Skip to main content

PlannerAgent

infotheory::aixi::planner_agent

Trait PlannerAgent 

Source 
pub trait PlannerAgent {
    // Required method
    fn run_cycle(
        &mut self,
        phase: PlannerPhase,
        step: usize,
        schedule: &PlannerSchedule,
        env: &mut PlannerEnvironment,
        observer: &mut dyn PlannerCycleObserver,
    ) -> Result<PlannerCycleOutcome, PlannerAgentError>;

    // Provided methods
    fn reseed_for_episode(&mut self, _random_seed: u64) { ... }
    fn reset_for_episode(&mut self, random_seed: u64) { ... }
}
Expand description

Executable planner-agent abstraction.

Required Methods§

Source

fn run_cycle( &mut self, phase: PlannerPhase, step: usize, schedule: &PlannerSchedule, env: &mut PlannerEnvironment, observer: &mut dyn PlannerCycleObserver, ) -> Result<PlannerCycleOutcome, PlannerAgentError>

Execute one planner-environment cycle.

Provided Methods§

Source

fn reseed_for_episode(&mut self, _random_seed: u64)

Reseed controller-side stochastic state.

This does not clear learned model state or retained history.

Source

fn reset_for_episode(&mut self, random_seed: u64)

Start a fresh environment episode while preserving learned model state.

Implementations should reset episode-local transient state such as a previous-action pointer or retained search tree. They should not discard learned predictor state unless the concrete controller documents that policy separately.

Implementors§