"Deep learning" build in MCE!?

[BerndB 349]

Hey Team &Gerald, may you remember that i´ve complained about some little issues with the latest Revision.

- MCE FO not responding anymore

- wrong altitude readouts

- constant repeating of "llocalizer alive, glideslope alive….etc."

Against the recommendation to go back to a former Version of MCE i´ve gave it a Chance. Now, after some more legs together withPF3 some of those issues disapeared by magic. My FO now reads out the correct altitudes again and Reports them to ATC like "Eurowings 7032 out of 12500, decending to 7000..." which was correct. And my FO doesn´t felt asleep anymore and somethingMCE seems to run stable again. Gerald, come on, you haven´t populate that you´ve integrated "Deep learning AI" allready into MCE 😉

Seriously i would like to know in simple words, from where MCE actually get Parameters like the actual altitude. Out of the SIM i guess. The additional Information must be something that MCE keeps in mind from the given atc instruction, like to which altitude we´re descending i. e., Right?

But the annoying "song" of the glideslope state is still left. The good News is that this seems to happen only on approaches into EDDH. I have the scenery fron JustSim in use. May be this helps to track down why this happen.

Cheers

[FS++ 712]

2 hours ago, BerndB said:

Hey Team &Gerald, may you remember that i´ve complained about some little issues with the latest Revision.

- MCE FO not responding anymore

- wrong altitude readouts

- constant repeating of "llocalizer alive, glideslope alive….etc."

Against the recommendation to go back to a former Version of MCE i´ve gave it a Chance. Now, after some more legs together withPF3 some of those issues disapeared by magic. My FO now reads out the correct altitudes again and Reports them to ATC like "Eurowings 7032 out of 12500, decending to 7000..." which was correct. And my FO doesn´t felt asleep anymore and somethingMCE seems to run stable again. Gerald, come on, you haven´t populate that you´ve integrated "Deep learning AI" allready into MCE 😉

Seriously i would like to know in simple words, from where MCE actually get Parameters like the actual altitude. Out of the SIM i guess. The additional Information must be something that MCE keeps in mind from the given atc instruction, like to which altitude we´re descending i. e., Right?

But the annoying "song" of the glideslope state is still left. The good News is that this seems to happen only on approaches into EDDH. I have the scenery fron JustSim in use. May be this helps to track down why this happen.

Cheers

Good to hear you're finally up and running with latest.

Probably something unusual in custom scenery, with regards glide slope. Report it to scenery builder. I believe they are part of the BGL files

Did you mean "The making of MCE"?..... Since yiou asked 😋

We had to actually build our own module that runs inside the simulator process.

fsInsider64.dll for P3D V4

fsInsider.dll for FSX or FS9

The dll gives us all the insight we need, regarding not only flight and engine parameters, but also, weather, and seemingly impossible things to access like read Native ATC, Radar Contact ATC and PF3 ATC options displayed in the sim, not to mention GSX or AES on-screen messages.

No need to use SimConnect at all, except with aircraft whose SDK require it (only PMDG 737, 747, 777 do)

This is why, unlike our competitor who needs to make assumptions, we don't have a panel where you click <After start> or <TAKEOFF> mode.

An intelligent FO should be able to work out the flight phase you're in automatically.😎

Aircraft switches are read and set via an ad-hoc dll (files in \Prepar3D v4\MCE dlls\ folder.

Depending on detected aircraft, one and only one of those dlls is loaded at any given time.

Trust me, you don't want to run the speech engine inside the simulator process via a gauge added to a panel.

You would be crossing your fingers, and stressed out, worrying about "upsetting the system" by saying the wrong thing at the wrong time, never mind speech engine going unresponsive on a long haul flight.

This is why MCE runs as an external process.

Should anything wrong happen to it (or the speech engine it hosts inside "mce.exe"), it could eventually die or you kill it on purpose. Your flight, the most important thing will hardly if ever be compromised.

You can re-start MCE mid-flight and it would feel like co-pilot went for a toilet break, with him getting immediately in sync with your flight phase.

Voice control won't be fun if you have to give instructions such as "say this before that", "don't do or say this" until you're in flight phase P2, etc.

We have trained him for good airman-ship (try to get him to retract the gear on ground, or extend flaps or gear at high speed).

It's our job to figure out all the critical stuff and prevent it being triggered accidentally. We want you to talk as naturally as possible.

This is why, even 10 years after initial release, we're still building and tweaking it.

We could have made 45 separate products (one for each fully supported aircraft).Nobody wants to go through 45 installations, so why would we inflict that on our user base?

One single product to cover all your voice control needs from Prepr3D V4 all the way down to FS9, and soon MSFS 2020 (Expected to be a free update if SDK happens to be similar to FSX and no royalties are due to Microsoft).

For those stumbling across this thread, be our guest, and have a go at the Demo

 

 

 

[BerndB 349]

On 11/1/2019 at 3:23 PM, FS++ said:

Did you mean "The making of MCE"?..... Since yiou asked

Damned…you uncovered my intention….😙

Seriously speaking, of course not. I would never want you offer your "Holy Gral". The reason why i asked is, that as you may remember, i´am using MCE in a Homecockpit environment together with Jeehell FMGS. And because of the much different interaction between the hardware, software via network and simconnect and the flightmodel simulation in P3D, this can tell me a little more about what to exspect from MCE behavior and could help to trace down issues if they are MCE internal or depending on my usage.

Thank you very much for your brief explanation. At least i´am happy that you´ve build your own independent "FSUIPC" with fsinsider, This would mean, that via this offset channel there shouldn´t be any interferences with simconnect nor Pete´s FSUIPC.

On 11/1/2019 at 3:23 PM, FS++ said:

For those stumbling across this thread, be our guest, and have a go at the Demo

Yeah, agree! To "those" honestly, MCE is a "must have" if you are seeking for "immersion" and what else matters in a simulation!?!.  

On 11/1/2019 at 3:23 PM, FS++ said:

No need to use SimConnect at all, except with aircraft whose SDK require it

 Because now where the latest revision of MCE seems to be the best and reliable one ever, you could spent all the sparetime you´re saving in not need to answer complaints anymore, to think about it if you may see a way in the future that MCE could interact even with Jeehell FMGS. Because Jeehell uses Simconnect for the communication between his FMGS Server and the SIM there is might be a way for MCE to step into it´s controls. The different FMGS applications like MCDU, FCU, PFD, OverheadPNL etc. are communicate via the FMGS internal channel which is called "widefmgs". But this is only needed if Jeehell FMGS is spreaded to more than one client via a network.

Cheers

Edited November 4, 2019 by BerndB