Struct rsrl_domains::MountainCar

pub struct MountainCar { /* fields omitted */ }

Classic mountain car testing domain.

This problem involves an under-powered car which must ascend a steep hill. Since gravity is stronger than the car's engine, even at full throttle, the car cannot simply accelerate up the steep slope. The car is situated in a valley and must learn to leverage potential energy by driving up the opposite hill before the car is able to make it to the goal at the top of the rightmost hill.¹

Technical details

The state is represented by a Vec with components:

Index	Name	Min	Max
0	Position	-1.2	0.6
1	Velocity	-0.07	0.07

References

• Moore, A. W. (1990). Efficient memory-based learning for robot control.
• Singh, S. P., & Sutton, R. S. (1996). Reinforcement learning with replacing eligibility traces. Recent Advances in Reinforcement Learning, 123-158. - Sutton, R. S., & Barto, A. G. (1998). Reinforcement learning: An introduction (Vol. 1, No. 1). Cambridge: MIT press.

---

1. See https://en.wikipedia.org/wiki/Mountain_car_problem ↩

Implementations

impl MountainCar

pub fn new(x: f64, v: f64) -> MountainCar

Trait Implementations

impl Default for MountainCar

fn default() -> MountainCar

Returns the "default value" for a type.

impl Domain for MountainCar

type StateSpace = ProductSpace<Interval>

State space representation type class.

type ActionSpace = Ordinal

Action space representation type class.

fn emit(&self) -> Observation<Vec<f64>>

Emit an observation of the current state of the environment.

fn step(&mut self, action: &usize) -> (Observation<Vec<f64>>, Reward)

Transition the environment forward a single step given an action, a.

fn state_space(&self) -> Self::StateSpace

Returns an instance of the state space type class.

fn action_space(&self) -> Ordinal

Returns an instance of the action space type class.

fn transition(
    &mut self,
    a: Action<Self>
) -> Transition<State<Self>, Action<Self>>

fn rollout<F>(
    self,
    pi: F,
    step_limit: Option<usize>
) -> Trajectory<State<Self>, Action<Self>> where
    F: FnMut(&State<Self>) -> Action<Self>,
    Self: Sized,