Tactics2D: A Reinforcement Learning Environment Library with Generative Scenarios for Driving Decision-making
Tactics2D is an open-source Reinforcement Learning environment library featured with auto-generation of diverse and challenging traffic scenarios. Its primary goal is to provide an out-of-the-box toolkit for researchers to explore learning-based driving decision-making models. This library implements both rule-based and data-driven approaches to generate interactive traffic scenarios. Noteworthy features of Tactics2D include expansive compatibility with real-world log and data formats, customizable traffic scenario components, and rich built-in functional templates. Developed with user-friendliness in mind, Tactics2D offers detailed documentation and an interactive online tutorial. The software maintains robust reliability, with over 90% code passing unit testing. For access to the source code and participation in discussions, visit the official GitHub page for Tactcis2D at https://github.com/WoodOxen/Tactics2D.
PDF Abstract
nuScenes
Argoverse 2
highD Dataset
inD Dataset
