The game playing agent consists of different moduls. And we use the provided components:
- the vision component, that analyzes the video frame
- the trajectory component , that calculated the trajectory and targeting
- the game playing component, orchestrates the game playing
These three components are written in Java and will only be modified / updated if time allows (probably not). The modules for the strategy on level selection with respect to competing agents. time lect in competition (competition strategy) and for shot selection to clear the level with maximum score are currently developed in Python. We do at the moment not plan to participate in the novelty track, i.e. the agent will not have to be retrained for a particular new set of levels with unknown items and challenges. Well we might participate, just to see how the agent performs in the new settings if competition rules allow that.
Not all details on integrations are decided.
Team members will be provided with instructions on how to get started and access to a common repository.