Struct vrp_core::algorithms::mdp::Simulator

pub struct Simulator<S: State> { /* fields omitted */ }

A simulator to train agent with multiple episodes.

Implementations

impl<S: State> Simulator<S>

pub fn new(
    learning_strategy: Box<dyn LearningStrategy<S> + Send + Sync>,
    policy_strategy: Box<dyn PolicyStrategy<S> + Send + Sync>
) -> Self

Creates a new instance of MDP simulator.

pub fn get_optimal_policy(
    &self,
    state: &S
) -> Option<(<S as State>::Action, f64)>

Return a learned optimal policy for given state.

pub fn get_state_estimates(&self) -> &StateEstimates<S>

Gets state estimates.

pub fn set_action_estimates(&mut self, state: S, estimates: ActionEstimates<S>)

Sets action estimates for given state.

pub fn run_episodes<A>(
    &mut self,
    agents: Vec<Box<A>>,
    parallelism: Parallelism,
    reducer: impl Fn(&S, &[f64]) -> f64
) -> Vec<Box<A>> where
    A: Agent<S> + Send + Sync, 

Runs single episode for each of the given agents in parallel.