Overview
What this challenge is about.
Pick an open multi-agent environment (PettingZoo's MPE 'simple_spread', Overcooked-AI, or SMAC). Implement or wrap three methods: IPPO (independent PPO per agent), MAPPO (centralized critic, decentralized actors), and a single-agent baseline that controls all agents. Run with 3, 5, and 8 agents; 5 seeds per condition. Report mean cooperative-return + 95% CI; analyze how each method scales with agent count. Write a 4-page workshop-style report with one striking plot.
The Brief
What you'll do, and what you'll demonstrate.
Benchmark cooperative MARL methods (IPPO, MAPPO, monolithic) across agent counts with proper statistics and write the workshop report.
Earning criteria — what you'll demonstrate
- Implement CTDE and fully-decentralized MARL methods
- Run a fair MARL benchmark with proper statistics
- Analyze how methods scale with agent count
- Write a workshop-style research report
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Research Scientist
Running a multi-seed MARL benchmark with workshop-style writeup is the rigor expected of a junior research scientist on a multi-agent research team.
This challenge sharpens
- multi-agent-reinforcement-learning
- experiment-design
- scientific-writing
ML Researcher
Comparing CTDE vs decentralized methods with proper statistics is the applied ML-research work that agent-research teams hire for.
This challenge sharpens
- multi-agent-reinforcement-learning
- ppo
- statistical-testing
Applied AI Scientist
Knowing the scaling story of MARL methods is the applied-AI skill that translates multi-agent research into deployable cooperative systems.
This challenge sharpens
- pytorch
- multi-agent-reinforcement-learning
- experiment-design