Scaling multi-agent reinforcement learning to full 11 versus 11 simulated robotic football
Problem
Simulated RoboCup football teams still rely on handcrafted strategies, and reinforcement learning had not been scaled to the full 22-player game.
Why it matters
Robotic football is described as a long-standing grand challenge in artificial intelligence.
What the AI did
Agents learned football strategies from tabula rasa via self-play, using an improved multi-agent adaptation of PPO on a custom 2D football simulator faster than the RoboCup simulator.
Results
Agents learned competent football strategies in a full 22-player setting using limited compute (one GPU and one CPU) from tabula rasa through self-play.
Impact
Demonstrates full-game MARL is possible under limited compute; no match-win benchmarks claimed.
Record
| Venue | Autonomous Agents and Multi-Agent Systems |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Andries Smit, Herman A. Engelbrecht, Willie Brink, Arnu Pretorius |
| Citations | 13 (OpenAlex, at import time) |
| References | 35 |
| Model or method | PPO (multi-agent adaptation with policy and critic networks) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1007/s10458-023-09603-y |
| Open access | hybrid |
| Keywords | multi-agent RL, PPO, self-play, RoboCup, robotic football |
Importance rating
incremental / niche
Abstract states robotic football is a long-standing grand challenge in AI, but the addressed problem is handcrafted strategies in simulated RoboCup leagues; a research benchmark with no stated real-world scale.
More in Robotics & Autonomy
| # | Paper | Year | Citations | Importance |
|---|---|---|---|---|
| 1 | Autonomous robotic laparoscopic surgery for intestinal anastomosis | 2022 | 384 |
7
|
| 2 | Solving Rubik's Cube with a Robot Hand | 2019 | 569 |
6
|
| 3 | PaLM-E: An Embodied Multimodal Language Model | 2023 | 346 |
6
|
| 4 | Assistive Navigation Using Deep Reinforcement Learning Guiding Robot With UWB/Voice Beacons and Semantic Feedbacks for Blind and Visually Impaired People | 2021 | 48 |
6
|
| 5 | Towards Deep Radar Perception for Autonomous Driving: Datasets, Methods, and Challenges | 2022 | 154 |
6
|