A successful lap
I’ve decided to upload the media and my project work on my Racing Car’s AI project, yay! There’s some videos, pictures, the code and the report itself. I’d not expect anyone to get it working on their own systems (the code requires a parallel port controller, and the correct camera view), however.
It might not be true academic AI, but I know it’d fool someone not in the know 🙂 Those rules do kick ass in almost all of my tests 😀 It was really interesting working with a non-virtual problem of AI, since it let me know that one major area non-virtual projects have trouble with is accurate data gathering which is interpreted correctly, which took the major part of my time to implement. At least I wasn’t working in 3D, whew!
Wright AI likes airports, I hope mine will be a bit more varied
I’ve decided to finally get into OpenTTD’s AI, and after some failed attempts got the build available to compile on Windows, since annoyingly no build is done automatically for the win32 platform. In any case, I’ve got the tutorials up and running, and it should be interesting. Since the game itself is passive-competitive, with players facing off against AI’s (with multiplayer available too), and the AI uses the same “interface” for commands, it is quite an equal game. That is, once the game’s inherent AI cheats are removed perhaps 🙂
If I get anything good going, I’ll post more on it. I prefer some personality rather then milking routes with a ton of vehicles doing the same thing. I’ll have to keep up with builds however, since it’s not a complete API yet, sadly, and there will be complexity abound with players and AI’s being able to build across routes, and alter terrain as parts of a route is built.
So, I’ve not done much over the last few weeks, apart from my final year project work. I handed in my report on Tuesday, missing the bank holiday to do more on it 🙂 Not a work of Shakespeare, and frankly I could have done with more proofreading. However, I can’t change it now, I just hope it states what I did well enough. 😉
The project itself is meant to achieve the aim of having a remote control car race around a track (as seen in the picture). I didn’t really do “AI” as such, but instead some simple rules to choose an action every second or so. The limitations of the hardware, it being a 1 core P4 machine with 512MB RAM meant that the 2 years of previously completed projects used enough resources anyway! Adding learning AI would have been tough.
The rules system worked fine however, the illusion of AI is there 😉 It also took a lot longer to bring the parts of the system that were required to gather information – image processing and analysis (to find the car and it’s direction), and route planning using waypoints (to find where to go next) – then it took to write rules to navigate around accurately. It got 10 laps completed easily, so I am pleased enough. 😀
I’ll get back to my radio show for the rest of the semester now, and I might put up a page on my project with the code and so forth. Also, Never Mind The Polygons is on tonight, which I might have had to miss if my project was still being worked on 🙂