Multiscalemicro:model_dynamics

Author: JusteRaimbault

library/src/main/scala/org/openmole/spatialdata/application/multimodelmacro/RealData.scala

@@ -38,7 +38,6 @@ object RealData {
                     distancesFile: String,
                     datesFile: String):  (Matrix, Matrix, Array[Double]) = {
     implicit val m: MatrixImplementation = Matrix.defaultImplementation
-    rng.setSeed(seed)
 
     // load populations, dates, distance matrix
     val populationMatrix: Matrix = Matrix(CSV.readMat(populationsFile))

library/src/main/scala/org/openmole/spatialdata/application/multimodelmacro/RunSDG.scala

@@ -22,7 +22,7 @@ object RunSDG extends App {
 
   // real data setup
   val initialState = RealData.setupRealData(
-
+    "","","" //TODO
   )
 
   val result = SDG.runMultiModelMacro(
@@ -55,7 +55,7 @@ object RunSDG extends App {
 
   utils.log(s"Indicators: Emissions = ${SDG.cumulatedFlows(result)}; Innovation = ${- SDG.averageUtility(result)}; Infrastructure = ${SDG.averageDistance(result)}; Eco inequality = ${SDG.giniEconomicWealth(result)}; Wealth = ${SDG.averageWealth(result)}")
 
-  val realresult: RealData.Result = RealData.Result(result)
+  val realresult: RealData.Result = RealData.Result(result.asInstanceOf[MultiMacroResult])
 
   utils.log(s"Fit on populations : logmse =  ${realresult.logmse} ; mselog = ${realresult.mselog}")
 

library/src/main/scala/org/openmole/spatialdata/application/multimodelmacro/SDG.scala

@@ -93,7 +93,7 @@ object SDG {
     // note: constructor are a mess: legacy vs mutation model, no proper for mutation
     // initial innov utility is 1
     val distrib = innovationUtilityDistrib match {case "normal" => InnovationUtilityNormalDistribution(); case "log-normal" => InnovationUtilityLogNormalDistribution()}
-    val innovModel = Innovation(populationMatrix,dmat,dates,rng,0.0,innovationWeight,innovationGravityDecay,innovationInnovationDecay,
+    val innovModel = Innovation(population,distances,dates,rng,0.0,innovationWeight,innovationGravityDecay,innovationInnovationDecay,
       mutationInnovation(_,_,_, innovationMutationRate, innovationNewInnovationHierarchy, innovationEarlyAdoptersRate, innovationUtilityStd, distrib),
       1.0
     )
@@ -102,12 +102,12 @@ object SDG {
     // eco exchanges - ! add gravitypotentials in eco state (total emission: cumulated interactions across models)
     // ! issue: not same unit for innov / eco (supply - demand gravity potentials) -> should compute something equivalent?
     // ! marius weight? - rescale pop updates?
-    val ecoModel = EconomicExchanges(populationMatrix, dmat, dates, ecoWeight, ecoSizeEffectOnDemand, ecoSizeEffectOnSupply, ecoGravityDecay, ecoWealthToPopulationExponent, ecoPopulationToWealthExponent)
+    val ecoModel = EconomicExchanges(population, distances, dates, ecoWeight, ecoSizeEffectOnDemand, ecoSizeEffectOnSupply, ecoGravityDecay, ecoWealthToPopulationExponent, ecoPopulationToWealthExponent)
     val ecoInitialState = EconomicExchanges.initialState(ecoModel)
 
     // coevolution
     // rq: flows also in other unit (pop rescaled) -> must compute pop-based only for comparability
-    val coevolModel = Coevolution(populationMatrix, Array(dmat), EmptyMatrix(), dates, 0.0, coevolWeight, coevolGamma, coevolGravityDecay,
+    val coevolModel = Coevolution(population, Array(distances), EmptyMatrix(), dates, 0.0, coevolWeight, coevolGamma, coevolGravityDecay,
       0.0, 1.0, 1.0, coevolNetworkGmax, coevolNetworkExponent, coevolNetworkThresholdQuantile)
     val coevolInitialState = Coevolution.initialState(coevolModel)
 

library/src/main/scala/org/openmole/spatialdata/application/multiscalemicro/MultiscaleMicro.scala

@@ -112,7 +112,16 @@ object MultiscaleMicro {
       * @param positioning positioning function
       * @return
       */
-    def addBuildings(currentState: State, numberToAdd: Int, buildingWidth: Double, buildingHeight: Int, shareHousing: Double, streetBuffer: Double, positioning: State => (Double, Double), patchSize: Double, lambdaDensity: Double)(implicit rng: Random): State = {
+    def addBuildings(currentState: State,
+                     numberToAdd: Int,
+                     buildingWidth: Double,
+                     buildingHeight: Int,
+                     shareHousing: Double,
+                     streetBuffer: Double,
+                     positioning: State => (Double, Double),
+                     patchSize: Double,
+                     lambdaDensity: Double
+                    )(implicit rng: Random): State = {
 
       val (xmin,xmax,ymin,ymax) = bbox(currentState)
       val (x, y) = positioning(currentState)
@@ -231,7 +240,7 @@ object MultiscaleMicro {
     def initialGrid(initialNetwork: TransportationNetwork, patchSize: Double, lambdaDensity: Double): Grid = {
       val nodesidseq = initialNetwork.network.nodes.map(_.id).toSeq
       // each node is the center of a patch
-      val (distToStationsMap, distToCentresMap) = distances(initialNetwork)
+      val (distToStationsMap, distToCentresMap) = TransportationNetwork.distances(initialNetwork)
       val coordsMap = initialNetwork.network.nodes.map(n => (n.id,n.position)).toMap
       val empty = emptyGrid(nodesidseq, coordsMap, distToStationsMap, distToCentresMap, patchSize)
       aggregateBuildingsToGrid(empty, Polygons.empty, lambdaDensity)
@@ -257,18 +266,7 @@ object MultiscaleMicro {
       )
     }
 
-    /**
-      * compute distance from nodes to stations and centres
-      *  indices are not needed as nw nodes are fixed
-      * @param nw network
-      * @return maps Node id => distance for stations and centres
-      */
-    def distances(nw: TransportationNetwork): (Map[Int,Double],Map[Int,Double]) = {
-      val nodesseq = nw.network.nodes.toSeq
-      val distToStationsMap: Map[Int,Double] = nw.network.shortestPathTo(nodesseq,nw.stations).toSeq.groupMap(_._1._1)(_._2._3).map{case (n,dists) => (n.id, dists.min)}
-      val distToCentresMap: Map[Int,Double] = nw.network.shortestPathTo(nodesseq,nw.centres).toSeq.groupMap(_._1._1)(_._2._3).map{case (n,dists) => (n.id, dists.min)}
-      (distToStationsMap, distToCentresMap)
-    }
+
 
     /**
       * recompute counts and densities in grid cells given buildings
@@ -324,6 +322,24 @@ object MultiscaleMicro {
     }
 
 
+    def updateDensityVariables(model: MultiscaleMicro, state: State): State = {
+      state.copy(
+        grid = aggregateBuildingsToGrid(state.grid, state.buildings, model.lambdaDensity)
+      )
+    }
+
+    /**
+     * Recompute distance and density variables
+     * @param model model
+     * @param state state
+     * @return
+     */
+    def updateVariables(model: MultiscaleMicro, state: State): State = {
+      val updatedNetworkDistances = TransportationNetwork.updateNetworkDistances(state)
+      updateDensityVariables(model, updatedNetworkDistances)
+    }
+
+
   }
 
   /**
@@ -370,6 +386,48 @@ object MultiscaleMicro {
       utils.log(s"Network nodes coordinates in [${nw.nodes.map(_.position._1).min}, ${nw.nodes.map(_.position._1).max}] x [${nw.nodes.map(_.position._2).min}, ${nw.nodes.map(_.position._2).max}]")
       TransportationNetwork(nw, centres, initialTrLinks.toSeq, centres)
     }
+
+
+    /**
+     * compute distance from nodes to stations and centres
+     *  indices are not needed as nw nodes are fixed
+     * @param nw network
+     * @return maps Node id => distance for stations and centres
+     */
+    def distances(nw: TransportationNetwork): (Map[Int,Double],Map[Int,Double]) = {
+      val nodesseq = nw.network.nodes.toSeq
+      val distToStationsMap: Map[Int,Double] = nw.network.shortestPathTo(nodesseq,nw.stations).toSeq.groupMap(_._1._1)(_._2._3).map{case (n,dists) => (n.id, dists.min)}
+      val distToCentresMap: Map[Int,Double] = nw.network.shortestPathTo(nodesseq,nw.centres).toSeq.groupMap(_._1._1)(_._2._3).map{case (n,dists) => (n.id, dists.min)}
+      (distToStationsMap, distToCentresMap)
+    }
+
+    /**
+     * Recompute network distances
+     * @param state state with updated network but not distances
+     * @return
+     */
+    def updateNetworkDistances(state: State): State = {
+      val nodesidseq = state.network.network.nodes.map(_.id).toSeq
+      val (distToStationsMap, distToCentresMap) = distances(state.network)
+      val updatedGrid = state.grid.copy(
+        patchDistancesToNetwork = nodesidseq.map(distToStationsMap(_)),
+        patchDistancesToCentre = nodesidseq.map(distToCentresMap(_))
+      )
+      state.copy(grid=updatedGrid)
+    }
+
+    /**
+     * Add long distance links to the network
+     * @param model model
+     * @param state state
+     * @return
+     */
+    def evolveNetwork(model:MultiscaleMicro, state: State): State = {
+      // TODO
+      state
+    }
+
+
   }
 
 
@@ -397,6 +455,21 @@ object MultiscaleMicro {
     //sealed trait DeveloperAgentsSetupMode
     //case object DeveloperAgentsUniformSetup extends DeveloperAgentsSetupMode
     //case object DeveloperAgentsHeterogenousSetup extends DeveloperAgentsSetupMode
+
+
+    /**
+     * Core function simulating microscopic dynamics of developers
+     *  -> compute utilities
+     *  -> compute a positioning function for addBuildings
+     * @param model model
+     * @param state state
+     * @return
+     */
+    def developerAgentsProjects(model: MultiscaleMicro, state: State): State = {
+
+      state
+    }
+
   }
 
   case class Result(
@@ -411,8 +484,15 @@ object MultiscaleMicro {
     * @param state state
     * @return
     */
-  def step(model: MultiscaleMicro, state: State): State = {
-    state
+  def step(model: MultiscaleMicro, state: State)(implicit rng: Random): State = {
+    def oneStepBuildings(state: State): State = DeveloperAgent.developerAgentsProjects(model, state)
+      //state.copy(
+      //buildings = State.addBuildings(// behavior of developer agents -> discrete choices
+      //)(rng))
+    val withNewBuildings = Iterator.iterate(state)(oneStepBuildings).take(model.mesoStepInterval).toList.last
+    val withUpdatedDensities = Grid.updateDensityVariables(model, withNewBuildings)
+    val withUpdatedNetwork = TransportationNetwork.evolveNetwork(model, withUpdatedDensities)
+    Grid.updateVariables(model, withUpdatedNetwork) // in practice last step
   }
 
   /**