Struct Experience 

pub struct Experience {
    pub agent_id: AgentId,
    pub observation: Vec<f32>,
    pub action: Vec<i64>,
    pub reward: f32,
    pub next_observation: Vec<f32>,
    pub terminated: bool,
    pub truncated: bool,
    pub value: f32,
    pub log_prob: f32,
}

Experience tuple sent from a simulator to a learner.

All tensor payloads from the pre-Burn (tch-coupled) implementation are replaced with plain Vec<f32> host buffers. Construct Burn tensors at the learner-side rollout buffer with Tensor::from_floats(...) once the buffer is full.

Fields

agent_id: AgentId

Agent that generated this experience.

observation: Vec<f32>

Observation vector [obs_dim].

action: Vec<i64>

Action taken. Length 1 for scalar discrete; num_dims for multi-discrete (matches crate::multi_agent::environment::MultiAgentEnvironment::agent_action_space).

reward: f32

Reward received.

next_observation: Vec<f32>

Next observation vector [obs_dim].

terminated: bool

Whether episode terminated (natural episode end).

truncated: bool

Whether episode was truncated (time limit / external reset).

value: f32

Value estimate at this state (from the rollout-time policy).

log_prob: f32

Log probability of the action taken under the rollout-time policy.

Implementations

impl Experience

pub fn new( agent_id: AgentId, observation: Vec<f32>, action: Vec<i64>, reward: f32, next_observation: Vec<f32>, terminated: bool, truncated: bool, value: f32, log_prob: f32, ) -> Self

Create a new experience tuple.

pub fn is_done(&self) -> bool

Check if this experience marks the end of an episode (terminated or truncated).

Trait Implementations

impl Clone for Experience

fn clone(&self) -> Experience

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Experience

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

Formats the value using the given formatter.