FAQ

What’s the difference between the Atari environments in OpenAI Gym / Gymnasium and ALE?

The environments provided in Gym are built on the ALE. It just provides a different API to the ALE. As of ALE 0.7 OpenAI Gym now uses ale-py and so there’s no difference between using ale-py and Gym.

What is Xitari? Should I use it?

Xitari is a fork of the ALE around version 0.4. It added support for more games and some other minor changes. This work has now been upstreamed to the ALE and it’s recommended you use the ALE directly.

I want to be able to extract from the game the number of lives my agent still has. How can I do it?

Previous versions of ALE did not support this. We started to support such feature since version 0.5.0, through the use of the function lives(). We strongly encourage RL researchers not utilize the loss of life signal for episode termination, and if you do to clearly report the setting used.

When extracting the screen I realized that I never see a pixel with an odd value. However, the pixel is represented as a byte. Shouldn’t it be up to 255 with odd an even values?

No, the Atari 2600 console (NTSC format) only supports 128 colours. Therefore, even though the colours are represented in a byte, in fact only the seven most significant bits are used. Consequently you have to right-shift the colour byte by 1 bit to produce consecutively numbered colour values.

Can I extract other state information such as the x,y position of sprites?

No, but there is a project which allows you to do just that: AtariARI.