AlpaSim: A modular, lightweight, and data-driven research simulator for autonomous driving

What is AlpaSim?

AlpaSim is an open-source autonomous vehicle simulation platform designed specifically for research and development. It allows users to test end-to-end AV policies in a closed-loop setting by simulating realistic sensor data, vehicle dynamics, and traffic scenarios within a modular and extensible testbed.

Suitable use cases include:

Algorithm Validation: Test new autonomous driving algorithms in realistic environments
Safety Analysis: Evaluate vehicle behavior in edge cases and challenging scenarios
Performance Benchmarking/Regression Testing: Compare different models and configurations
Debugging: Understand and debug complex autonomous driving behaviors

Sensor Fidelity

Pluggable renderer service with default NuRec support and OmniDreams video-model rendering through FlashDreams
High-fidelity camera feeds with configurable field-of-view, resolution, and frame rates
Stateful video-model rendering for stronger dynamic-object and non-rigid visual fidelity
Realistic sensor noise and environmental conditions

Research Hackability

Python-based implementation built for rapid prototyping and experimentation
Modular grpc interface design allows researchers to swap out components with custom implementations
Extensive configuration options and debugging tools

Horizontal Scalability

Microservices architecture enabling distributed computing
Scale individual components for optimal load balancing
Support for multi-node deployments

To learn more about the design principles and architecture, check out the system design docs.

Driving Policies

AlpaSim currently supports the following driver policies:

Alpamayo-R1 - NVIDIA Alpamayo, a VLA driving policy with chain-of-causation reasoning
Alpamayo 1.5 - Public Alpamayo 1.5 checkpoint support is also available
VaVAM - an autoregressive video-action driving policy
Transfuser - Latent TransFuser v6 (LTFv6) policy developed for NAVSIM (provisional)

Stay tuned for additional model support. Contributions from the community are appreciated.

Support

📣 Usage questions and discussion about AlpaSim: please join us on the Alpamayo NV Developer Forum.

🐛 Code-level bugs, documentation issues, and feature requests: file a GitHub issue using the appropriate template (Bug report, Documentation request, or Feature request). The relevant NVIDIA responder is auto-assigned via the assignees: field on the template.

🚨 Security vulnerabilities: please use NVIDIA's Vulnerability Disclosure Program. Do not file security issues publicly here.

Getting Started

To run simulations locally (Docker Compose, single machine), see the Tutorial. The default tutorial path uses NuRec; for OmniDreams as the renderer backend, see the Video Model Renderer guide. For cluster or SLURM deployment, see src/tools/run-on-slurm.

Documentation & Resources

Onboarding Guide: Initial setup and access instructions
Tutorial: Step-by-step usage guide
Manual Driver: Interactive keyboard control of the ego vehicle
Video Model Renderer: Running OmniDreams through FlashDreams as the AlpaSim renderer backend
Operations Guide: Performance tuning, configuration, and troubleshooting
Telemetry Guide: Prometheus metrics, persistence, and central discovery
Data Pipeline: ASL log format and reading logs
Design Documentation: Technical architecture and design decisions
Plugin System: Extending AlpaSim with custom models, configs, and tools
API Reference: gRPC API documentation
Changelog: Major updates and migration notes
Contributing: Code style, coordinate conventions, and PR process
Developer Quick Reference: Build commands, testing, and CI workflow
Test Suites & Scenes: Available scene datasets

Sample Data

Hugging Face Dataset: PhysicalAI-Autonomous-Vehicles-NuRec
Sample Artifacts: Included in the repository via Git LFS

Contributing

We welcome contributions from the research community! Please see our Contributing Guide for details on:

Code style and conventions
Testing requirements
Pull request process
Development setup

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

Citation

If you use this software, please cite it as follows:

@software{alpasim_2025,
  author       = {
    NVIDIA and
    Yulong Cao and
    Riccardo de Lutio and
    Sanja Fidler and
    Guillermo Garcia Cobo and
    Zan Gojcic and
    Maximilian Igl and
    Boris Ivanovic and
    Peter Karkus and
    Janick Martinez Esturo and
    Marco Pavone and
    Aaron Smith and
    Ellie Tanimura and
    Michal Tyszkiewicz and
    Michael Watson and
    Qi Wu and
    Le Zhang
  },
  title        = {AlpaSim: A Modular, Lightweight, and Data-Driven Research Simulator for Autonomous Driving},
  year         = {2025},
  month        = {October},
  url          = {https://github.com/NVlabs/alpasim},
}

Project Contributors:

Contributors in each topic in alphabetical order

Project Lead: Maximilian Igl

Tech Leads: Michal Tyszkiewicz, Michael Watson

Architecture Design & Networking: Michal Tyszkiewicz

Open Sourcing: Guillermo Garcia Cobo, Maximilian Igl, Peter Karkus, Ellie Tanimura, Michael Watson

Infrastructure & Wizard: Maximilian Igl, Aaron Smith, Michal Tyszkiewicz, Michael Watson, Qi Wu (SLURM deployment), Le Zhang (Data management)

Runtime: Maximilian Igl, Aaron Smith, Ellie Tanimura, Michal Tyszkiewicz, Michael Watson

CICD: Maximilian Igl, Aaron Smith

Data Pipeline: Riccardo de Lutio, Janick Martinez, Le Zhang

Product Manager: Matt Cragun

Testing & debugging: Guillermo Garcia Cobo, Peter Karkus, Ellie Tanimura

Service Modules:

Driver integration: Maximilian Igl, Peter Karkus, Michal Tyszkiewicz
Evaluation: Yulong Cao, Maximilian Igl
Controller: Michael Watson
Physics: Riccardo de Lutio
Trafficsim: Maximilian Igl, Boris Ivanovic

Senior Mgmt: Sanja Fidler, Zan Gojcic, Boris Ivanovic, Marco Pavone

Acknowledgements for additional contributions: Fabian Barajas, Kashyap Chitta, Ankit Gupta, Laura Leal-Taixe, Nicole Yang

Built for researchers, by researchers
Accelerating autonomous vehicle development through realistic simulation

alpasim