API for connecting CARLA-GUI to real-world simulator

This page introduces the method for the CARLA-GUI to receive throttle, steer and brake commands from external I.O. devices (e.g. a real-world simulator).

---

HumanEgoControlClient

This is the client class providing the interface for connecting the external I.O. device with the CARLA-GUI using python-socket

Method

• apply_ego_commands(self, throttle = 0.0, steer = 0.0, brake = 0.0)
  add the ego vehicle to the init section

  • Parameters

    • throttle : float, optional
      throttle of the vehicle, within [0.0,1.0]

    • steer : float, optional the steer of the vehicle, within [-1.0,1.0]

    • brake : string, optional
      the brake of the vehicle, within [0.0,1.0].

---

Demo Code Recipe

Here is a short code recipe showing the usage of the client class. The code is also available at test_client.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Tue Aug 11 11:35:37 2020
@author: shijiliu
"""


import sys
sys.path.append("..")

from backend.human_ego_control import HumanEgoControlClient

def main():
    client = HumanEgoControlClient()
    for ii in range(1000):
        client.apply_ego_commands(throttle = 0.5 + (ii % 10) * 0.01, steer = 0.0, brake = 0.0)

if __name__ == '__main__':
    main()

---

Notification

The code recipe above should be easy to understand and use. Here lists several issues that may occur when you are creating your own project.

• Import issue
  Here is a skeleton drawing showing the file structure of the CARLA-GUI project

   - CARLA-GUI package
         \   
        - gui-interface
        |    \ 
        |     test_client.py (code recipe above)
        |
        |
        |
        - backend
        |    \
        |     human_ego_control.py (containing the HumanEgoControlClient class)
        |
        - your-package
  

Import HumanEgoControlClient class using the method described in the code recipe above will only be successful if the script is put inside the gui-interface package or back-end package or any package in parallel with these packages. If you decide not putting your package in parallel with the backend, please remember to download the human_ego_control.py and put it in your own directory. Since we are using socket here, the directory won't affect the usefulness of the code.

• Timing issue If user choose to drive the ego vehicle, the backend will be waiting for the human command at reach update step. During the waiting time, the whole simulation will be paused. To ensure the simulation to be realistic, please make sure your code is functioning fast enough. (The back-end loop is in theory updating each 0.02s, or 50 Hz.)

---

author: shijiliu

date: 2020-08-11

email: shijiliu@umich.edu

---