Struct TMaze 

pub struct TMaze { /* private fields */ }

T-maze — the provably memory-hard POMDP benchmark (Bakker 2001).

See the module docs for the full task, observation, reward, and determinism specification.

Implementations

impl TMaze

pub fn new() -> Self

Create a T-maze with the default corridor length (DEFAULT_CORRIDOR_LENGTH = 10) and a seeded cue RNG for a reproducible cue sequence.

pub fn with_corridor_length(corridor_length: usize) -> Self

Create a T-maze with a custom corridor length and the default seed.

Panics

Panics if corridor_length is zero. A zero-length corridor would place the cue and the junction at the same step, leaking the cue into the junction observation and destroying the memory task.

pub fn with_seed(seed: u64) -> Self

Create a T-maze with the default corridor length and a custom seed.

pub fn with_seed_and_corridor_length(seed: u64, corridor_length: usize) -> Self

Create a T-maze with a custom seed and corridor length.

Two instances built with the same seed and corridor_length draw the same cue on every reset, independent of the actions taken.

Panics

Panics if corridor_length is zero (see TMaze::with_corridor_length).

pub fn corridor_length(&self) -> usize

The corridor length N (the junction is at position N).

pub fn cue(&self) -> Cue

The cue for the current episode (primarily for diagnostics and tests).

pub fn position(&self) -> usize

The current position along the corridor in 0..=N.

pub fn at_junction(&self) -> bool

Whether the agent is currently at the junction (position == N).

Trait Implementations

impl Debug for TMaze

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for TMaze

fn default() -> Self

Returns the “default value” for a type.

impl Environment for TMaze

type Action = i64

Discrete junction action. 0 = up, 1 = down; only consequential at the junction (ignored, as a no-op, while in the corridor).

type State = TMazeState

Snapshot type capturing position, cue, and the cue RNG. See TMazeState.

fn reset(&mut self)

Reset the environment and return initial observation

fn get_observation(&self) -> Vec<f32>

Get the current observation

fn step(&mut self, action: i64) -> StepResult

Step the environment with an action

fn observation_space(&self) -> SpaceInfo

Get the observation space dimensions

fn action_space(&self) -> SpaceInfo

Get the action space dimensions

fn render(&self) -> Vec<u8>

Render the current environment state

fn close(&mut self)

Close the environment

fn clone_state(&self) -> TMazeState

Snapshot the current env state for later restoration.

fn restore_state(&mut self, state: &TMazeState)

Restore the env to a previously-snapshotted state.