VAIL-UCLA

VAIL-UCLA / MetaDrive-Arena

Public

Multi-agent racing tournament on MetaDrive

18
4
100% credibility
Found Mar 31, 2026 at 18 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

MetaDrive Arena is an online evaluation platform where students upload AI racing agents to compete in simulated 1v1 races on various tracks, with live leaderboards, video replays, and automated matchmaking.

How It Works

1
🏁 Discover the racing challenge

You hear about an exciting AI racing competition in class where you train smart cars to race against classmates.

2
📧 Get your private login

The teacher sends you a special code to join the competition safely.

3
🚀 Log in and explore

Enter your code on the website to see the leaderboard and get ready to compete.

4
📤 Upload your racing bot

Share your trained smart car so it can race others automatically.

5
⚔️ Watch the races unfold

Your bot queues up for 1v1 duels on twisty tracks with video replays.

6
📊 Check the live leaderboard

See win rates, speeds, and rankings update as matches finish.

🥇 Celebrate your victories

Climb to the top, watch highlight reels, and enjoy the thrill of outracing everyone.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 18 to 18 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is MetaDrive-Arena?

MetaDrive-Arena is a Python-based server for hosting multi-agent racing tournaments in the MetaDrive simulator. Instructors spin up a FastAPI web app where students upload zipped RL agents—think PPO or SAC policies trained on lidar observations—and it auto-matches them in 1v1 races across tracks like circuit, oval, or serpentine, posting results to a live leaderboard with ELO, win rates, and video replays. CLI tools handle user rosters, tokens, and round-robin tournaments, turning class projects into competitive arenas.

Why is it gaining traction?

Unlike raw multi-agent GitHub code repos or platforms like LangGraph setups, this delivers turnkey evaluation: upload triggers races against everyone, GPU-scaled workers handle dozens concurrently, and BEV/3D videos plus highlights (fastest laps) make debugging intuitive. Custom maps test generalization, and student limits prevent abuse, hooking educators tired of manual sim runs.

Who should use this?

RL profs running multi-agent courses, especially autonomous driving or racing sims, needing a drop-in tournament backend. TAs grading 50+ PPO/SAC agents without babysitting evals. Multi-agent GitHub devs prototyping Claude-style or Copilot-assisted policies in a competitive loop.

Verdict

Grab it for classrooms—starter code and admin tools make setup dead simple, despite 18 stars and 1.0% credibility signaling early maturity. Polish docs for broader adoption, but it crushes niche use cases today.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.