Action Valuation of On- and Off-Ball Soccer Players Based on Multi-Agent Deep Reinforcement Learning.
Hiroshi NakaharaKazushi TsutsuiKazuya TakedaKeisuke FujiiPublished in: IEEE Access (2023)
Keyphrases
- reinforcement learning
- multi agent
- action selection
- soccer games
- soccer game
- robot soccer
- robocup soccer
- action space
- partially observable domains
- state space
- state action
- reward shaping
- function approximation
- temporal difference
- game theory
- multiagent systems
- cooperative
- multi agent systems
- single agent
- transfer learning
- multi agent environments
- continuous state
- learning agents
- model free
- optimal policy
- transition model
- intelligent agents
- multiple agents
- soccer video
- reinforcement learning agents
- autonomous agents
- traffic signal control
- online game
- human actions
- function approximators
- agent learns
- partially observable
- learning algorithm