Struct Experience 

pub struct Experience<O, A> {
    pub observation: O,
    pub action: A,
    pub reward: f64,
    pub next_observation: O,
    pub status: EpisodeStatus,
}

A single transition: (s, a, r, s', status).

The fundamental unit of experience stored in replay buffers and used for agent updates. Corresponds to one (s, a, r, s’, done) tuple in classical RL literature, but with a richer status field that distinguishes natural termination from truncation.

Fields

observation: O

The observation at the start of this transition.

action: A

The action taken.

reward: f64

The scalar reward received.

next_observation: O

The observation after taking the action.

status: EpisodeStatus

Whether the episode ended and why.

Algorithms that bootstrap value estimates (DQN, PPO, SAC) must inspect this to handle terminal states correctly:

• Terminated: bootstrap with zero value
• Truncated: bootstrap with V(next_observation)
• Continuing: bootstrap with V(next_observation)

Implementations

impl<O, A> Experience<O, A>

pub fn new( observation: O, action: A, reward: f64, next_observation: O, status: EpisodeStatus, ) -> Self

pub fn is_done(&self) -> bool

Returns true if this transition ends an episode.

pub fn bootstrap_mask(&self) -> f64

Returns the bootstrap mask: 1.0 if the episode continues or was truncated (i.e. the next state has non-zero value), 0.0 if terminated.

Multiply value estimates by this mask when computing TD targets: target = reward + gamma * bootstrap_mask() * V(next_obs)

pub fn map_obs<O2>(self, f: impl Fn(O) -> O2) -> Experience<O2, A>

Map the observation to a different type.

Useful for observation-wrapping layers that preprocess before storage.

pub fn map_action<A2>(self, f: impl Fn(A) -> A2) -> Experience<O, A2>

Map the action to a different type.

Trait Implementations

impl<O: Clone, A: Clone> Clone for Experience<O, A>

fn clone(&self) -> Experience<O, A>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<O: Debug, A: Debug> Debug for Experience<O, A>

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

Formats the value using the given formatter.