[mirekkr]

Good evening,
Among those wonderful features of OR (just as the rotating turntable - many thanks, Carlo!), I find something strange time from time.
This time I made an activity where an AI pusher couples to my rear to help me climb uphill. This scenario works OK in OR.
But ... The Helper engine connected this way really does not like its work: its power starts to drop down very soon after the player train accelerates.
Enclosed is a simple activity for the default Marias Pass to reproduce the behavior.
- Please do NOT use Autopilot when testing
- Use HUD with Locomotive Information displayed
- Start your train before Essex (for the 1st test ignore the Dash9 behind you) and continue to Essex Yard.
- When prompted, stop your train after the switch and wait for the AI Helper to start (9:05) and connect to your rear
- Go forward; as you throttle and accelerate, watch the loco values in the HUD. (If you are patient enough, you can try to climb up to Java.)
- You can also zoom to the wheels of the AI Helper and watch that they rotate very lazy (but they do so even if you stop :-))
For the comparison, I've prepared another Dash9 (not AI) at the very start of the activity. As a 2nd test, you can connect to it at the activity start, and you'll have loco values to compare.

I may have done something wrong in the activity design - or is it a bug?

Thank to all the team for excellent work,
Mirek
 AI_Helper_Act.ZIP (3.97K)
Number of downloads: 182

[Csantucci]

Does this occur both with advanced adhesion on and off? Which one of the Power values present on the Loco info page decreases incorrectly?

[edwardk]

I did experience a broken path error with the above test activity.

Edward K.

[copperpen]

First run with this test activity showed the AI helper as having about 15 more HP that the lead engine on the run up to Java. This is common for flipped engines. Will have another run tomorrow and observe the helper wheels as well.

[mirekkr]

Thanks for answers, guys.
I've enclosed some pictures from my run of the activity.

Comments:
Pict 1 - AI just connected (values are OK)
Throttled up to N3, then to Idle, then used Dynamics.
Pict 2 - After Essex Gallery, throttled up to N2 (the helper "gets lazy" and does not work)
Leaving throttle at N2.
Pict 3 and 4: notice the values of Power for the Helper: 96, ..., 85, ... 74 ...
Because the rails go down for a while, I must throttle to zero : this resets the values to those at start (pict 5)
Climbing up to Java - ok.
At Java bridge, the situation repeats (pict 6).
Pict 7 shows how it looks like at Java West (snapped from behind). We are at 63 hp now and the value is constantly and slowly decreasing. I can hardly go up this grade. Also notice that the Helper produces no exhaust compared to the lead engine.
If there was no need to throttle down to zero, the value of Power of Helper would continue decreasing.

To Carlo: In the attached zip, there's Openrails log with the settings you asked for. I dare say that this has nothing to do with adheasion because the non-AI helper engine works correctly.

To Edward: OR Track Viewer does not show any broken path for me.

Thanks all for help,
Mirek

[mirekkr]

...and the attachment:

Attached File(s)

•  AI_Helper_act_pictures.ZIP (14.23MB)
  Number of downloads: 190

[mirekkr]

... and yet one more picture, which shows the Loco Info where:
-- 1st Dash9 is the player
-- 2nd Dash9 is non-AI helper
-- 3rd Dash9 is AI helper
And one new experience: if you save and restore game, the problem occurs never more.

Regards,
Mirek

Attached File(s)

•  Comparison_AI_nonAI.ZIP (284.19K)
  Number of downloads: 176

[Csantucci]

I have committed a patch in x.3563. The problem was due to the fact that AI trains have the AntiSlip flag forced to true, while player trains and static consists have the AntiSlip flag practically forced to false (because the AntiSlip flag is present in the Wagon section of the .eng files, but OR looks for it in the Engine section). Locos with AntiSlip flag to true aren't correctly managed by the OR physics software using Advanced Adhesion (e.g. their WheelSpeedMpS is not updated), and therefore their physics behave weirdly. I have created a patch that, when an AI train couples to a player train, sets the AntiSlip flag of the former AI train locos to the value of the AntiSlip flag of the player locomotive.
However this should be considered only a temporary patch, waiting that the antislip logic is implemented.

[mirekkr]

Carlo,
thank you very much for your work.
I'll test your patch tomorrow when I have more time. I'll let you know about my result.

Again, many thanks to you and all the people here!
Mirek

[mirekkr]

I can confirm that it is corrected now.
AI Helper now does help :-)

Thanks,
Mirek