diff --git a/unified_planning/model/multi_agent/action_parallelizer.py b/unified_planning/model/multi_agent/action_parallelizer.py
new file mode 100644
index 000000000..5668da28c
--- /dev/null
+++ b/unified_planning/model/multi_agent/action_parallelizer.py
@@ -0,0 +1,381 @@
+import unified_planning as up
+from unified_planning.engines.sequential_simulator import UPSequentialSimulator as SequentialSimulator
+import unified_planning.plans as plans
+
+class Parallelizer():
+
+    def __init__(self):
+
+        self.problem = None
+
+
+    # generates the adjacency lists needed by 'GeneratePOP' to generate the final POP plan
+    def Update_POPdict(self, POP_dict, GAMMA, lastAct_ins):
+      if lastAct_ins is not None:
+        POP_dict[lastAct_ins] = [GAMMA[0]]  # the start action
+      lenG = len(GAMMA)
+
+      if lenG == 2:
+        POP_dict[GAMMA[0]] = [GAMMA[1]]   # the endAction is in the adjacency list of the startAction
+
+      if lenG > 2:  # at least one action was parallelized to the start action
+        if lastAct_ins is not None:
+          if POP_dict.get(lastAct_ins) is None:
+            POP_dict[lastAct_ins] = [GAMMA[1]]
+          else:
+            POP_dict[lastAct_ins].append(GAMMA[1])
+        for i in range(1, lenG-1):
+          POP_dict[GAMMA[i]] = [GAMMA[i+1]]
+        POP_dict[GAMMA[0]] = [GAMMA[lenG-1]]
+
+      lastAct_ins = GAMMA[lenG-1]
+
+      return POP_dict, lastAct_ins
+
+    #-------------------------------------------------------------------------------------------------------#
+
+    # generates the POP of the parallelized plan
+    def GeneratePOP(self, listActs):
+      POP = plans.PartialOrderPlan(listActs)
+
+      return POP
+
+    #-------------------------------------------------------------------------------------------------------#
+
+    # writes in a file .plan the resulting parallelized plan
+    def UpdateParallelizedPlan(self, lista, f):
+
+        L = len(lista)
+
+        if self.actionMovement in str(lista[0]) and L>1:
+          stringedAction = str(lista[0])
+          f.write(stringedAction+"; \n")
+
+        if L==0:
+          return
+
+        # in 'list' there is only one action, that is or: an atomic action or a start action that cannot be parallelized
+        elif L==1:
+          stringedAction = str(lista[0])
+          f.write(stringedAction+"; \n")
+
+        # I only have a 'start' and an 'end' action
+        elif L==2:
+          for act in lista:
+            stringedAction = str(act)
+            f.write(stringedAction+"; \n")
+
+        # in the case something has been parallelized to a start action
+        else:
+          stringedAction = str(lista[0])
+          f.write("{" + stringedAction+"}||{")
+          for act in lista[1:L-1]:
+            stringedAction = str(act)
+            f.write(stringedAction+"; ")
+          f.write("} \n")
+          stringedAction = str(lista[L-1])
+          f.write(stringedAction+"; \n")
+
+    #-------------------------------------------------------------------------------------------------------#
+
+    # check if by adding a subplan of 'goto' actions, another action becomes parallelizable
+    def defineAndSolveSubproblem(self, new_goals, curren_state):
+        curr_state = curren_state
+        problem2 = self.problem.clone()
+        initial_values = self.problem.initial_values
+
+        # I set the initial values of prob2 equals to the current state of the simulation
+        for fluent in initial_values:
+            problem2.set_initial_value(fluent, curr_state.get_value(fluent))
+
+        # I set the goal of prob2 equal to the unsatisfied conditions
+        problem2.clear_goals()
+        for necessaryPrecondition in new_goals:
+            problem2.add_goal(necessaryPrecondition)
+
+        with OneshotPlanner(name = "fast-downward") as planner:
+            assert planner.supports(problem2.kind)
+            out_plan2 = planner.solve(problem2).plan
+
+        foundSubplan = False
+        # check the plan is made up pf all movement actions
+        if out_plan2 is not None:
+            foundSubplan = True
+            contatoreAz = 0
+            contatoreGoto = 0
+            for act in out_plan2.actions:
+                contatoreAz += 1
+                curr_state = self.sim.apply(curr_state, act)
+                if self.actionMovement in act.action.name:
+                    contatoreGoto += 1
+            if contatoreGoto != contatoreAz:   # not only movement actions in the subplan --> I do not return the subplan
+                return curren_state, None, False
+
+        return curr_state, out_plan2, foundSubplan
+
+    #-------------------------------------------------------------------------------------------------------#
+
+    # initial check that verifies if a certain action B is parallelizable to 'start_actionA'; that is if B is
+    # applicable even if A is terminated.
+    def initialCheck_applicability(self, actCercasi, actStart, actEnd, current_stateOriginal, current_stateBB):
+
+        if current_stateBB == []:
+            curr_state = current_stateOriginal
+        else:
+            curr_state = current_stateBB
+
+        curr_state = self.sim.apply(curr_state, actCercasi)
+        applicability = self.sim.is_applicable(curr_state, actStart)
+        if applicability:
+            curr_state = self.sim.apply(curr_state, actStart)
+            applicability = self.sim.is_applicable(curr_state, actEnd)
+
+        return applicability
+
+    #-------------------------------------------------------------------------------------------------------#
+
+    # I search for the first action to be parallelized to the start_action
+    def findFirstParallelizableAction(self, i, j, durativeAct, start_action, end_action, current_state2, applicability):
+
+        while j<self.len_plan and not(applicability):
+            cercasi_action = self.seq_plan.actions[j]
+
+            # it isn't possible to parallelize a start_action to another one
+            if durativeAct in cercasi_action.action.name:
+                j+=1
+                continue
+
+            current_stateBB = []
+            firstParalActs_list = []
+            appl = self.sim.is_applicable(current_state2, cercasi_action)
+            if not(appl):
+                unsat_preconds = self.sim.get_unsatisfied_conditions(current_state2, cercasi_action)[0]
+                current_stateBB, subplan, foundSubplan = self.defineAndSolveSubproblem(unsat_preconds, current_state2)
+                if not foundSubplan:
+                    j+=1
+                    continue
+                else:
+                    state = current_state2
+                    for act in subplan.actions:
+                        applicability = self.initialCheck_applicability(act, start_action, end_action, state, [])
+                        if applicability:
+                            state = self.sim.apply(state, act)
+                            appl = True
+                        else:
+                            appl = False
+                            break
+                if appl:
+                  for act in subplan.actions:
+                    firstParalActs_list.append(act)
+            else:
+              appl = True
+            # check the applicability of the first action we want to parallelize, even if the 'start_action' is still ongoing
+            if appl:
+              applicability = self.initialCheck_applicability(cercasi_action, start_action, end_action, current_state2, current_stateBB)
+
+            j+=1
+        j-=1
+
+        return cercasi_action, j, applicability, current_state2, firstParalActs_list
+
+    #-----------------------------------------------------------------------------------------------------------------------------
+
+    # application of the first parallelizable action to the current state
+    def applyFirstParallelizzableAction(self, cercasi_action, j, current_state, GAMMAbis, list_acts_TOinsert, firstParalActs_list):
+
+        current_stateBIS = current_state
+        for act in firstParalActs_list:
+            current_stateBIS = self.sim.apply(current_stateBIS, act)
+            GAMMAbis.append(act)
+        current_stateBIS = self.sim.apply(current_stateBIS, cercasi_action)
+        GAMMAbis.append(cercasi_action)
+        list_acts_TOinsert.append(j)
+
+        return current_stateBIS, GAMMAbis, list_acts_TOinsert
+
+    #------------------------------------------------------------------------------------------------------------------------------
+
+    # application of end_action to the current state. It is done after the parallelization of a sequence to start_action was terminated
+    def applyEndAction(self, end_action, i, start_action, GAMMAbis, list_acts_TOinsert, current_stateBIS):
+        GAMMAbis.append(end_action)
+        list_acts_TOinsert.append(i+1)
+        current_stateBIS = self.sim.apply(current_stateBIS, start_action)
+        current_stateBIS = self.sim.apply(current_stateBIS, end_action)
+
+        return GAMMAbis, list_acts_TOinsert, current_stateBIS
+
+    #-----------------------------------------------------------------------------------------------------------------------------
+
+    # after having parallelied a first action, this function searches for the rest of actions
+    def findRestSequenceToParallelize(self, j, current_stateBIS, GAMMAbis, list_acts_TOinsert):
+        w = 1   # one action has already been parallelized
+        while w < self.N_ACTIONStoPARALLELIZE and j+w<self.len_plan:
+            # j = index of the first action parallelized. It can be different from i+2
+            o = j+w
+            act = self.seq_plan.actions[o]
+            appl = False
+            if "start" in act.action.name:
+                break
+            else:
+                appl = self.sim.is_applicable(current_stateBIS, act)
+
+            if not(appl):
+                break
+            GAMMAbis.append(act)
+            list_acts_TOinsert.append(o)
+            current_stateBIS = self.sim.apply(current_stateBIS, act)
+            w+=1
+
+        return current_stateBIS, GAMMAbis, list_acts_TOinsert, w
+
+    #-----------------------------------------------------------------------------------------------------------------------------
+
+    # check the applicability of the sequence of actions found to be parallelized
+    def checkInRange(self, m, n, current_stateCHECK, applicab):
+        for z in range(m, n):
+            act_z = self.seq_plan.actions[z]
+            applicab = self.sim.is_applicable(current_stateCHECK, act_z)
+
+            if not(applicab):
+                unsat_preconds = self.sim.get_unsatisfied_conditions(current_stateCHECK, act_z)[0]
+                current_stateCHECK, subplan, foundSubplan = self.defineAndSolveSubproblem(unsat_preconds, current_stateCHECK)
+                applicab = self.sim.is_applicable(current_stateCHECK, act_z)
+
+            if applicab:
+                current_stateCHECK = self.sim.apply(current_stateCHECK, act_z)
+            else:
+                break
+
+        return applicab, current_stateCHECK
+
+    #--------------------------------------------------------------------------------------------------------------------
+
+    # changing the order of appliance of some actions, we have to check that the rest of the sequential plan remains applicable
+    def checkApplicability_restOfActions(self, current_stateBIS, i, j, w):
+        applicability = True
+        current_stateCHECK = current_stateBIS
+        if i+2 < j:
+            applicability, current_stateCHECK = self.checkInRange(i+2, j, current_stateCHECK, applicability)
+
+        if applicability and j+w < self.len_plan:
+            applicability, current_stateCHECK = self.checkInRange(j+w, self.len_plan, current_stateCHECK, applicability)
+
+        return applicability
+
+
+    #--------------------------------------------------------------------------------------------------------------------
+    #--------------------------------------------------------------------------------------------------------------------
+
+
+    def parallelize(self, problem, seq_plan, LenSeq_daParallelizzare, pathFile, durativeAct, fluentWhere, actionMovement):
+
+        # parameters:
+        # - problem = the compiled problem
+        # - seq_plan = sequential plan solution found through a classical solver
+        # - LenSeq_daParallelizzare = maximum length of the sequence we want to parallelize to a start-end action
+        # - pathFile = file path on which we want to write the final parallelized plan
+        # - durativeAct = action that we want to parallelize to a sequence of other actions.
+        #                 Usually it is an action that takes a lot of time to be executed
+        # - actionMovement = name of the action of your planning formalism that implies a movement of an agent between two locations
+        # 
+        # output:
+        # - POP_adjList : adjacency list of the parallelized plan. Used for generating POP_plan
+        # - POP_plan : final resuling Partial Order Plan
+
+        self.problem = problem
+        self.seq_plan = seq_plan
+        self.len_plan = len(self.seq_plan.actions)   # number of actions in the sequential plan
+        self.sim = SequentialSimulator(problem)
+        self.current_state = self.sim.get_initial_state()  # Initial value of the fluents of the problem
+        self.actionMovement = actionMovement
+        self.list_actions_inserted = []  # actions of the sequential plan, already inserted in the parallelized plan
+        self.N_ACTIONStoPARALLELIZE = LenSeq_daParallelizzare
+
+        f = open(pathFile,"w")  #Sequential file .plan where the parallelized plan will be written
+
+        # needed in order to create the POP structure
+        lastAct_ins = None
+        POP_adjList = {}
+
+        i = 0
+        while i<self.len_plan:
+
+            GAMMA = []  # actions that will be inserte in the parallelized plan at the end of the curret iteration
+            curr_action = seq_plan.actions[i]  # the i-th action of the sequential plan
+
+            # case 1: action already inserted in the output parallelizedd plan
+            if i in self.list_actions_inserted:
+                i+=1
+                continue
+
+            #------------------- AZIONE ATOMICA -----------------------
+            # case 2: atomic action
+            # I simply write the action in the plan and simulate the respective event.
+            if "start" not in curr_action.action.name:
+                firstParalActs_list = []
+
+                if not(self.sim.is_applicable(self.current_state, curr_action)):
+                    unsat_preconds = self.sim.get_unsatisfied_conditions(self.current_state, curr_action)[0]
+                    self.current_state, subplan, foundSubplan = self.defineAndSolveSubproblem(unsat_preconds, self.current_state)
+                    for act in subplan.actions:
+                        firstParalActs_list.append(act)
+
+                self.current_state = self.sim.apply(self.current_state, curr_action)
+
+                firstParalActs_list.append(curr_action)
+                for act in firstParalActs_list:
+                    GAMMA.append(act)
+
+                self.list_actions_inserted.append(i)
+                i+=1
+
+            # durative action: a start action
+            else:
+                start_action = curr_action   # curr_action is the 'startDurative' action
+                GAMMA.append(start_action)
+                self.list_actions_inserted.append(i)
+                end_action = seq_plan.actions[i+1]
+
+                applicability = False
+                j=i+2  #+2 because I have to skip the 'start durative action' and the 'end durative action'
+                while j<self.len_plan and not(applicability):
+                    cercasi_action, j, applicability, current_stateBIS, firstParalActs_list = self.findFirstParallelizableAction(i, j, durativeAct, start_action, end_action, self.current_state, applicability)
+                    # j is the index of the first parallelizable action, if one exists
+
+                    if applicability:
+                        list_acts_TOinsert = []
+                        GAMMAbis = GAMMA.copy()
+
+                        current_stateBIS, GAMMAbis, list_acts_TOinsert = self.applyFirstParallelizzableAction(cercasi_action, j, current_stateBIS, GAMMAbis, list_acts_TOinsert, firstParalActs_list)
+
+                        current_stateBIS, GAMMAbis, list_acts_TOinsert, w = self.findRestSequenceToParallelize(j,  current_stateBIS, GAMMAbis, list_acts_TOinsert)
+
+                        if applicability:
+                            GAMMAbis, list_acts_TOinsert, current_stateBIS = self.applyEndAction(end_action, i, start_action, GAMMAbis, list_acts_TOinsert, current_stateBIS)
+
+                            applicability =  self.checkApplicability_restOfActions(current_stateBIS, i, j, w)
+
+                    j+=1
+                # end of the while used to search for a sequence of actions to be parallelized to 'start action'
+
+                if applicability:
+                    self.current_state = current_stateBIS
+                    GAMMA = GAMMAbis
+                    for k in list_acts_TOinsert:
+                        self.list_actions_inserted.append(k)
+
+                else:
+                    self.current_state = self.sim.apply(self.current_state, start_action)
+
+                i+=1
+
+            # update of the .plan file and of the POP
+            self.UpdateParallelizedPlan(GAMMA, f)
+            POP_adjList, lastAct_ins = self.Update_POPdict(POP_adjList, GAMMA, lastAct_ins)
+
+            # end of the main while
+
+        POP_plan = self.GeneratePOP(POP_adjList)
+        f.close()
+
+        return POP_adjList, POP_plan