splatsim

3D Gaussian Splatting based simulator with DDS and CARLA integration.

Prerequisites

• Python 3.10+
• CUDA 12.4
• uv
• (Docker builds) Docker with BuildKit and NVIDIA Container Toolkit

Installation

Local

# Core only (headless)
uv sync

# With GUI viewer
uv sync --extra gui

# With DDS support
uv sync --extra dds

# With CARLA support (includes DDS)
uv sync --extra carla

# All optional dependencies
uv sync --extra all

CARLA extra requires system libraries for lanelet2. On Ubuntu 22.04:

sudo apt-get install -y \
  libboost-dev libboost-serialization-dev libboost-filesystem-dev \
  libboost-program-options-dev libboost-python-dev libboost-system-dev \
  libeigen3-dev libpugixml-dev libgeographic-dev librange-v3-dev python3-dev

Docker

The Dockerfile uses a multi-stage build with the dds extra to keep the image minimal.

# Build
docker buildx build \
  -f docker/Dockerfile \
  -t splatsim .

# Run (requires NVIDIA Container Toolkit)
docker run --rm -it --gpus all splatsim

To customize CUDA or Ubuntu versions:

docker buildx build \
  -f docker/Dockerfile \
  --build-arg CUDA_VERSION=12.4.1 \
  --build-arg UBUNTU_VERSION=22.04 \
  -t splatsim .

Usage

# Launch the viewer
splatsim-viewer

# Run a scenario
spawn-scenario

Development

uv sync --dev
pre-commit install