Software

functional-wise-move [code]
wise-move designed with a different philosophy. Allows for more control over environment design, including canvas elements, default policy specification, and a functional reward specification.

wise-move [code]
Hierarchical framework to investigate safe reinforcement learning, using incremental learntime verification of temporal logic constraints. Provides a gym environment and learning backends including MCTS.

GridDriving [code]
Extension of the original CarRacing-v0 environment to have a configurable city-like road network, with traffic lights and multiple vehicles, each of which can be controlled.