30 players online racing with pings larger than 1000 ms :o)))

As you all know you get pings around 1000 ms to the other side of the world over the internet. This makes online racing troublesome if we want to race globally, which we of course want :o), especially since most of the fastest sim racers live in Australia, at least according to the results in the Grand Prix Legends demo at [link lost in import] I've previously posted to this NG a suggestion to a solution were you just have collision detection turned off, like the ghosts in F1RS. Some liked this idea and some didn't :o). We all would like to have collision detection turned on, but this is very much more difficult to implement in a racing simulator, compared to the ghost multiplayer solution and when you have to allow pings over 1000 ms. Byron Forbes was one of those who really wanted collision detection on and I've had some really nice discussions with him on this matter. I've also discussed a lot with Ian Lake. The result of these discussion is a suggestion to a multiplayer feature which allows for pings between players to be well over 1000 ms, and here it is: The rules of larger than 1000 ms pings multiplayer online racing: 1. You can collide with the AI cars in your own simulation. 2. The AI cars in your simulation are trying to have the same position as the real player cars. 3. The first real player who crosses the finish line wins. 4. You get your position in relation to the real player in front of you as a time (ex. +0.543 Ian Lake), upto three times per lap (split times) depending on the track length. Note, this is the time to the real player, not the AI representing him, so you could be in front of the AI and still get a positive time. 5. You get your position in relation to the real player behind you as a time (ex: -1.367 Byron Forbes), upto three times per lap (split times) depending on the track length. Note, same as the note in 4. 6. An AI behind you will not try to pass you until the real player (represented by the AI) is at least 0.5 seconds (probably not the best choice of time, perhaps a full second is better?) in front of you. 7. The AI to AI racing in your sim is non existing, since they just try to show where the real players are. 8. The AI follow some prerecorded driving line from each player. This could be either the qualifying lap or a couple of offline race laps (prefered). The AI try to keep up with each real player by following the prerecorded driving lines. This is done without warping or other unnatural driving from the AI. 9. Bumping into an AI wont hurt the AI, only you will be hurt, thus the real player represented by the AI will not be hurt either. 10. The AI in your sim only change place with each other when the whole car of the real player is in front of the whole car of the other real player. This will be done as a slipstreaming or an outbraking manouver. 11. When an AI has to pass a real player to keep up with the real player it is representing. It either do it slip streaming, if possible without adding unrealistic acceleration or top speed, or by putting itself side by side to the real player just before braking into the corner and then perfectly outbrake the real player. In the later case an unrealistic increase of top speed is allowed. 12. The sim will only allow your participation in the online race if you use exactly the same setup for the race as in the provided offline laps??? In qual you may use a different setup. 13. The simulator corrects the driving line in the offline laps, if they are to much off from a normal line. For example zickzacking on the straights and blocking while entering a corner will be corrected. 14. When a real player enters the pits the AI car representing the real player in all the other players simulation will be moved from the track straight into the pit, perhaps by dissolving on the track and reappearing in the pit. 15. An alternative to 14. When you get called into the pit you have to deny it before the second split time or you will be forced to pit. This is so the AI representing you in the other simulations will have time to do the same thing as you do. This rule could perhaps be combined with rule 14, so if you have some accident before the pit entry you could go straight into the pit and rule 14 will apply. This suggestion doesn't need a server. All players can connect to all others via for example Kali. Kali will only be used for the simulator to get the ip-addresses of all the other players, after that the simulators wont need Kali. A 28.8k modem should have no trouble sending 29 ip- packets and receiving 29 ip-packets, all within one second, which is all that is needed. Two or three seconds isn't actually a problem either :o). Getting the position of another real player every third second isn't a problem, since all the AI does is to increase or decrease it's pace to be approximately where the real player are. What do you all think, could this happen in the near future of the sport we all like to call Simulator Racing :o)? /Christer, a 34 year old software engineering consultant, and one serious sim racer :o)))

>What do you all think, could this happen in the near future of the sport >we all like to call Simulator Racing :o)? No, because it is too far removed from real *racing*. A lot of the skill, and IMHO most of the enjoyment of sim racing, comes from the passing and defending manoeuvers. Unless this is done in real time with the guy in front knowing that he has to defend his line and the guy behind reacting, in real time, to those movements, you're not really racing. Sorry, Christer, but you're really barking up the wrong tree with this idea. You would do better to devote your time to finding someone in the US to get you a TEN account. Oh, and then you'll realise that the only reason why the Australians are so fast in GPL is that everyone else is racing on TEN.... <g> btw - on TEN I've raced a guy from Korea and it was mostly very good with very little warping, perfectly capable of close drafting at Dega and going round the turns door to door (*much* better than anything I ever managed with F1RS via Kali (or even on a LAN for that matter!), so don't let your experiences with Kali/F1RS cloud your judgement). Australia isn't really that much further away than Korea. Within the time it would take to develop your idea and iron out all the bugs, the technology may well be there to allow global racing in real time <g> Cheers, Richard -- We all bump into each other every day of our lives, and we render our opinions whether we know anything or not, and if anybody catches us out we lie...

Thanks a lot for following up my post, Richard :o), I was afraid I would get no response at all. Now read my follow up below. says...

Quoted text (14 lines)

> >What do you all think, could this happen in the near future of the sport > >we all like to call Simulator Racing :o)? > No, because it is too far removed from real *racing*. A lot of the skill, > and IMHO most of the enjoyment of sim racing, comes from the passing and > defending manoeuvers. Unless this is done in real time with the guy in > front knowing that he has to defend his line and the guy behind reacting, > in real time, to those movements, you're not really racing. > Sorry, Christer, but you're really barking up the wrong tree with this > idea. You would do better to devote your time to finding someone in the US > to get you a TEN account. Oh, and then you'll realise that the only reason > why the Australians are so fast in GPL is that everyone else is racing on > TEN.... <g>

I guess they all beat you silly then, since you are as slow as we are :o). I guess the TEN drivers would have no trouble beating everyone in the GPL demo either :o)? I bet they cant, thus challenging them :o). Even if we could communicate at the speed of light it would still take 125 ms for a signal to travel to the other end of our globe. 125 ms is 10 meters at 300 kph, or two car lengths. There is no technology today which makes it possible to communicate with speeds greater than the speed of light. I'm hoping for such a technology, but if we are unlucky there might never be one :o(((. I'm guessing TEN already use a similar solution which we are proposing, which would actually mean that TEN are fooling you into believing that you're racing in real time <g>. Sounds horrible doesn't it :o). Why not see beyond this and see the increadible possibilities which lies within exchanging ideas? I really dont think Einstein could of come up with the relativity theory if he didn't have anyone to exchange ideas with at all. I know he had some really great conversations with Bohr, for example. I also bet he used many of the ideas currently available at that time. I've have had some really great conversations with Byron and Ian, but among us it's probably Ian who's the Einstein :o))). Byron would be Bohr, I guess, and I'm just the publisher :o). I've had some great conversations with you to, Richard, and I've certainly used the result of our conversations to change my ideas. Hope you dont mind :o). I of course cant judge if the ideas have improved or not, I leave that to the market :o). /Christer, use others to improve ideas, so I can get more fun when driving a racing simulator :o)))

It will be just as racing the AI's offline, but they will race at the pace of the real drivers participating in the race. The AI will work as a filter, though, so taking out other drivers will be impossible :o). If you crash it's all your own fault :o). If you disconnect you're a DNF :o). /Christer remove "has.no.life." in email, cause it's the SPAM filter

This is a very interesting idea. I'ld love to see it happen.

>I guess they all beat you silly then, since you are as slow as we are >:o). I guess the TEN drivers would have no trouble beating everyone in >the GPL demo either :o)? I bet they cant, thus challenging them :o). Most couldn't, some can. I got down to 1:07.3 before getting bored with just driving round; it's a great driving model but.... <g> Yes, N2 on TEN does use code to 'predict' where the other car is, but this is significantly different to what you are proposing. Remember that whilst a car may be moving 10 metres at 300kph, it is the *change* in the speed that is important. If it is continuing in a straight line at constant velocity then a time delay doesn't matter at all. So it is the amount of change in direction and speed that can happen within the latency period that is important. Granted that this is rather less with a NASCAR than a F1 car but by and large it works very well using the TEN mechanism. Some form of prediction will always be required, but it needs to be based on where the cars *actually* are, not on some pre-recorded laps! Take the following point for instance: OK, so I drive soem offline laps being careful to take a tight line into each and every corner (not blocking, just being careful). If you then use those laps to base the prediciton stuff on, no-one will *ever* pass my car since with an accurate F1 type sim it simply isn't possible (or at least shouldn't be <g>) to drive round the outside of a car through a corner (anyone for a Brabbham fan car? <g>). Alternatively, what if I always took a wide line into a corner because that was the fastest line. Well, it would be very easy to stick the nose of the car inside and make the pass if 'my' AI always took that line, but if we were actually *racing* then I would have defended the line. Take a F1RS ghost lap and try to pass the ghost without ever overlapping with it? If you manage it, now do so 50 times, if you ever overlap, it's equivalent to crashing out. In what way does this really differ from what you are proposing? Too right he did <g> Einstein is very much over-rated. He had the genius to take some already published stuff and make the connection to produce the Special Theory, but that was very much his high spot. The General Theory proved to be more or less correct but it took better mathematicians than him to really work it out, and let's not forget that he point blank refused to accept many of the implications of his 'own' theories later in his life. I'm not against the exchange of ideas, but I really think it's a shame to waste so much energy trying to come up with a solution that is so diluting the whole point of online racing. It's OK to admit that some things are technically impossible or simply not worth it - and concentrate on improving those things that are! Cheers, Richard -- We all bump into each other every day of our lives, and we render our opinions whether we know anything or not, and if anybody catches us out we lie...

Ahh, nice to see someone who sees the possibilities instead of the limitations :o) /Christer says...

says... > >I guess they all beat you silly then, since you are as slow as we are > >:o). I guess the TEN drivers would have no trouble beating everyone in > >the GPL demo either :o)? I bet they cant, thus challenging them :o). > Most couldn't, some can. I got down to 1:07.3 before getting bored with > just driving round; it's a great driving model but.... <g> It will be real nice to see this driving model in both N3 and ICR3 :o))). So, if I slammed the brake down at 300 kph, then your simulator gets that information 125 ms later, wouldn't this mean that the car representing me in your sim would brake 10 meters later, or two car lengths??? We haven't worked out all the details and dont think we have to either, but to answer this one: Both you and I know that you will not drive a perfect race :o). So when you miss out on a corner, taking it a bit slower than you usually do, then the AI car representing you in all the other players sim have to slow down to match it's pace to yours. This could be done in several ways and without warping. Remember the AI representing you will try to be where you are on the track. One solution could be that the AI is a bit slow out of the corner in which you were slow entering, then the real player behind your AI can get a chance to slip-stream passed you. There could be other solutions to this, but I guess you see my point. Again, we have not worked out the details. For example the AI could be as smart as it defends the racing line itself in relation to how fast your offline race laps were. The faster your offline laps are the better your AI defends it's line. In the way that you, in our suggestion, race the AI's as you would in an offline race, but the AI's pace are about the same as the real drivers around the world also participating in the race. If you are on the inside of an AI into the corner it will of course not drive into you, it will do what a normal AI would do in an offline race. Einstein was only an example. I believe all great thinkers have been inspired and influenced by other thinkers, good or bad. IMO, if you spend your time solving problems, even if the problems are already solved or you come to the wrong conlusions, then you have not wasted your time at all. I believe that any time spent solving problems is good for once brain. I could be wrong, though, but the latest research on the brain point in that direction. Yes, I know, Richard, you think I should spend more time solving problems :o), and I will, trust me :o). BTW, the solution we propose, could very much, in some degree, already be used at NROS. If they for example changed the warping to be a change of pace (increase or decrease) of the AI to match the real players real track position, then it would look a bit more like the idea we propose. I use my right to disagree :o). IMO, it's not OK to admit that some things are technically impossible or not worth it, unless one are really convinced, perhaps not even then :o). I guess we have stumbled on the classic halt-problem - When do we stop trying to solve a problem? /Christer, we really dont have a clue, do we :o)?

Quoted text (7 lines)

> At the end of the day, as long as I have access to a low enough latency > environment to race I will do so in preference to an 'AI' solution. I > already have that environment and it can only get better as the technology > improves, hence why I don't see a great future in your proposal for myself > at least. But if you really want to race those Australians atm then your > solution is worth pursuing. Now all you've got to do is persuade a software > company that it is worthwhile.... ;-)

Let me (Bohr??) enter here. LOL. In increasing order of goodness we have the following; 1/ Offline racing 2/ Ghosts 3/ Online racing where the packets influence the AI ie the AI works in unison with the incoming packets 4/ Online racing (very low pings required) Now, to me, 1/ is getting boring, 2/ doesn't interest me too much and 4/ is out of the question globally as we speak. And 3/ is a stepping stone to a truely global 4/ ! My ideas reguarding 3/ is that, simply, the AI in a sim is able to speed up/slow down according to what info it gets from the incoming packets from other drivers. This is obviously not as good as 4/ but is far better than 1/ and 2/. Though we would only be interacting with the AI when in close proximity to other cars we still have all the pressures of real time racing. The most important thing is to know what your position is relative to the other drivers and to thus know how to pace yourself. To me, this is a no brainer for any company doing a racing sim for the time being and probably for some time to come. Christer mentioned each driver submitting laps to dictate the AI. What my ideas on this are is that simply, when packets get delayed, that the AI's speed is dictated by that drivers qualifying speed. Nothing to complex. Also, as you can see I have 4/ as being better than 3/?????? Is it really? 3/ gives full international racing whereas 4/ gives only local racing. It's a tossup as to which is better. Imagine when ICR3 comes out what it would be like to be sitting there waiting to race against 30+ other drivers from all around the world! There are a few issus to be ironed out like pits and knowing your exact position/time relative to others (especially on the last lap) but nothing that hard to solve. Pits - a driver must hit a key to let the AI representing him on the other computers that he is pitting on this lap. Much like GP2 is already. Last lap - like ICR2's F2 screen you could get a readout of every other drivers relativity (there's that relativity stuff again - spooky) to each other. This on its own would be great even!

Imagining, imagining, imagining... There it is... My heart rate just went up :o))))))))). /Christer, just got a smile all over my face :o)

Quoted text (12 lines)

> > I think you are vastly underestimating the amount of effort required to > > implement this. Just a little example - what happens at the start of the > > race? With large latencies, even the simple synchronisation of the starts > > is likely to be difficult. Now take into account the fact that in the rush > > for the first corner there's going to be a lot of passing and by the time > > half a lap is complete the chances are that each of the drivers' PCs is > > going to have a completely different view of what happened at the start! > > Working out algorithms to sort this out is not a 'no brainer', indeed far > > from it. > So hard?? Why? How is it different from the AI now? All that needs to be done is to make > cars faster/slower as needed to match incoming packets. Where's all the fuss?

Exactly, this is actually a "no brainer" when you realize how easy it is. Say you're third on the grid and make a really lousy start, then the AI (representing the other driver) will come woshing passed you into the first corner. All AI perhaps wont pass you even if their matching real drivers are in front of you in real time, but these AI will soon get real eager to pass you, if you cant keep up with the pace of the AI's real driver. If for example another driver has done a lousy start and is in real time behind you, but he's AI is in front of you. Then he's AI will decrease it's pace and you will catch up very soon and be able to pass. The AI will never race each other, they will just keep the order in which the real players actually have in real time. This could be dramatized nicely by the simulator, with for example slipstreaming, outbraking manouvers, fighting through corners, so it would actually seem as if the AI's are fighting each other, which they are of course not :o). There will be no warping :o))). Seriously, I would prefer the AI behind me in this situation to just be glued to my behind and then either slipstream me or outbrake me. If I'm all over the track trying to block it I risk taking myself out and loosing a lot of time, and time is important, since the driver with the best race time wins the race. This was actually the only part in this post where I didn't agree with Byron (Bohr) :o), which is a first I believe :o). It was actually Byron that was the driving force behind this AI proposal. I was satisfied with the ghost solution and told Byron repeatadly that a solution with collision detection on would be to hard to implement. I dont believe that anymore :o), thanks Byron :o). /Christer

Quoted text (9 lines)

> I think you are vastly underestimating the amount of effort required to > implement this. Just a little example - what happens at the start of the > race? With large latencies, even the simple synchronisation of the starts > is likely to be difficult. Now take into account the fact that in the rush > for the first corner there's going to be a lot of passing and by the time > half a lap is complete the chances are that each of the drivers' PCs is > going to have a completely different view of what happened at the start! > Working out algorithms to sort this out is not a 'no brainer', indeed far > from it.

So hard?? Why? How is it different from the AI now? All that needs to be done is to make cars faster/slower as needed to match incoming packets. Where's all the fuss? I think you confused 3/ with 4/ here but anyway. My point is that, though with 4/ you get the real deal, with 3/ you get a good deal and can race everyone. Right in front of anyone who can see it's potential. Like to finance me? :)) Also, this type of system could be designed to go to real time if pings allowed. There are all types of possibilities here. AFAIK there is no guarantee at all for what you say about ICR3 and real time racing. Have you developed some new technology? Either you did not think this out or did not understand what I meant. Lets say we have a 50 lap race. On the completion of lap 49 my modem sends my "actual" elapsed time to all others and vice versa. Of course, you would already have a good idea of how hard you needed to push at about the 3 or so laps to go mark.So on the last lap (and all others) you know your exact position relative to those closest to you for position. The winner is simply the player with the least elapsed time from Start to Finish as in offline racing for example. On earlier laps this might also let you know if your holding up the car behind you that may in fact be just ahead of you. For example, in the case you are holding up a car that is actually ahead of you, if the gap gets to 3+ seconds say, then that car could just warp to its real location since it would be at a safe distance. Once again, where's the big deal?

Quoted text (8 lines)

> > if the gap gets to 3+ seconds say, then that car could just warp to its real location since it > > would be at a safe distance. Once again, where's the big deal? > There will be no warping :o))). Seriously, I would prefer the AI behind > me in this situation to just be glued to my behind and then either > slipstream me or outbrake me. If I'm all over the track trying to block > it I risk taking myself out and loosing a lot of time, and time is > important, since the driver with the best race time wins the race.

Yes. This is a matter of preference I suppose. As you mention, there is a penalty for holding up an AI anyway and the further you fell back behind the actual car the faster (more power, traction, etc) that AI behind you gets.Also, I think it would be good to have a way to know if that AI should actually be in front of you. Perhaps it could flash red with an increasing frequency depending on how far ahead it should be. That would enable you to just pull to one side on a straight and allow the nusance to pass :) Would probably be best dealt with by having it as an option if this type of AI ever happens. Which, of course, it will. Now it's just a matter of the software houses seeing the light :))

I dont think you have to know if the AI behind you should be in front of you or not, just race as you would a normal race, sooner or later that AI will start attacking you, or perhaps you will start to pull away :o). Some more rules after emailing with Byron, Ian and Richard :o). 15. If a real driver makes a mistake then his AI makes a similar mistake at a similar track section a bit later. Say you make a mistake entering turn 1, then the AI representing you makes a similar mistake entering turn 2 or 3, depending on current pings. Remember we have delays upto one or two seconds here :o). 16. If a real driver makes a mistake, so it's very easy for the AI behind him to pass, then they will pass, no matter where the AI's real drivers are. For example at the start, or a real driver going off the track. These 16 rules would of course work much better in a server solution than if we would use no server, but then I guess we would have to pay to the company maintaining the server. Say around 10 USD a month, would be quite acceptable to me :o). /Christer, wonder if we have any readers at all left in this thread now :o)