need track replay CH-47F AI Wingmen Exhibit Erratic and Self-Destructive Behavior During Unplanned Landings

[rgk86cavalry]

Game Mode: Single Player
Mission Type: Player CH-47F with 3 CH-47F AI wingmen
Map Tested: any

Issue Summary:

When attempting unplanned or dynamic landings (e.g. in open lots or roadways), AI CH-47F wingmen behave erratically and often crash violently despite having clear, safe landing zones.

Observed Behaviors:

AI wingmen approach normally, then:

Suddenly drop all collective from ~100 ft AGL and smash into the ground.

Attempt to flare, then pull max pitch and perform a backflip, overshooting the LZ and crashing.

Hover near the ground, wobble, then commit to a nose-first impact.

Break formation and RTB without command, even when ordered to land at a visible and accessible LZ.

Nearly land safely, then abort at the last second with a high-g pitch-up, loop, and fatal stall.

This behavior occurs without enemy presence, under calm conditions, and in flat, obstruction-free areas.

Expected Behavior:

AI CH-47F wingmen should:

Follow the player’s example or landing order and land safely in open terrain.

Maintain controlled descent profiles and adjust dynamically if terrain allows.

Not randomly abort or suicide-dive on final approach.

How to Reproduce:

Create a mission with:

Player CH-47F

2–3 CH-47F wingmen in formation

Fly to an open flat area not designated as an airfield or FARP.

Attempt to land without using scripted LZ or waypoint commands.

Observe wingmen behavior.

Additional Notes:

This appears to be a rotorcraft landing logic issue with the CH-47F. It makes ad-hoc logistics, troop insertions, and player-led dynamic landings nearly impossible in single-player scenarios.

[BIGNEWY]

please include a short as possible track replay example on Caucasus if possible. 

thank you 

[rgk86cavalry]

[rob10]

They're looking for a TRACK FILE, not a video.  You can save a track file from the debrief window after you close the mission.  A trackfile has information that ED can use to debug beyond what you may see in the a video.

[rgk86cavalry]

AI Landing Behavior – Detailed Report for Developer Review

thanks again for looking into this.

There are two separate AI landing behaviors observed in this mission track:

1. First Landing Attempt:
Location: Open field near the city, just off the airfield. The zone is mostly clear—only a single tree nearby.
Behavior: The AI helicopters begin their approach normally and hover at around 50–75 feet AGL. Then, without warning, they dump collective and crash hard into the ground. In several cases, the aircraft bounce off the surface with enough force to trigger rotor strikes.

2. Second Landing Attempt:
After I land, the AI begin their descent. They get close to the LZ, then suddenly abort the landing, climb out, and fly halfway back toward the airfield (roughly a 5-minute flight). Then they turn around and come back to attempt another landing—again resulting in a very hard impact.

This pattern repeats consistently. It seems like there may be a logic issue with the AI’s landing resolution, possibly tied to terrain or collision clearance checks.

ch47f test.trk

[rgk86cavalry]

CH-47F AI Wingman Landing Bug – Expanded Report with TacView Analysis

 

To the ED development team,

 

Thank you again for following up on the AI behavior issues. I've performed a deeper analysis using both the .trk and TacView .acmi logs and wanted to share additional findings that confirm systemic issues in AI CH-47F landing logic.

---

 

Previously Reported Issues (Recap)

1. First Landing Attempt – Sudden Collective Dump

AI helicopters hover at ~182 meters AGL, then abruptly drop collective.

They fall ~178 meters and violently impact the ground.

All three AI wingmen (Rotary-1-2, -3, -4) exhibit this behavior.

 

Time stamps of the drop:

Rotary-1-2: 17.65s

Rotary-1-3: 37.66s

Rotary-1-4: 39.73s

 

---

2. Second Landing Attempt – Abort, Fly Away, Return

After I land, AI units initiate a landing approach, then abort, flying away from the LZ.

Max distance before returning:

Rotary-1-2: 1.43 km

Rotary-1-3: 2.73 km

Rotary-1-4: 2.91 km

They eventually return and attempt another landing, but again drop collective and crash hard.

---

New Findings from TacView Review

Flight Behavior – Rotary-1-4

Following the abort, Rotary-1-4 continued flying away and reached over 3.5 km from my location before turning back.

Timeframe of this flight segment: ~90s to ~110s mission time.

The AI followed a wide arc, likely re-evaluating the LZ or resetting its approach logic.

Collective & Pitch Behavior

During both outbound and inbound flights, telemetry showed little to no pitch or collective modulation, suggesting limited AI decision-making in approach vectors.

On return, instead of flaring or hovering, the AI descended rapidly from 182 meters to 4.67 meters, matching the earlier crash pattern.

Impact Estimate

Descent: ~177 meters in under 2 seconds.

Estimated vertical speed: 30–40 m/s (~110–145 km/h)

Result: Ground bounce, potential blade strikes, and failure to complete a safe landing sequence.

---

Conclusion

The AI behavior for CH-47F helicopters, particularly during landings, appears to lack:

Proper flare or hover logic

Controlled descent rate management

Stable landing decision trees after aborts

This issue happens both with a clear LZ and after dynamic behavior like landing aborts. The current logic results in predictable, repeatable crashes regardless of terrain or AI skill level.

Track and TacView logs are available and highlight these issues clearly.

 

Thanks again for the hard work—hope this helps nail down the bug.

Tacview-20250421-201020-DCS.zip.acmi

Edited April 22 by rgk86cavalry