The race line

[Combustion]

I'm thinking of making a robot for my 3rd year project, it would be a small cpu with various sensors mounted on an electric rc car. The idea is that it would (via measuring distances) figure out the perfect race line, eg after a few runs of the track, *learn* the turns and figure out the best line. It would also do a few pre race tests with the car, eg find out how many Gs the tyres can handle and what braking distances are etc, so it will get the (theoretically) best time possible for a track.

Maybe in the future it could figure the race line *before* going round the turn, ie in real time!

I am now trying to figure out just HOW the hell do you work out the race line? Taking the simplest turn, a 90 degree box turn:

_______________
|
|       _____
|      |
|      |

I know that the best line is a circle where the center point is when the radius is equal to the distance from the apex is equal to the distance to the outside lane (point of turning).

I've got a few ideas how to work out when the car is at the point where the turn should take place, I'm asking if anyone has done this sort of calculation before?

 

[Ev0]

Well, your project sounds very ambitious, but it would be very difficult to get working properly. I remember recently there was a robot car race in the desert where normal SUVs were fitted with equipment to navigate across a small stretch of desert on their own (no human intervention). I think every single competitor failed to get to the finish line. But, a paved track has much fewer variables than a desert, so the programming side would probably be easier.

And as for the racing line for the corner depicted, I have no idea about how to carry out the calculations for that, but try checking out some books or websites on racing theory, they might have the mathematical calculations if the book is not entirely beginner oriented.

 

[JNasty4G63]

There was an article in Racecar Engineering a couple months back discussing this very issue. They were noting that with more and more modern race machines, the "perfect line" through a corner is changing from less of a smooth arc to more of a squared off 3 part turn. With more and more trail braking taking place because of the greater stability of todays machines, cars can actually square off the turn a bit more and get through quicker. Just a thought. Worth reading up on it since this sounds like a fun project.

Hilg

 

[retsmah]

I remember recently there was a robot car race in the desert where normal SUVs were fitted with equipment to navigate across a small stretch of desert on their own (no human intervention). I think every single competitor failed to get to the finish line. But, a paved track has much fewer variables than a desert, so the programming side would probably be easier.

It sounds like you are talking about the DARPA grand challenge, some of them used SUVs although you can use whatever type of vehicle you like. The vehicles are given coordinates that they have to navigate to on their own. It's not a small stretch though, the point is 175 miles across the desert, and they are given ten hours. Traveling across a desert filled with obstacles is tough, doing it at an average speed of 17mph is even harder.

The general theory behind the racing line stuff is making the turn as large of a radius as you can. So you just fit as big of a circle as you can inside the track. Of course, there are a lot of corners that won't work for, and I'm sure you could come up with plenty of situations where that wouldn't work best... but it shouldn't work too poorly either.

Overall I think it sounds pretty tough

 

[Combustion]

Overall I think it sounds pretty tough

I know it's going to be tough, but thats otherwise whats the point of doing something thats easy

Maybe I can do a simulator on the computer before using actual hardware, but hardware would be much cooler.

Anyone have a link to a theory of racing lines? Also maybe I will implement drifting techniques into version 2.0 of the program =p