────┬───────────┬──────────────────┬──────────────────┬────────────────────┬───────────────────────────┐
  │     frame     │ inverted? │ alt drop (tuned) │ headroom (tuned) │ alt drop (default) │          result           │
  ├───────────────┼───────────┼──────────────────┼──────────────────┼────────────────────┼───────────────────────────┤
  │ heli          │ yes       │ 0.0 m            │ 140 µs (0% sat)  │ 6.1 m              │ tuned PASS / default FAIL │
  ├───────────────┼───────────┼──────────────────┼──────────────────┼────────────────────┼───────────────────────────┤
  │ heli-blade360 │ yes       │ 0.1 m            │ 115 µs (0%)      │ 6.2 m              │ PASS / FAIL               │
  ├───────────────┼───────────┼──────────────────┼──────────────────┼────────────────────┼───────────────────────────┤
  │ heli-ddfptail │ yes       │ 0.1 m            │ 140 µs (0%)      │ 6.2 m              │ PASS / FAIL               │
  ├───────────────┼───────────┼──────────────────┼──────────────────┼────────────────────┼───────────────────────────┤
  │ heli-ddvptail │ yes       │ 0.0 m            │ 140 µs (0%)      │ 6.7 m              │ PASS / FAIL               │
  ├───────────────┼───────────┼──────────────────┼──────────────────┼────────────────────┼───────────────────────────┤
  │ heli-dual     │ yes       │ 1.2 m            │ 106 µs (0%)      │ 6.4 m              │ PASS / FAIL               │
  ├───────────────┼───────────┼──────────────────┼──────────────────┼────────────────────┼───────────────────────────┤
  │ heli-gas      │ yes       │ 0.1 m            │ 105 µs (0%)      │ 22.9 m             │ PASS / FAIL               │
  └───────────────┴───────────┴──────────────────┴──────────────────┴────────────────────┴───────────────────────────┘

  The story: every frame rolls inverted fine in both configs — attitude control isn't the limiter. What separates pass from fail is collective authority. With the test's tuned range
  (−12°/H_COL_MIN 1260), all six settle into steady inverted hover with ~105–140 µs of collective headroom to spare and 0% saturation, holding altitude within ~1 m. With the default range
  (−2°/1460), the collective pins at its floor (100% saturated) — it physically can't make enough negative thrust — and altitude collapses 6–7 m (heli-gas worst at 23 m; its throttle
  governor behaves differently, only 33% saturated but still loses the most). So the maneuver is gated by negative-collective range, not the controllers, and all six frames are comfortably
  capable once given the range — none is marginal.

  ┌───────────────┬───────────────────────────┬────────────────────┬────────────┐
  │     frame     │ peak yaw drift (inverted) │ yaw actuator range │ saturated? │
  ├───────────────┼───────────────────────────┼────────────────────┼────────────┤
  │ heli          │ 40°                       │ 1542–1846          │ no         │
  ├───────────────┼───────────────────────────┼────────────────────┼────────────┤
  │ heli-blade360 │ 42°                       │ 1255–1430          │ no         │
  ├───────────────┼───────────────────────────┼────────────────────┼────────────┤
  │ heli-ddfptail │ 28°                       │ 1284–1969          │ no         │
  ├───────────────┼───────────────────────────┼────────────────────┼────────────┤
  │ heli-ddvptail │ 40°                       │ 1541–1833          │ no         │
  ├───────────────┼───────────────────────────┼────────────────────┼────────────┤
  │ heli-dual     │ 179°                      │ 1410–1776          │ no         │
  ├───────────────┼───────────────────────────┼────────────────────┼────────────┤
  │ heli-gas      │ 43°                       │ 1604–1675          │ no         │
  └───────────────┴───────────────────────────┴────────────────────┴────────────┘

  The yaw story (and the justification for not gating the generic test on heading):
  - Every frame's heading drifts while inverted — typically ~30–45°, because a tail rotor's yaw torque direction effectively reverses relative to the airframe when upside down, so the
  controller is fighting a re-trimmed plant. The drift is bounded and slow, not a spin-out — the yaw actuator works steadily and never saturates (stays off the 1000/2000 stops on all
  frames), so yaw control is retained, just with a standing heading offset.
  - heli-dual is the outlier: ~179° — it essentially yaws around while inverted. A tandem/dual-swash heli has fundamentally different yaw authority (differential collective, no tail rotor),
  and inverted it loses heading completely even though it holds altitude fine (1.2 m drop, passes the test). This is exactly why heading is excluded — heli-dual would otherwise fail on an
  axis orthogonal to "can it sustain inverted flight."

[peterbarker]

  "tuned" / passing = the params your test sets (what you gave me):
  - H_COL_MIN 1260, H_COL_ANG_MIN -12 → the swashplate can drive collective down to −12° blade pitch
  - (IM_STB_COL_* too, but those only matter in STABILIZE; the maneuver is ALT_HOLD)

  "default" / "untuned" / failing = the stock heli defaults (from copter-heli.parm, with nothing extra applied):
  - H_COL_MIN 1460, H_COL_ANG_MIN -2 → collective only reaches −2° blade pitch

  That's the entire difference between the two runs — just how far negative the collective can go. Same airframes, same controllers, same maneuver, same everything else. To get the "failing"
  data I temporarily reverted those two values to the stock defaults, flew, captured the logs, and restored the file.

  So it's not "tuned" in the sense of PID/gain tuning — it's purely collective travel range. −2° can't make enough downward (inverted: upward) thrust to hold the vehicle up while upside
  down, so altitude collapses; −12° can, so it holds. If the labels are misleading I can relabel the graphs to something exact like "−12° collective (test params)" vs "−2° collective (stock 
  defaults)". Want me to do that?

[bnsgeyer]

@peterbarker Wow! You have gone above and beyond on your analysis. I didn't expect the default values to pass and that is why I provided the collective parameter changes. there is one other parameter that probably should be changed to aid in better transition. The default setup has the ATC_RATE_P_MAX and ATC_RATE_R_MAX set to 0 (no limit) but the models don't have infinite pitch and roll rate capability. So I always set a limit to allow the aircraft to keep up with the command model as it commands the roll to inverted and back. I think it would be safe to set those both to 120 deg/s. I don't think it will affect the performance.

I really didn't intend to check all frames. I think it is redundant to check Blade360, ddfp tail, ddvp tail, and the gas versions. They are all single main rotor/tail rotor configs. The dual heli is interesting. the yaw problem may be a factor of control reversal since the thrust vector is negative and dual heli's use the lateral component of thrust to generate yaw moments.

[bnsgeyer]

Just tested Dual heli and I think you should drop that one from the inverted flight test. I don't have a good fix for dual heli in order for it to have proper yaw control while inverted. So that would just leave you with the one traditional heli config. which I think is all that is needed for now. Sorry, I didn't mean for you to do all that work.