StateSpace.hs

{-# OPTIONS_GHC -Wall -fno-warn-orphans #-}
{-# LANGUAGE FlexibleContexts, FlexibleInstances, TypeFamilies #-}
{-# LANGUAGE Safe #-}

{- | 
Module      :  Physics.Learn.StateSpace
Copyright   :  (c) Scott N. Walck 2014-2019
License     :  BSD3 (see LICENSE)
Maintainer  :  Scott N. Walck <walck@lvc.edu>
Stability   :  experimental

A 'StateSpace' is an affine space where the associated vector space
has scalars that are instances of 'Fractional'.
If p is an instance of 'StateSpace', then the associated vectorspace
'Diff' p is intended to represent the space of (time) derivatives
of paths in p.

'StateSpace' is very similar to Conal Elliott's 'AffineSpace'.
-}

module Physics.Learn.StateSpace
    ( StateSpace(..)
    , (.-^)
    , Time
    , DifferentialEquation
    , InitialValueProblem
    , EvolutionMethod
    , SolutionMethod
    , stepSolution
    , eulerMethod
    )
    where

import Data.AdditiveGroup
    ( AdditiveGroup(..)
    )
import Data.VectorSpace
    ( VectorSpace(..)
    )
import Physics.Learn.Position
    ( Position
    , shiftPosition
    , displacement
    )
import Physics.Learn.CarrotVec
    ( Vec
    , (^*)
    , (^-^)
    )

infixl 6 .+^, .-^
infix  6 .-.

-- | An instance of 'StateSpace' is a data type that can serve as the state
--   of some system.  Alternatively, a 'StateSpace' is a collection of dependent
--   variables for a differential equation.
--   A 'StateSpace' has an associated vector space for the (time) derivatives
--   of the state.  The associated vector space is a linearized version of
--   the 'StateSpace'.
class (VectorSpace (Diff p), Fractional (Scalar (Diff p))) => StateSpace p where
  -- | Associated vector space
  type Diff p
  -- | Subtract points
  (.-.)  :: p -> p -> Diff p
  -- | Point plus vector
  (.+^)  :: p -> Diff p -> p

-- | The scalars of the associated vector space can be thought of as time intervals.
type Time p = Scalar (Diff p)

-- | Point minus vector
(.-^) :: StateSpace p => p -> Diff p -> p
p .-^ v = p .+^ negateV v

instance StateSpace Double where
    type Diff Double = Double
    (.-.) = (-)
    (.+^) = (+)

instance StateSpace Vec where
    type Diff Vec = Vec
    (.-.) = (^-^)
    (.+^) = (^+^)

-- | Position is not a vector, but displacement (difference in position) is a vector.
instance StateSpace Position where
    type Diff Position = Vec
    (.-.) = flip displacement
    (.+^) = flip shiftPosition

instance (StateSpace p, StateSpace q, Time p ~ Time q) => StateSpace (p,q) where
  type Diff (p,q)   = (Diff p, Diff q)
  (p,q) .-. (p',q') = (p .-. p', q .-. q')
  (p,q) .+^ (u,v)   = (p .+^ u, q .+^ v)

instance (StateSpace p, StateSpace q, StateSpace r, Time p ~ Time q
         ,Time q ~ Time r) => StateSpace (p,q,r) where
  type Diff (p,q,r)      = (Diff p, Diff q, Diff r)
  (p,q,r) .-. (p',q',r') = (p .-. p', q .-. q', r .-. r')
  (p,q,r) .+^ (u,v,w)    = (p .+^ u, q .+^ v, r .+^ w)

inf :: a -> [a]
inf x = x : inf x

instance AdditiveGroup v => AdditiveGroup [v] where
    zeroV   = inf zeroV
    (^+^)   = zipWith (^+^)
    negateV = map negateV

instance VectorSpace v => VectorSpace [v] where
    type Scalar [v] = Scalar v
    c *^ xs = [c *^ x | x <- xs]

instance StateSpace p => StateSpace [p] where
    type Diff [p] = [Diff p]
    (.-.) = zipWith (.-.)
    (.+^) = zipWith (.+^)

-- | A differential equation expresses how the dependent variables (state)
--   change with the independent variable (time).
--   A differential equation is specified by giving the (time) derivative
--   of the state as a function of the state.
--   The (time) derivative of a state is an element of the associated vector space.
type DifferentialEquation state = state -> Diff state

-- | An initial value problem is a differential equation along with an initial state.
type InitialValueProblem state = (DifferentialEquation state, state)

-- | A (numerical) solution method is a way of converting
--   an initial value problem into a list of states (a solution).
--   The list of states need not be equally spaced in time.
type SolutionMethod state = InitialValueProblem state -> [state]

-- | An evolution method is a way of approximating the state
--   after advancing a finite interval in the independent
--   variable (time) from a given state.
type EvolutionMethod state
    = DifferentialEquation state  -- ^ differential equation
    -> Time state                 -- ^ time interval
    -> state                      -- ^ initial state
    -> state                      -- ^ evolved state

-- | Given an evolution method and a time step, return the solution method
--   which applies the evolution method repeatedly with with given time step.
--   The solution method returned will produce an infinite list of states.
stepSolution :: EvolutionMethod state -> Time state -> SolutionMethod state
stepSolution ev dt (de, ic) = iterate (ev de dt) ic

-- | The Euler method is the simplest evolution method.
--   It increments the state by the derivative times the time step.
eulerMethod :: StateSpace state => EvolutionMethod state
eulerMethod de dt st = st .+^ de st ^* dt