SimulatedEnvMixin

class maze.core.env.simulated_env_mixin.SimulatedEnvMixin

Environment interface for simulated environments.

The main addition to StructuredEnv is the clone method, which resets the simulation to the given env state. This interface is used by Monte Carlo Tree Search.

abstract clone_from(maze_state: Any) → None

Clone an environment by resetting the simulation to its current state.

step_without_observation(action: Dict[str, Union[int, numpy.ndarray]]) → Tuple[Any, bool, Dict[Any, Any]]

Environment step function that does not return any observation.

This method can be significantly faster than the full step function in cases with expensive state to observation mappings.

Parameters

action – the selected action to take.

Returns

reward, done, info