modified repo

CoPO

@@ -0,0 +1 @@
+Subproject commit 86975e924df9f7d0bee701955228e6855ae3f9ff

README.md

@@ -6,20 +6,31 @@ wykorzystując algorytmy **wieloagentowe** (MA), na co
 najmniej trzech różnych mapach. Omów otrzymane wyniki oraz zwizualizuj działanie
 wytrenowanych agentów.
 
-## How to run:
 
-### Install metadrive 
+## Setup
 ```
-pip install metadrive-simulator
+conda create -n copo python=3.7
-python -m metadrive.pull_asset
+conda activate copo
+
+# Install MetaDrive version 0.2.5
+pip install git+https://github.com/metadriverse/metadrive.git@releases/0.2.5
+
+pip install torch
+
+git clone https://github.com/decisionforce/CoPO
+cd CoPO/copo_code
+pip install -e .
+
+pip install -U ray==1.2.0 "ray[rllib]==1.2.0"
+pip install -U "numpy<1.24.0"
+pip uninstall opencv-python
+pip uninstall opencv-python-headless
+pip install opencv-python==4.5.5.64
+pip install pydantic==1.9.0
 ```
 
-#### Verify
+## How to train a RL agents
-`python -m metadrive.examples.profile_metadrive`
 
-Install swig, python and pip
 
-Install other libraries required by program:
 
-`pip install -r requirements.txt`
 

script/main.py

@@ -1,83 +0,0 @@
-#!/usr/bin/env python
-"""
-This script demonstrates how to setup the Multi-agent RL environments.
-
-
-Usage: python -m metadrive.examples.drive_in_multi_agent_env --env pgma
-
-Options for --env argument:
-    (1) roundabout
-    (2) intersection
-    (3) tollgate
-    (4) bottleneck
-    (5) parkinglot
-    (6) pgma
-
-"""
-import argparse
-from metadrive.component.sensors.rgb_camera import RGBCamera
-from metadrive import (
-    MultiAgentMetaDrive, MultiAgentTollgateEnv, MultiAgentBottleneckEnv, MultiAgentIntersectionEnv,
-    MultiAgentRoundaboutEnv, MultiAgentParkingLotEnv
-)
-from metadrive.constants import HELP_MESSAGE
-from metadrive.policy.idm_policy import ManualControllableIDMPolicy
-
-if __name__ == "__main__":
-    envs = dict(
-        roundabout=MultiAgentRoundaboutEnv,
-        intersection=MultiAgentIntersectionEnv,
-        tollgate=MultiAgentTollgateEnv,
-        bottleneck=MultiAgentBottleneckEnv,
-        parkinglot=MultiAgentParkingLotEnv,
-        pgma=MultiAgentMetaDrive
-    )
-
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--env", type=str, default="roundabout", choices=list(envs.keys()))
-    parser.add_argument("--top_down", "--topdown", action="store_true")
-    args = parser.parse_args()
-    env_cls_name = args.env
-    extra_args = dict(film_size=(800, 800)) if args.top_down else {}
-    assert env_cls_name in envs.keys(), "No environment named {}, argument accepted: \n" \
-                                        "(1) roundabout\n" \
-                                        "(2) intersection\n" \
-                                        "(3) tollgate\n" \
-                                        "(4) bottleneck\n" \
-                                        "(5) parkinglot\n" \
-                                        "(6) pgma" \
-        .format(env_cls_name)
-    env = envs[env_cls_name](
-        {
-            "use_render": True if not args.top_down else False,
-            "crash_done": False,
-            "sensors": dict(rgb_camera=(RGBCamera, 400, 300)),
-            "interface_panel": ["rgb_camera", "dashboard"],
-            "agent_policy": ManualControllableIDMPolicy
-        }
-    )
-    try:
-        env.reset()
-        # if env.current_track_agent:
-        #     env.current_track_agent.expert_takeover = True
-        print(HELP_MESSAGE)
-        env.switch_to_third_person_view()  # Default is in Top-down view, we switch to Third-person view.
-        for i in range(1, 10000000000):
-            o, r, tm, tc, info = env.step({agent_id: [0, 0] for agent_id in env.agents.keys()})
-            env.render(
-                **extra_args,
-                mode="top_down" if args.top_down else None,
-                text={
-                    "Quit": "ESC",
-                    "Number of existing vehicles": len(env.agents),
-                    "Tracked agent (Press Q)": env.engine.agent_manager.object_to_agent(env.current_track_agent.id),
-                    "Keyboard Control": "W,A,S,D",
-                    # "Auto-Drive (Switch mode: T)": "on" if env.current_track_agent.expert_takeover else "off",
-                } if not args.top_down else {}
-            )
-            if tm["__all__"]:
-                env.reset()
-                # if env.current_track_agent:
-                #     env.current_track_agent.expert_takeover = True
-    finally:
-        env.close()

script/requirements.txt

@@ -1,4 +0,0 @@
-wheel
-metadrive-simulator
-gymnasium
-gym