Merge branch 'main' into pump_spec_tests

Author: mohsen-ghaffari1992

src/main/scala/symsim/Agent.scala

@@ -35,7 +35,7 @@ trait Agent[State, ObservableState, Action, Reward, Scheduler[_]]:
     * convenient to maintain discretizations with the agent, as this makes the
     * examples compact.
     */
-  def discretize (s: State): ObservableState
+  def observe (s: State): ObservableState
 
   /** Return the reward assigned to state s when action a is given (the
     * reward estimation function)

src/main/scala/symsim/ExactRL.scala

@@ -49,7 +49,7 @@ trait ExactRL[State, ObservableState, Action, Reward, Scheduler[_]]
   def learningEpisode (f: VF, s_t: State): Scheduler[VF] =
     def p (f: VF, s: State, a: Action): Boolean = agent.isFinal (s)
     for
-      a <- chooseAction (f) (agent.discretize (s_t))
+      a <- chooseAction (f) (agent.observe (s_t))
       fin <- Monad[Scheduler].iterateUntilM (f, s_t, a) (learningEpoch) (p)
     yield fin._1
 

src/main/scala/symsim/QLearning.scala

@@ -24,8 +24,8 @@ trait QLearning[State, ObservableState, Action, Reward, Scheduler[_]]
       sa_tt <- agent.step (s_t) (a_t)
       (s_tt, r_tt) = sa_tt
 
-      ds_t = agent.discretize (s_t)
-      ds_tt = agent.discretize (s_tt)
+      ds_t = agent.observe (s_t)
+      ds_tt = agent.observe (s_tt)
 
       // Q-learning is off-policy (p.844 in Russel & Norvig)
       a_tt = bestAction (q) (ds_tt)

src/main/scala/symsim/Sarsa.scala

@@ -24,10 +24,10 @@ trait Sarsa[State, ObservableState, Action, Reward, Scheduler[_]]
          sa_tt <- agent.step (s_t) (a_t)
          (s_tt, r_tt) = sa_tt
          // SARSA: on-policy (p.844 in Russel & Norvig)
-         a_tt <- chooseAction (q) (agent.discretize (s_tt))
+         a_tt <- chooseAction (q) (agent.observe (s_tt))
 
-         ds_t = agent.discretize (s_t)
-         ds_tt = agent.discretize (s_tt)
+         ds_t = agent.observe (s_t)
+         ds_tt = agent.observe (s_tt)
          old_entry = q (ds_t) (a_t)
          correction = r_tt + gamma * q (ds_tt) (a_tt) - old_entry
          qval = old_entry + alpha * correction

src/main/scala/symsim/examples/concrete/breaking/Car.scala

@@ -43,7 +43,7 @@ object Car
       s.v == 0.0 || s.p >= 10 || Math.abs (s.p) >= 1000.0
 
 
-    def discretize (s: CarState): CarObservableState =
+    def observe (s: CarState): CarObservableState =
       require (s.v >= 0, s"s.v = ${s.v} is not non-negative")
       require (s.p >= 0, s"s.p = ${s.p} is not non-negative")
 

src/main/scala/symsim/examples/concrete/golf/Golf.scala

@@ -45,7 +45,7 @@ object Golf
          s == 9
 
       // Golf is discrete
-      def discretize (s: GolfState): GolfObservableState =  s
+      def observe (s: GolfState): GolfObservableState =  s
 
       private def golfReward (s: GolfState) (a: GolfAction): GolfReward = (s, a) match
          case (6, (Club.P, _)) => -100.0

src/main/scala/symsim/examples/concrete/mountaincar/MountainCar.scala

@@ -42,7 +42,7 @@ object MountainCar
     s.p >= 0.5
 
 
-  def discretize (s: CarState): CarObservableState =
+  def observe (s: CarState): CarObservableState =
     require (s.p >= -1.2, s"s.p = ${s.p} is not within the boundaries")
     require (s.p <= 0.5, s"s.p = ${s.p} is not within the boundaries")
     require (s.v >= -1.5, s"s.v = ${s.v} is not within the boundaries")

src/main/scala/symsim/examples/concrete/pumping/Pump.scala

@@ -42,7 +42,7 @@ end PumpState
 
 /** Discretize cut points for flow, head, and tank variables./
   *
-  * We discretize to the first point on the list that is lower than the actual
+  * We observe to the first point on the list that is lower than the actual
   * value of the variable.  So we 'round down'. Values below the last element
   * in the cut-point list are rounded up (to the last value).
   *
@@ -106,7 +106,7 @@ object Pump extends
     s.t >= 4000 || s.h < HEAD_MIN || s.tl > TANK_MAX || s.tl < TANK_MIN
 
 
-  def discretize (s: PumpState): ObservablePumpState =
+  def observe (s: PumpState): ObservablePumpState =
     require (s.tl >= TANK_MIN, s"s.tl = ${s.tl} >= $TANK_MIN")
 
     val df = closest (s.f) (flowCutPoints)

src/main/scala/symsim/examples/concrete/simplebandit/Bandit.scala

@@ -39,7 +39,7 @@ class Bandit (banditReward: List [Randomized[BanditReward]])
   override def isFinal (s: BanditState): Boolean = s
 
   // Bandit is discrete
-  override def discretize (s: BanditState): BanditState = s
+  override def observe (s: BanditState): BanditState = s
 
   override def step (s: BanditState) (a: BanditAction): Randomized[(BanditState, BanditReward)] =
     for r <- banditReward (a) yield (true, r)

src/main/scala/symsim/examples/concrete/simplemaze/Maze.scala

@@ -56,7 +56,7 @@ object Maze
     s == (4, 3) || s == (4, 2)
 
   // Maze is discrete
-  def discretize (s: MazeState): MazeObservableState =  s
+  def observe (s: MazeState): MazeObservableState =  s
 
   private def mazeReward (s: MazeState): MazeReward = s match
     case (4, 3) => 1.0   // Good final state