Čapek

Python library for developing TORCS racing robots.

Getting started

1. Install TORCS.

   The most convenient way is to install TORCS from this repository. The repository contains the most recent version of the simulator with SCR patches that allow communicating with TORCS via UDP.

   $ git clone https://github.com/fmirus/torcs-1.3.7.git
   $ sudo apt-get install libglib2.0-dev libgl1-mesa-dev libglu1-mesa-dev \
        freeglut3-dev libplib-dev libopenal-dev libalut-dev libxi-dev \ 
        libxmu-dev libxrender-dev libxrandr-dev libpng-dev
   $ cd torcs-1.3.7
   $ ./configure
   $ make
   $ sudo make install
   $ sudo make datainstall
   

2. Install Čapek.

   $ pip install capek --user
   

3. Program a robot's logic.

   Reference robots are located in the examples directory.

4. Test the robot.

   1. Start TORCS.
   2. Select Race → Practice → Configure Race → Accept.
   3. Drop scr_server 1 to the Selected menu. This is the robot.
   4. Select Accept → Accept → New Race.
   5. Observe.

5. Fix, if necessary.

Creating robots

To create a robot using Čapek, you have to re-implement the Driver class. The common structure of a robot is demonstrated below.

from capek import Driver, Client

class MyDriver(Driver):
    # ...
    # Initialize necessary classes or variables.
    # ...

    def drive(self, state, control):
        # ...
        # Define behavior on gear change, steering, etc.
        # ...

        if state.speed_X < 300:
            control.accel = 1  
        else:
            control.accel = 0

        # Changes in control are sent to TORCS on the next tick.

        return control

client = Client(verbosity=1)
# Enters loop until reaches maximum learning episodes.
# Or is interrupted.
client.run(driver=MyDriver)

The only methods that are supposed to be re-implemented are drive, on_shutdown and on_restart. The two latter exist because of convention with C++ and Java clients.