Welcome illustration: © 2003 Jean-Christophe Durieu.
Watching a race is fun. Watching a driver you have programmed take its first corner is something else! This tutorial helps you build your own TORCS robot, understand its decisions, and improve it by experimenting in the simulator.
A robot is a program that drives a car. You will write it in C++ and build it as a plugin, a small program that TORCS can load. TORCS calls functions in your robot, supplies information about the car and the race, and reads back your steering, accelerator, brake, and gear commands. The simulator handles the vehicle physics; your robot decides how to drive.
There is plenty to explore. A controller that follows a straight may struggle in a corner. Driving quickly introduces braking and traction problems. Add opponents, fuel, damage, and pit stops, and even a simple racing environment becomes an interesting programming playground.
The setup and first chapter take you from a stationary car to one that follows the track's centerline. You will learn where the driving information comes from, how a steering decision is calculated, and how to build and run your changes. The example intentionally starts small: first gear, modest throttle, and a simple steering correction.
The driving sequence takes you through recovery, speed control, steering, sharing the track, and pit stops. Custom car setups, textures, and implementing multiple robots in one module belong to the optional appendices.
You need:
Start with Getting Ready and choose your platform. It takes you through building a stationary robot and selecting it in Practice. Then follow Chapter 1 to make it drive. Keep your platform's rebuild instructions handy: you will reuse that short step after each experiment.
Read the explanation, try the supplied example (a checkpoint), and watch what the car does. Change one thing at a time. Predict the effect, then compare your prediction with what you observe. If a result surprises you, sketch the car and track on paper. Understanding an unsuccessful experiment is part of the exercise.
Each checkpoint has a complete source example. You can type the driving code yourself or use the supplied version to recover when you get lost. The simple algorithms are teaching examples; their limitations provide the motivation for the next improvements.
A coding agent can help explain an expression, investigate a build failure, or implement a small experiment. Give it the current tutorial and TORCS interfaces, ask for one change at a time, and evaluate the resulting behaviour in the simulator. A compiling driver is only the start of that evaluation.
The source examples and illustrations are included with the tutorial, so it can be read and used offline.
Stuck, or found something confusing?
Ask in the TORCS Discord or on the TORCS-users mailing list. For the mailing list, subscribe before posting.
Tell us which step you reached, whether you use Windows or Linux, and what happened. If an error appeared, include its text. You do not need to know what caused the problem to ask for help.
Teaching with this tutorial?
We would love to hear which explanations worked, where students got stuck, and what would make the tutorial easier to use in class.
Tried an interesting experiment?
Share what you changed and how your robot behaved!
1.14.0 