This is a sneak peek! I have been inspired by projects like Hide and Seek by OpenAI as well as Capture the Flag by DeepMind, specifically the team based competitive aspect in both projects. In the future I want to work on how agent’s that have competitive goals can figure out to collaborate temporarily, but for now I just want to look at collaboration within predefined competitive teams. The setup is an “Amazon Air Delivery” szenario, where multiple teams of drones from different companies “fight” for packages for delivery across a city. The team with the most deliveries wins. There are multiple steps to solve this problem. First I have been looking at replacing the typical drone controller called PID (proportional-integral-derivative) with a Neural Network, which is shown below. Secondly there will be procedurally generated city scapes to replace the block environment. Thirtly multiple teams of agents will be implemented to collect and deliver packages.