FAQs.md

FAQs in MetaUrban

🛠 Installation

Windows and Linux

There may be some issues when installing ORCA on Windows that cannot be easily resolved. We highly recommend conducting experiments on WSL2, which has undergone more thorough verIfication.

The installation of MetaUrban requires the following tools:

make, cmake, gcc

Please check the installation of these tools via

make -v
cmake --version
gcc -v

If not installed, use commands on WSL2 and Linux

sudo apt-get update
sudo apt-get install make cmake gcc

MacOS

Please check the installation of these tools via

make -v
cmake --version
gcc -v

If not installed, use commands on MacOS

sudo brew install make cmake gcc

If there is no brew on your PC, use

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

nano ~/.zshrc
export PATH="/opt/homebrew/bin:$PATH"
source ~/.zshrc

to install brew

Additionally, we highly recommend that use bash via

bash

General

If you meet issues about path like

no 'bind' (in /PATH/TO/METAURBAN/metaurban/policy/orca_planner.py)

Try

export METAURBANHOME='/PATH/TO/METAURBAN'

💡 Rendering

Headless mode

If your PC is without monitor, you can only use offscreen rendering mode by setting

use_render=False,
manual_control=False,

and add some specIfic lines in scripts to warm up the environment

from metaurban.envs.base_env import BASE_DEFAULT_CONFIG
from metaurban.engine.engine_utils import initialize_engine, close_engine
config_warmup = BASE_DEFAULT_CONFIG.copy()
config_warmup["debug"] = True
initialize_engine(config_warmup)
close_engine()

before spawning environment

env = SidewalkStaticMetaUrbanEnv(config)
o, _ = env.reset(seed=0)

Rendering pattern

There may be some minor dIfference of rendering qualities among dIfferent systems, which is normal.

Testing rendering results

You can test the cameras via

metaurban/tests/latest_updated/test_semantic_camera_synbody.py

If everything functions properly, you should observe paired images like the example below

🚀 RL training

If you plan to train RL agents on your PC, make sure to adjust the number of environments in the scripts:

n_envs=20

This will help prevent out-of-memory (OOM) errors or high CPU usage. However, due to the inherent characteristics of RL, setting the number of environments too low may compromise the effectiveness of agent training.

🤖 Run a Pre-Trained (PPO) Model

We provide RL models trained on the task of navigation, which can be used to preview the performance of RL agents.

python -m metaurban.examples.drive_with_pretrained_policy

and the target result would be