Automobilista 2 V1.5 Officially RELEASED - Now updated to V1.5.0.5

Worth mentioning there are still quite a few places that suck you off the track when you get close to the edge of the tarmac. So I really don't want the grass becoming a insta-wreck until that's fixed.

 

Is this from dirt build-up on the track (that you can't always see clearly)? I have experienced sudden loss of grip under power or acceleration at certain places, but this has improved with the new tread adjustments on most vehicles. Please give a few examples and let's see if it is an area the AI throw a lot of dirt on the track.

 

Perhaps I didn't pass the message correctly.
I also think the grass grip is fine, I was talking more about the unsettle caused by ground elevations changes (imperfections), that have the potential to throw the car "off the rails". Like when we cut the track into the grass/gravel and sausage kerbs, for instance.
We race cars at their limit and a small lateral/longitudinal force (other car) or vertical force (ground) have the potencial to upset the car stability. I thing AMS2 is very driver friendly , and should stay that way for those who like it that way, but can have an evolutionary setting towards simulation.

I should also mention that a big part of the grow potential is related to driver feedback, most noticeable because we can change FFB settings/files (an AMS2 plus over all other sims!!) - the same game, same physics, but it changes completly "the feeling" of it. Picking our own choice is really a big plus!
To me, visual feedback could take another go (we already have legacy and the "current" one), but I would like one steadier ("normal" track circunstances) and harsher (when the car is shaked by something). Other sims do this very well (we can really feel the car is getting apart ).
I see/feel a lot of low pass filters and sine waves that shouldn't be there, I want the raw experience .

 

So is the public build and beta build the same now? I know sometimes there are silent updates that don't change the version numbers.

 

The build number (the last part) always changes with an update, even though occasionally the game version number does not (if the .exe has not been revised).

 

Will do tomorrow!

 

Yeah, that is why I wondered if they were the same now or not, I am not near my pc to find out.

 

Horray!! Thankyou so very much!!! I've not tried all of these yet but, I've never been able to have a competitive race with changing weather with the GTE's because they used to just slow down to a snails pace after it started to rain. Now however, after a quick test (might need a longer one to double check), it seems they are equal to me in either wet or dry at my choosen skill level! So just wanted to say thanks and BRAVO!!!

 

P.S. The GT3's have never been as bad as the GTE's (which are now great in this regaurd!), but have either been a bit too slow or fast when it gets wet. Just done a quick test, and these now seem pretty damn spot on aswell! Can't wait to try the other classes! I'm doing all my tests Nurbury Veedol as the track has also had its A.I. paths and behaviour refined many times now. Bravo!

 

Just tried Group C. Id say overall balance between wet and dry is good, maybe slightly faster than me in wet, but also, in terms of speed overall, at the same skill setting as tests with GTE's & GT3's, they are slower. They used to be much faster, so wish overall speed of A.I. with these cars was just a litttle faster. But so far, in terms of blance between wet and dry skill, all tests are positive, much better and more refined than previously!

 

I'm not sure but I feel driving on grass is a bit more forgiving than before, did Reiza change anything about the grip level on the grass ? I think it's cool, you still loose a ton of time when you go offtrack but it doesn't feel like oil on ice like before.

 

in other Sims, this isn't possible..predict fans of other sims that you had to crash instantly, Alonso rallies at 300 km / h with an F1 ... I think it's Simcade

 

Just tried GT4's. Overall speed is very good, very competitive at same skill level as all tests (which has been 110 btw). From dry to wet though, id say they are definately a touch too fast in wet. Still very happy with improvements though!

 

Mostly refering to other games

 

AMD desided by Graphic's card Software was too old, will need to update to get my recording capablilties back

 

I found the same in the GT4s, but it was a week or so ago before the latest AI adjustments. On the two tracks I tested in the wet, AI were only two seconds off dry pace on a soaked track, so I'm glad it's a bit better now, even if it needs more tweaking.

 

Well in theory you don't need to change the game or get reiza to create a "real" or maybe the better term is machine learning AI like Sophy. Any one with the machine learning chops, a copy of the game, the time and interest could work on it. And perhaps access to enough processing power to actual do the training which may be substantial.

A human player gets output from the game in terms of FFB, some audio samples and (in VR) 2 images giving stereoscopic vision. And the input they give is basically a steering angle and a couple of pedal positions and a few button presses.

To AM2 the "machine learning" player would look like a new brand of FFB wheel and pedals, little different from it seeing a 'logitech driving force GT' 'fanatec blah blah' or whatever.

You give your machine learning algo the same outputs we get and get it to create inputs which you can then feed into the game in a loop via whatever MS SDK exists for FFB controllers / keyboards or whatever. If AM2 has motec data / telemetry (I assume it does but I've only played for 12 hours) you can probably get another bunch of useful input in real time from that.

I'd guess Sony probably use information from the game engine internals which would potentially suit the AI better than getting images, but may not be strictly necessary for machine learning to work (especially if the telemetry data has track coordinates)

i.e someone could work on this and create it and they wouldn't have to be the game developer or change a single line of AM2 code. You could add a machine learning player developed like this just by running a copy of the game - albeit it would probably be performance intensive.

It's not easy of course, but maybe if you found some smart university student looking for something to do for his comp sci final project and who is into sim racing...

The downsides? If you created it (and it drove as well as the Sony version) it would be an obvious vector for cheating.

 

Sounds a bit too easy IMHO.
How should Reiza not have to change a single line of code if they wanted to implement such an AI when it comes to the reaction of the AI opponents (following driving lines, overtaking and defending or adapting setups and strategies)?

 

I just have to say! I love this! I'd not seen this before today whilst checking something on the Steam AMS2 store page. Its great that this is the first thing that you see on the page now, a great overview of the game showing its massive amount of features, reallly pulls you in and gets you interested! Very nicely narrated aswell, is that you @steelreserv ? Very nicely done, Inspiring you might say, certainly would make me wanna rty and buy if I did'nt already own the sim!

After seeing it on steam I went and found the same video on Reiza's YT channel to share. If you've not seen it yet, its definately worth a watch to refresh yourself of just how much is going on in the sim! I've not really got too much time to drive it today, but after watching this I've just gotta go for a quick lap or two in it. Bravo!

 

yeah, now they just need to update the Demo as well, maybe add 1 or 2 different cars to the demo, and the latest physics, as you still see people trying the demo and not liking it, the demo should also be something that draws people into AMS2 IMHO