When it comes to robots playing soccer, bipedal robots are used in most competitions. Robot dogs can also play the game as this video from Hybrid Robotic shows. Their latest work shows a multi-agent reinforcement learning (MARL) framework that allows robot dogs to play soccer autonomously.
As the researchers explain, first, walking, dribbling, and kicking are trained. On top of that, ” a high-level strategic planning policy is trained with Multi-Agent Proximal Policy Optimization (MAPPO) via Fictitious Self-Play (FSP).”
This approach brings more complex behavior to robots, such as coordinated passing and interception.
[HT]