On the Potential of Rocket League for Driving Team AI Development
| Autor: | Yannick Verhoeven, Mike Preuss |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Sports analysis
Rocket (weapon) Game mechanics Test case Computer science 020204 information systems Learning environment Control (management) 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing 02 engineering and technology League Data science |
| Zdroj: | SSCI |
| Popis: | Several Games have been used as test cases for new AI developments in the last years, e.g. the games of the Atari Learning Environment, Go, and also StarCraft. As many single player problems are treated successfully now, research is turning more to Team AI problems: AI/human and AI/AI coordinated decision making in mixed cooperative/competitive environments. OpenAI Five and RoboCup and related environments as MuJoCo soccer are pushing into this direction. Rocket League is an interesting alternative as it has quite different game mechanics compared to soccer, and an API with a number of different bots already exists. We summarize the current state, relate it to the situation in the fields that may provide useful approaches as e.g. sports analysis, and argue that it makes sense to do more research on Rocket League bot control and coordination. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|