Elodin
Show HN: Open-Source AI Racing Harness
Last verified:
What is Elodin?
Elodin AI Grand Prix Race Sim Harness is an open-source practice simulator for Anduril's AI Grand Prix autonomous drone-racing competition, a $500K competition hosted by DCL (Drone Champions League). The harness allows competitors to practice and iterate on their autonomy stack before the official competition simulator ships, providing a realistic practice loop that matches the published competition constraints and message format.
The tool runs against real Betaflight firmware, which requires at least 1000 sensor samples per second to run real-time correctly. It includes a custom camera sensor generated directly in the simulation loop using the Rust Bevy game engine, eliminating the need to connect to heavy game engines like Unreal. This approach is optimized for the simple visual needs of the drone racing challenge while enabling quick improvements to visual fidelity.
This harness is designed for engineers, university teams, and independent developers competing in the AI Grand Prix drone race. It addresses the iteration bottleneck that competitors face while waiting for Round 1 virtual qualifier access, enabling them to develop and test autonomy software for fully autonomous drones without hardware modifications or human pilots.
Elodin pricing
Pricing model: Freemium
Free and open-source. The AI Grand Prix harness is available on GitHub under elodin-sys/ai-grand-prix. Elodin platform offers 60 free minutes of hosted Monte Carlo simulation time per month when creating an account at app.elodin.systems. No paid plans are mentioned for the race sim harness itself.
Elodin pros
- Open-source and freely available
- Matches official competition constraints and message format
- Runs against real Betaflight firmware
- Supports 1000+ sensor samples per second real-time
- Built with Rust Bevy game engine for graphics
- Custom camera sensor in simulation loop
- No need for heavy Unreal Engine connection
- Enables iteration before official simulator ships
- Realistic practice loop for drone autonomy
- Based on Elodin physics simulation platform
- GPU-accelerated simulation using XLA and JAX
- Python and Rust API support
- Supports Monte Carlo simulations up to 100,000 simultaneously
- Works on Linux, macOS, and Windows (WSL)
- Facilitates sim-to-real transfer for autonomy software
Elodin cons
- Not the official competition simulator
- Visual fidelity currently basic, not fancy looking
- SDK only supported on Linux and macOS
- Windows requires WSL for simulation server
- Only supports glibc 2.35+ systems
- Limited to autonomous drone racing use case
- Requires Python or Rust programming knowledge
- No official visual assets from DCL competition
Frequently asked questions about Elodin
What is the Elodin AI Grand Prix Race Sim Harness?
It is an open-source practice simulator for Anduril's AI Grand Prix autonomous drone-racing competition. The harness lets competitors practice and iterate on their autonomy stack before the official competition simulator ships, built to match the published competition constraints and message format.
What does Betaflight require to run correctly?
Real Betaflight requires at least 1000 sensor samples per second to run real-time correctly. The harness is designed to meet this requirement for accurate simulation of drone flight control.
How does the camera sensor work in this harness?
The competition warranted introducing a new feature to generate the camera sensor directly in the simulation loop. This eliminates the need to connect to Unreal or similar game engines, which are very heavy. For the simple needs of this challenge, creating samples directly in the loop is handy and easy to use.
What game engine powers the visualization?
The harness uses the Rust Bevy game engine for its graphics. While it is not fancy looking currently, Bevy should allow the team to improve visual fidelity quickly.
Who is this harness for?
It is for anyone competing in the Anduril AI Grand Prix drone race - university teams and independent engineers worldwide who are eagerly awaiting the competition simulation harness and want to practice their autonomy stack beforehand.
Is this the official competition simulator?
No, this is not the official simulator. It is a practice harness to tide competitors over while waiting for the Round 1 virtual qualifier simulation. The official competition simulator will be released separately.
What programming languages does Elodin support?
Elodin currently allows you to write physics simulations in Rust or Python. The software you are testing with simulations can be written in any language. The team plans to add support for more languages in the future.
What platforms can run Elodin?
The SDK is only supported on Linux (glibc 2.35+ including Ubuntu 22.04+, Debian 12+, Fedora 35+) and macOS (including Apple Silicon). Windows users can use Elodin by installing and running the simulation server in WSL (Windows Subsystem for Linux).
What is Monte Carlo simulation in Elodin?
Monte Carlo simulations explore the state space by running massive numbers of tests. Elodin's cloud platform can run up to 100,000 simultaneous Monte Carlo simulations cost-effectively using ECS architecture, IREE compilation, and cloud elasticity. Users get 60 free minutes per month with an account.
What is the AI Grand Prix competition?
The AI Grand Prix is a $500,000 autonomous drone-racing competition operated by DCL (Drone Champions League) for Anduril. It invites university teams and independent engineers to compete for the prize pool and job opportunities at Anduril. The championship race takes place in November 2026 in Columbus, Ohio, using fully autonomous Neros Technologies drones with no human pilots or hardware modifications allowed.