google-deepmind/android_env
A Python library that exposes Android devices as reinforcement learning environments for training agents to perform tasks via touchscreen interaction.

Velocity · 7d
+0.7
★ / day
Trend
→steady
star history
AndroidEnv provides a flexible platform for defining custom RL tasks on top of the Android operating system. Agents interact with the device through a universal action interface, sending localized touch and lift events, while the system returns pixel observations and rewards as defined by task specifications. It enables research into training RL agents to complete real-world tasks on mobile devices.