Struct SignalingGame 

pub struct SignalingGame { /* private fields */ }

2-agent referential signaling game over a fixed vocabulary.

See the module docs for the full description and the comms layout table.

Implementations

impl SignalingGame

pub const NUM_AGENTS: usize = 2

Number of agents (always 2: speaker + listener).

pub fn new(vocab: usize) -> Self

Construct a signaling game over a vocab-token vocabulary with hidden token 0.

Panics

Panics if vocab == 0 — a signaling game needs at least one token.

pub fn with_hidden(vocab: usize, hidden: i64) -> Self

Construct a signaling game with an explicit hidden token (for tests / deterministic rollouts).

Panics

Panics if vocab == 0 or hidden is outside 0..vocab.

pub fn vocab(&self) -> usize

Vocabulary size V.

pub fn hidden(&self) -> i64

The hidden token the speaker must transmit this episode.

pub fn received(&self) -> i64

The message token the listener received on the most recent step, or -1 if no step has occurred since the last reset.

pub fn set_hidden(&mut self, hidden: i64)

Set the hidden token (0..vocab) and clear the received message.

Panics

Panics if hidden is outside 0..vocab.

Trait Implementations

impl Clone for SignalingGame

fn clone(&self) -> SignalingGame

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl CommunicatingEnvironment for SignalingGame

fn message_vocab(&self, agent_id: usize) -> Vec<usize>

Per-agent message action-space layout (one bin count per message dim), analogous to MultiAgentEnvironment::agent_action_space. These dims are appended after the task action dims in each agent’s action vector. An empty vec means the agent sends nothing.

fn message_obs_size(&self, agent_id: usize) -> usize

Number of observation dims that carry received messages for this agent. Zero means the agent receives nothing. The env is responsible for placing received messages into the observation vector returned by get_agent_observation / step_multi (see place_message).

fn delivery(&self) -> Delivery

Delivery policy for messages emitted this step. Defaults to broadcast, matching the Bucket Brigade signal semantics.

impl Debug for SignalingGame

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

Formats the value using the given formatter.

impl Default for SignalingGame

fn default() -> Self

A 2-token signaling game (the smallest non-trivial vocabulary).

impl Environment for SignalingGame

type State = (i64, i64)

Full snapshot: the env consumes no RNG, so restoring reproduces every subsequent step exactly.

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

Single-agent surface returns the speaker’s observation. Multi-agent consumers should use MultiAgentEnvironment::get_agent_observation.

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

Degenerate single-agent step: interpret action as the speaker’s emitted message and broadcast it. Multi-agent consumers should use MultiAgentEnvironment::step_multi.

type Action = i64

Action type accepted by Environment::step.

fn reset(&mut self)

Reset the environment and return initial observation

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) -> (i64, i64)

Snapshot the current env state for later restoration.

fn restore_state(&mut self, state: &(i64, i64))

Restore the env to a previously-snapshotted state.

impl JointEnv for SignalingGame

Thin adapter so the joint PPO trainer can drive the signaling game directly (issue #275, Phase 2). reset_joint / step_joint delegate to the existing MultiAgentEnvironment surface — all message routing is already handled inside SignalingGame::step_multi, so no extra plumbing lives here.

fn reset_joint(&mut self, _seed: Option<u64>) -> Vec<Vec<f32>>

Reset the env in-place. Returns per-agent observations.

fn step_joint(&mut self, actions: &[Vec<i64>]) -> JointStepResult

Step the env with per-agent actions.

impl MultiAgentEnvironment for SignalingGame

fn num_agents(&self) -> usize

Number of agents in this environment.

fn get_agent_observation(&self, agent_id: usize) -> Vec<f32>

Get observation for a specific agent.

fn agent_action_space(&self, agent_id: usize) -> Vec<usize>

Per-agent action-space layout.

fn step_multi(&mut self, actions: &[Vec<i64>]) -> MultiAgentResult

Step the environment with multiple actions (one per agent).

fn active_agents(&self) -> Vec<bool>

Get which agents are currently active (not terminated).