Closed GoodarzMehr closed 1 year ago
Thanks for this great PR! This is really useful. I will carefully review it and merge.
I tried the PR; it works well! I will add a document to tell users how to build up the docker. Good work!
You're very welcome, happy to help!
This PR adds a Dockerfile that performs all the necessary tasks regarding the generation of a generic OpenCDA Docker image, including installation of CARLA and its additional maps, perception components (PyTorch and YOLOv5), and SUMO. The user has the option of whether to install additional maps, perception components, and SUMO when building the image.
I have tested it on a system running a fresh installation of Ubuntu 22.04.2 with AMD Ryzen 9 3900X and RTX 2080 Ti. Apart from a few minor modifications (np.int and np.float being deprecated), examples involving PyTorch worked fine. SUMO does not seem to work however - spitting out an error regarding traci - even though the Dockerfile follows OpenCDA's installation instructions. This may be an issue with the version of SUMO that's being installed.