fix RL weird errors

Author: flimdejong

docker/runner/docker-compose.yml

@@ -1,5 +1,3 @@
-version: '3'
-
 services:
 
   roboteam_primary_ai:
@@ -225,4 +223,24 @@ services:
       - ./ssl-game-controller-config:/config
       - ./ssl-game-controller-data:/data
     command: -address :8081
-    profiles: ["simulator","diff","game", "RL"]
+    profiles: ["simulator","diff","game", "RL"]
+
+  simulator:
+    image: ${SUMATRA_IMAGE_SIMULATOR:-sumatra:latest}
+    command: "-hl --moduli sim_server --autoRef --maxSpeed --initialCommand PREPARE_KICKOFF_BLUE --waitForAis --playingTime ${SUMATRA_TARGET_STAGE_TIME:-0} --timeout ${SUMATRA_TIMEOUT:-0} ${SUMATRA_SIMULATOR_ADDITIONAL_ARGS:-}"
+    user: "${USER_ID:-1000}:${GROUP_ID:-1000}"
+    volumes:
+      - "./temp/${PROJECT_NAME}/simulator/build:/Sumatra/data"
+      - "./temp/${PROJECT_NAME}/simulator/data:/Sumatra/data"
+    networks:
+      - default
+    profiles: ["RL"]
+
+  ai-blue:
+    image: ${SUMATRA_IMAGE_BLUE:-sumatra:latest}
+    command: "-hl --moduli sim_client --host simulator --aiBlue ${SUMATRA_AI_ADDITIONAL_ARGS:-}"
+    user: "${USER_ID:-1000}:${GROUP_ID:-1000}"
+    environment:
+      - SUMATRA_INFLUX_DB_URL=${SUMATRA_INFLUX_DB_URL:-http://influxdb:8086}
+      - SUMATRA_INFLUX_DB_PASSWORD=${SUMATRA_INFLUX_DB_PASSWORD:-}
+      - SUMATRA_INFLUX_DB_USERNAME=${SUMATRA_INFLUX_DB_USERNAME:-}

docker/runner/ssl-game-controller-config/state-store.json.stream

@@ -1,6 +1,6 @@
 # Nova AI 
 
-RL implementation to swap out the play system
+RL implementation to swap out the play system. Current goal is to make the role division dynamic.
 
 ## Features
 
@@ -12,7 +12,22 @@ RL implementation to swap out the play system
 
 
 ## Explanation .py scripts
-- getRefereeState.py gets the state of the referee
 - GetState.py gets a combined state and contains 2 functions, one to get the ball position and one to get robot position
 - sentActionCommand sends a command using proto to the legacy AI system
 - teleportBall.py to tp the ball to a location we can define in our environment
+
+
+## High-level RL explanation
+
+# Loop
+Agent receives a state in the form of where all the robots are, if we are dribbling and where the ball is.
+
+
+Every loop the game needs to be reset. This means resetting:
+- Time
+- Current stage
+- 
+
+Also when a ball placement happens, we need to extract the designated position and teleport the ball there for faster simulation.
+
+

roboteam_ai/src/RL/README.md

@@ -12,7 +12,7 @@
 
 # Make sure to go back to the main roboteam directory
 current_dir = os.path.dirname(os.path.abspath(__file__))
-roboteam_path = os.path.abspath(os.path.join(current_dir, "..", "..", "..", ".."))
+roboteam_path = os.path.abspath(os.path.join(current_dir, "..", "..", ".."))
 
 # Add to sys.path
 sys.path.append(roboteam_path)

roboteam_ai/src/RL/getState.py

@@ -0,0 +1,57 @@
+
+import sys
+import os
+from google.protobuf.timestamp_pb2 import Timestamp
+from google.protobuf.duration_pb2 import Duration
+
+# Make sure to go back to the main roboteam directory
+current_dir = os.path.dirname(os.path.abspath(__file__))
+roboteam_path = os.path.abspath(os.path.join(current_dir, "..", "..", ".."))
+
+# Add to sys.path
+sys.path.append(roboteam_path)
+
+# Assuming you have generated Python classes from your .proto files
+from roboteam_networking.proto_CI.ssl_gc_state_pb2 import State, GameState, Command
+from roboteam_networking.proto_CI.ssl_gc_referee_message_pb2 import Referee
+from roboteam_networking.proto_CI.ssl_gc_common_pb2 import Team
+
+
+def reset_game_state_blue_kickoff():
+    # Create a new State message
+    new_state = State()
+
+    # Set the GameState to KICKOFF for Blue team
+    new_state.game_state.type = GameState.KICKOFF
+    new_state.game_state.for_team = Team.BLUE
+
+    # Set the current command to KICKOFF for Blue team
+    new_state.command.type = Command.KICKOFF
+    new_state.command.for_team = Team.BLUE
+
+    # Set Blue team as the first kickoff team
+    new_state.first_kickoff_team = Team.BLUE
+
+    # Reset stage and timers
+    new_state.stage = Referee.NORMAL_FIRST_HALF
+    new_state.stage_time_elapsed.Clear()
+    new_state.stage_time_left.Clear()
+    new_state.match_time_start.GetCurrentTime()  # Set to current time
+
+    # Reset team states
+    for team, is_blue in [("Blue", True), ("Yellow", False)]:
+        team_info = new_state.team_state[team]
+        team_info.name = team
+        team_info.goals = 0
+        team_info.yellow_cards.clear()
+        team_info.red_cards.clear()
+        team_info.timeouts_left = 4  # Or appropriate number
+        team_info.timeout_time_left.CopyFrom(Duration(seconds=300))  # 5 minutes
+        team_info.ball_placement_failures = 0
+        team_info.on_positive_half = is_blue  # Blue on positive half, Yellow on negative
+
+    return new_state
+
+# Usage
+new_state_change = reset_game_state_blue_kickoff()
+# Use this new_state_change to update your game controller state

roboteam_ai/src/RL/resetReferee.py

@@ -4,7 +4,7 @@
 
 # Make sure to go back to the main roboteam directory
 current_dir = os.path.dirname(os.path.abspath(__file__))
-roboteam_path = os.path.abspath(os.path.join(current_dir, "..", "..", "..", ".."))
+roboteam_path = os.path.abspath(os.path.join(current_dir, "..", "..", ".."))
 
 # Add to sys.path
 sys.path.append(roboteam_path)

roboteam_ai/src/RL/sentActionCommand.py

@@ -13,7 +13,7 @@
 
 # Make sure to go back to the main roboteam directory
 current_dir = os.path.dirname(os.path.abspath(__file__))
-roboteam_path = os.path.abspath(os.path.join(current_dir, "..", "..", "..", ".."))
+roboteam_path = os.path.abspath(os.path.join(current_dir, "..", "..", ".."))
 
 # Add to sys.path
 sys.path.append(roboteam_path)

roboteam_ai/src/RL/teleportBall.py

@@ -1,29 +1,35 @@
 import asyncio
 import websockets
 from google.protobuf.json_format import MessageToJson, Parse
-from ssl_gc_api_pb2 import Input, Output
+from ssl_gc_api_pb2 import Input
+from ssl_gc_change_pb2 import Change
+from ssl_gc_common_pb2 import Team
+from ssl_gc_state_pb2 import Command
 
-async def reset_match(uri='ws://localhost:8081/api/control'):
+async def reset_and_stop_match(uri='ws://localhost:8081/api/control'):
     async with websockets.connect(uri) as websocket:
-        # Create an Input message
-        input_message = Input()
-        input_message.reset_match = True
-
-        # Convert the message to JSON
-        json_message = MessageToJson(input_message)
-
-        # Send the message
-        await websocket.send(json_message)
-        print(f"Sent reset command: {json_message}")
+        
+        # Step 1: Reset the match
+        reset_message = Input(reset_match=True)
+        await websocket.send(MessageToJson(reset_message))
+        response = await websocket.recv()
 
-        # Wait for a response
+        # Step 2: Send STOP command
+        stop_message = Input(
+            change=Change(
+                new_command_change=Change.NewCommand(
+                    command=Command(
+                        type=Command.Type.STOP,
+                        for_team=Team.UNKNOWN
+                    )
+                )
+            )
+        )
+        await websocket.send(MessageToJson(stop_message))
+        print(f"Sent STOP command: {MessageToJson(stop_message)}")
         response = await websocket.recv()
-        output_message = Parse(response, Output())
-        print(f"Received response: {MessageToJson(output_message)}")
 
-async def main():
-    await reset_match()
-    print("Reset command sent to SSL Game Controller")
+        print("Reset and STOP commands sent to SSL Game Controller")
 
 if __name__ == "__main__":
-    asyncio.run(main())
+    asyncio.run(reset_and_stop_match())