Included method to tackle emergence in two_rooms_one_door_modified + Better access of different settings in marl_adapted + Added and modified a lot of config files

marl_factory_grid/configs/custom/MultiAgentConfigs/dirt_quadrant_eval_config.yaml

@@ -19,31 +19,21 @@ Agents:
   Sigmund:
     Actions:
       - Move4
-      #- Clean
       - Noop
     Observations:
-      # - Walls
-      # - Other
       - DirtPiles
       - Self
     Positions:
       - (9,1)
-      #- (9,9)
-      #- (4,5)
   Wolfgang:
     Actions:
       - Move4
-      #- Clean
       - Noop
     Observations:
-      # - Walls
-      # - Other
       - DirtPiles
       - Self
     Positions:
       - (9,5)
-      #- (9,9)
-      #- (4,5)
 
 Entities:
   DirtPiles:

marl_factory_grid/configs/custom/MultiAgentConfigs/dirt_quadrant_train_config.yaml

@@ -22,8 +22,6 @@ Agents:
       #- Clean
       #- Noop
     Observations:
-      # - Walls
-      # - Other
       - DirtPiles
       - Self
     Positions:
@@ -39,8 +37,6 @@ Agents:
       #- Clean
       #- Noop
     Observations:
-      # - Walls
-      # - Other
       - DirtPiles
       - Self
     Positions:

marl_factory_grid/configs/custom/MultiAgentConfigs/two_rooms_one_door_modified_eval_config.yaml

@@ -14,40 +14,30 @@ General:
 # In "two rooms one door" scenario 2 agents spawn in 2 different rooms that are connected by a single door. Their aim
 # is to reach the destination in the room they didn't spawn in leading to a conflict at the door.
 Agents:
-  Wolfgang:
-    Actions:
-      - Move8
-      - DoorUse
-      - Noop
-    Observations:
-      - DirtPiles
-      - Self
-    #Positions:
-      #- (1,1)
-      #- (2,1)
-      #- (3,1)
-      #- (4,1)
-      #- (5,1)
-      #- (6,1)
   Sigmund:
     Actions:
-      - Move8
+      - Move4
       - DoorUse
       - Noop
     Observations:
       - DirtPiles
       - Self
-    #Positions:
-      #- (1,13)
-      #- (2,13)
-      #- (3,13)
-      #- (4,13)
-      #- (5,13)
-      #- (6,13)
+    Positions:
+      - (3,1)
+  Wolfgang:
+    Actions:
+      - Move4
+      - DoorUse
+      - Noop
+    Observations:
+      - DirtPiles
+      - Self
+    Positions:
+      - (3,13)
 
 Entities:
   DirtPiles:
-    coords_or_quantity: (3,12), (3,2) # This order is required, because agent 0 needs to reach (3, 12) and agent 1 (3, 2)
+    coords_or_quantity: (2,1), (3,12), (2,13), (3,2) # Static form: auxiliary pile, primary pile, auxiliary pile, ...
     initial_amount: 0.5 # <1 to ensure that the robot which first attempts to clean this field, can remove the dirt in one action
     clean_amount: 1
     dirt_spawn_r_var: 0
@@ -58,8 +48,8 @@ Entities:
 
 Rules:
   # Environment Dynamics
-  DoorAutoClose:
-    close_frequency: 10
+  #DoorAutoClose:
+    #close_frequency: 10
 
   # Utilities
   # This rule defines the collision mechanic, introduces a related DoneCondition and lets you specify rewards.

marl_factory_grid/configs/custom/MultiAgentConfigs/two_rooms_one_door_modified_eval_config_emergent.yaml

@@ -14,19 +14,19 @@ General:
 # In "two rooms one door" scenario 2 agents spawn in 2 different rooms that are connected by a single door. Their aim
 # is to reach the destination in the room they didn't spawn in leading to a conflict at the door.
 Agents:
-  Wolfgang:
+  Sigmund:
     Actions:
-      - Move8
+      - Move4
       - DoorUse
       - Noop
     Observations:
       - DirtPiles
       - Self
     Positions:
-      - (3,1) # Agent spawnpoint
-  Sigmund:
+      - (3,1)
+  Wolfgang:
     Actions:
-      - Move8
+      - Move4
       - DoorUse
       - Noop
     Observations:
@@ -37,7 +37,7 @@ Agents:
 
 Entities:
   DirtPiles:
-    coords_or_quantity: (3,12), (3,2)
+    coords_or_quantity: (3,12), (3,2) # Static form: auxiliary pile, primary pile, auxiliary pile, ...
     initial_amount: 0.5 # <1 to ensure that the robot which first attempts to clean this field, can remove the dirt in one action
     clean_amount: 1
     dirt_spawn_r_var: 0
@@ -48,8 +48,8 @@ Entities:
 
 Rules:
   # Environment Dynamics
-  DoorAutoClose:
-    close_frequency: 10
+  #DoorAutoClose:
+    #close_frequency: 10
 
   # Utilities
   # This rule defines the collision mechanic, introduces a related DoneCondition and lets you specify rewards.
@@ -58,5 +58,5 @@ Rules:
 
   # Done Conditions
   #DoneOnAllDirtCleaned:
-  #DoneAtMaxStepsReached:
-    #max_steps: 100
+  #DoneAtMaxStepsReached: # Mayne Required since door blocking will result in infinite loop
+    #max_steps: 1000

marl_factory_grid/configs/custom/dirt_quadrant_eval_config.yaml

@@ -19,11 +19,8 @@ Agents:
   #Sigmund:
     #Actions:
       #- Move4
-      #- Clean
       #- Noop
     #Observations:
-      # - Walls
-      # - Other
       #- DirtPiles
       #- Self
     #Positions:
@@ -33,17 +30,13 @@ Agents:
   Wolfgang:
     Actions:
       - Move4
-      #- Clean
-      #- Noop
     Observations:
-      # - Walls
-      # - Other
       - DirtPiles
       - Self
     Positions:
       - (9,5)
-      #- (9,9)
-      #- (4,5)
+      - (9,9)
+      - (4,5)
 
 Entities:
   DirtPiles:

marl_factory_grid/configs/custom/dirt_quadrant_train_config.yaml

@@ -19,11 +19,7 @@ Agents:
   #Sigmund:
     #Actions:
       #- Move4
-      #- Clean
-      #- Noop
     #Observations:
-      # - Walls
-      # - Other
       #- DirtPiles
       #- Self
     #Positions:
@@ -36,11 +32,7 @@ Agents:
   Wolfgang:
     Actions:
       - Move4
-      #- Clean
-      #- Noop
     Observations:
-      # - Walls
-      # - Other
       - DirtPiles
       - Self
     Positions:

marl_factory_grid/configs/custom/two_rooms_one_door_modified_eval_config.yaml

@@ -0,0 +1,62 @@
+General:
+  env_seed: 69
+  # Individual vs global rewards
+  individual_rewards: true
+  # The level.txt file to load from marl_factory_grid/levels
+  level_name: two_rooms_modified
+  # View Radius; 0 = full observatbility
+  pomdp_r: 0
+  # Print all messages and events
+  verbose: false
+  # Run tests
+  tests: false
+
+# In "two rooms one door" scenario 2 agents spawn in 2 different rooms that are connected by a single door. Their aim
+# is to reach the destination in the room they didn't spawn in leading to a conflict at the door.
+Agents:
+  #Sigmund:
+    #Actions:
+      #- Move4
+      #- DoorUse
+    #Observations:
+      #- DirtPiles
+      #- Self
+    #Positions:
+      #- (3,1)
+      #- (2,1)
+  Wolfgang:
+    Actions:
+      - Move4
+      - DoorUse
+    Observations:
+      - DirtPiles
+      - Self
+    Positions:
+      - (3,13)
+      - (2,13)
+
+Entities:
+  DirtPiles:
+    coords_or_quantity: (2,13), (3,2) # (2,1), (3,12)
+    initial_amount: 0.5 # <1 to ensure that the robot which first attempts to clean this field, can remove the dirt in one action
+    clean_amount: 1
+    dirt_spawn_r_var: 0
+    max_global_amount: 12
+    max_local_amount: 1
+
+  Doors: { }
+
+Rules:
+  # Environment Dynamics
+  #DoorAutoClose:
+    #close_frequency: 10
+
+  # Utilities
+  # This rule defines the collision mechanic, introduces a related DoneCondition and lets you specify rewards.
+  WatchCollisions:
+    done_at_collisions: false
+
+  # Done Conditions
+  #DoneOnAllDirtCleaned:
+  #DoneAtMaxStepsReached: # Mayne Required since door blocking will result in infinite loop
+    #max_steps: 1000

marl_factory_grid/configs/custom/two_rooms_one_door_modified_train_config.yaml

@@ -0,0 +1,75 @@
+General:
+  env_seed: 69
+  # Individual vs global rewards
+  individual_rewards: true
+  # The level.txt file to load from marl_factory_grid/levels
+  level_name: two_rooms_modified
+  # View Radius; 0 = full observatbility
+  pomdp_r: 0
+  # Print all messages and events
+  verbose: false
+  # Run tests
+  tests: false
+
+# In "two rooms one door" scenario 2 agents spawn in 2 different rooms that are connected by a single door. Their aim
+# is to reach the destination in the room they didn't spawn in leading to a conflict at the door.
+Agents:
+  #Sigmund:
+    #Actions:
+      #- Move4
+    #Observations:
+      #- DirtPiles
+      #- Self
+    #Positions:
+      #- (3,1)
+      #- (1,1)
+      #- (3,1)
+      #- (5,1)
+      #- (3,1)
+      #- (1,8)
+      #- (3,1)
+      #- (5,8)
+  Wolfgang:
+    Actions:
+      - Move4
+    Observations:
+      - DirtPiles
+      - Self
+    Positions:
+      - (3,13)
+      - (2,13)
+      - (1,13)
+      - (3,13)
+      - (1,8)
+      - (2,6)
+      - (3,10)
+      - (4,6)
+
+Entities:
+  DirtPiles:
+    coords_or_quantity: (2,13), (3,2) # (2,1), (3,12)
+    initial_amount: 0.5 # <1 to ensure that the robot which first attempts to clean this field, can remove the dirt in one action
+    clean_amount: 1
+    dirt_spawn_r_var: 0
+    max_global_amount: 12
+    max_local_amount: 1
+
+  #Doors: { }
+
+Rules:
+  # Environment Dynamics
+  #DoorAutoClose:
+    #close_frequency: 10
+
+  # Utilities
+  # This rule defines the collision mechanic, introduces a related DoneCondition and lets you specify rewards.
+  WatchCollisions:
+    done_at_collisions: false
+
+  # Done Conditions
+  DoneOnAllDirtCleaned:
+  #DoneAtMaxStepsReached:
+    #max_steps: 100
+
+  AgentSpawnRule:
+    spawn_rule: "order"