The online racing simulator
AI in/out sim like interface
(9 posts, started )
#1 - Woz
AI in/out sim like interface
It would be good if there was an in/out sim like interface for the AI. This would allow 3rd party development of AI drivers. It would require some form of position information (already in outsim) along with track details. Finally the ability to send control requests to the AI car.

I think this would open the door to some great AI for LFS and even make things like AI leagues possible to find the best AI etc.

It might even make sense to include the current AI codebase, tweaked to use the interface, so people can see how it works and where the weak points are.
Now that definitely sounds interesting
+1

Nice idea, indeed - I think most of the bits are there, no?
It seems like it would "only" need the control input.
Wow nice idea - i see many new users coming to LFS to take part in a world wide AI competition. Would be much better then these robot-footballmatches

I hope this will be implemented sometimes

Greets,
Warper
#5 - Vain
Already proposed and most propably ignored by devs.
Great potential anyway.

Vain
Quote from Brethon :It's not very viable for a company to start outsourcing things for free. Can you imagine being let go at your job, because someone is willing to do it for free in their spare time? And then the whole mechanics of the company's financial system is shot to heck, and it goes from there.

No one mentioned outsourcing. You're barking up the wrong tree.
#7 - Woz
Quote from Brethon :It's not very viable for a company to start outsourcing things for free. Can you imagine being let go at your job, because someone is willing to do it for free in their spare time? And then the whole mechanics of the company's financial system is shot to heck, and it goes from there.

This has nothing to do with outsourcing. As you are new to LFS you might not understand the dynamics of the dev team.

There are 3 people in the LFS team and only one of those is a coder.

There are already a number of is/out sim interfaces that allow control over aspects of LFS. These have been used in the past to implemeent fuel etc before they came into LFS.

This idea is about modding LFS, something that many games companies allow. iD Games only got to the level they have reached today because of 3rd parties creating content and mods for their games and giving those to the community for free.

Allowing 3rd party developers to create AI for LFS will have NO impact on the dev team jobs because you still need LFS to use the mods.
#8 - bozo
There are existing racing games that allow for the development of AI car control. RARS has been around for many years, enabling some good academic research, see http://rars.sourceforge.net/

TORCS has now replaced RARS, see http://torcs.sourceforge.net/ - not sure how popular this is though; some of the more recent test events don't seem to have attracted any participants!

If you start digging around the research in this area, you'll find that developing AI control of a racing car is no simple matter. Apart from the constraints of a real-time environment, there are huge challenges working out a fast racing line, compounded by complications like overtaking and defending your position on track.

However, when you look at events like Robocup you'll see how collaboration between the teams (code is made freely available after each competition) has aided the development of multi-agent systems. As an existing game with a good following, I'm sure if a car-control interface was provided for LFS, there would be a small number of researchers interested in developing their own AI drivers.

Thinking slightly bigger, Robocup succeeds because it has a grand aim ... developing a humanoid football team that can beat the 2050 World Cup winning team (Brazil vs the robots!). Perhaps the racing equivalent would be an AI Sim Driver winning a race against F1 drivers by 2050. Given that F1 cars are already largely controlled by electronics rather than mechanical devices, maybe that wouldn't be so hard to imagine ... just would need sufficient resources working on the AI side.

Food for thought.
Bleh, an AI can beat and F1 driver now. Just put a ZX-81 into a Ferrari and the title is Sinclair's.

I was toying with the idea of making a program to race LFS for you, not as a hack, but as a user-programming challenge. The idea being you'd set LFS to keyboard control and write a script that drives the car, i'd just develop a script engine that lets you program the parameters of what to do at what points of the track in what conditions, and instructions on how to interact with other cars, and then you've got a whole new daft racing league.

It's something I might do in the future, but in the meantime I dont think an LFS-AIguage would actually be very much use, for a start there would be some latency between applications and also, as detailed above, truly intelligent AI can be very difficult to code.

When i've written AI in the past, and I havn't done a good AI in years because I tend to focus on multiplayer stuff (the technical aspects scare most programmers, but I think it's easier because then I dont have to make an AI!), I try to make the AI do what I would do - in the case of LFS this would still result in two fundamental probems:

It'd still be slow, and it'd still crash into you !

AI in/out sim like interface
(9 posts, started )
FGED GREDG RDFGDR GSFDG