So this is a copy of Carla Simulator 0.8.2 PythonClient folder.
This is part of a project at Blue Horizon AI to make an autonomous wheelchair. This simulator environment is being created to set up an environment for collecting data, training and testing autonomous wheelchairs.
On a fresh download Carla has 2 scripts that allow the simulator to run in server mode.
-
manual_example.py - This script allows the simulator to be controlled through manual controls
-
client_example.py - This script allows the images from the simulator to be saved
I wanted to have a script that allowed me to control the simulator manually and save the data as I drove.
The main script that is being made is Manual-Control-Data-Collection.py
This script is based off the manual_example.py
Current Features
- Manual control through Xbox controller
- Saving image data as a .jpg file to a folder
- Store the steering angle, throttle and other values such as timestamp to a csv file
- Record data and replay that data to ensure that the recorded data is accurate for imitation learning
- Allow the simulator to be controlled in realtime through an autonomous model (using .h5 file and .json file)
Future Features
- Have the car model be replaced with a wheelchair
- This will be done on Unreal Engine 4 editor so this github wont be updated with that feature
HOW TO USE
- Install Carla simulator 0.8.2
- Copy the Manual-Control-Data-Collection.py file into your PythonClient folder
- Mulitple changes may need to be done to ensure that it works on your computer
- filepaths, controller setup and a couple modules
- The version which allows models to be used requires a miniconda environment (If you want it message me on github)
- Run this file as you would run client_example.py following the documentation on Carla wiki for stable branch (8.2)
- Some documentation is added which is missing in the manual_control.py - The comments I have added may help you change manual_control.py to what you need it to do