Update Docker Environment for Assignments

[j3soon]

Hi @author31, as in our previous offline discussion, we plan to port the assignments in ai-capstone26 into this repo by switching to Isaac Sim.

You mentioned the first step is to have a docker environment with Ubuntu 24.04, Isaac Sim 5.1.0, and ROS 2 Jazzy, and the docker container should be runnable on both local (Ubuntu) and on the cloud via NVIDIA Brev.

The following is a current draft Brev launchable and its accompanying docs:

and docs, can be tested using our org.

I plan to open a more complete PR for the dockerfile (adapted for this course), along with docker compose file, and minimal step-by-step instructions to this repo, a few things need confirmation:

• Do we need CUDA/PyTorch? If so, what versions are preferred?
  Below is the previous dockerfile used in this repo.
  

  aicapstone/Dockerfile

  Line 2 in c6f0c0d

  FROM nvidia/cuda:12.8.1-devel-ubuntu22.04

• Do we need Isaac Lab? If needed, is v2.3.2 suitable?
  Below is the Isaac Lab version (v2.3.2) in previous dockerfile:
  https://github.com/HCIS-Lab/aicapstone/tree/main/dependencies
• Isaac Sim 6.0.0 is released a while ago, do we want to use the latest version? or stick to v5.1.0?
• Any other dependencies?
  Below are some dependencies in the previous dockerfile.
  

  aicapstone/Dockerfile

  Lines 86 to 99 in c6f0c0d

  	RUN python -m pip install --upgrade pip \
  	&& python -m pip install -U \
  	torch==2.7.0 \
  	torchvision==0.22.0 \
  	--index-url https://download.pytorch.org/whl/cu128 \
  	&& python -m pip install --upgrade \
  	"isaacsim[all,extscache]==5.1.0" \
  	--extra-index-url https://pypi.nvidia.com \
  	&& python -m pip install \
  	pip==23 \
  	setuptools==65 \
  	flatdict==4.0.0 \
  	huggingface-hub==0.35.3 \
  	transformers==4.57.6

• Currently, I only briefly tested the environment, and it would be helpful if we have a real project for more thorough testing.
  Do we have a Isaac Sim/Lab code/workflow suitable for testing? Or should I take an existing example scene to test further?
• Where should I put the dockerfile, compose file, and docs/screenshots? This is needed for the PR.

Thank you!

[author31]

Hi @j3soon, thanks for following up.

Since the updated target machines for these assignments are consumer-grade gaming laptops, with the minimum GPU being an RTX 5060 with 8GB of VRAM, my main concern is whether the ROS 2 + Isaac Sim stack can run properly on this hardware configuration. Could you please help me confirm this?

Previously, I considered using only the Isaac Lab stack to implement the robot control stack. More information can be found in scripts/hw1.py in this PR: #46. Could you please help me review whether this stack is feasible?

Besides that, I’m currently looking for a regular consumer-grade laptop to test whether the workload fits well and whether students’ hardware can run the assignment. May I ask for your opinion on whether an RTX 5060 with 8GB of VRAM would be sufficient?

[j3soon]

Hi,

I would suggest at least 16GB of VRAM, as running your hw1.py alone consumes 6~7GB VRAM.

In addition, I just tested the script, and can confirm it works on my side using Brev: , will add docs later.