Air Traffic Controller

Project Description

ATC was a simulation game designed to help new Air Traffic Controllers learn how to communicate effectively in a low stakes environment. The game works by spawning planes each with an objective, and it is up to the Air Traffic Controller to get each plane resolved, before too much backup spawns. The game utilized AI pilots which could communicate based on what the ATC was saying to them.

My Roles

• AI Synthesized responses

  • The AI is using the ChatGPT API and when the player speaks to them, using a speech to text plugin, that input is sent to ChatGPT along with a prompt, explaining how the pilot can respond, and does so. 

• Text to Speech

  • The AI takes its response and uses that text, along with the selected voice mode on the pilot class, and sends that back to OpenAI using its text to speech endpoint, and plays the resulting audio file.

• Allowed AI to take action based on what the player says to them

  • For example when the player states that a plane is cleared for takeoff, the AI was prompted to place a code at the start of their next response, that is parsed out and used to run the takeoff actions on the actual planes in game.

• Grading System

  • After a plane is successfully resolved, a new prompt is sent to the ChatGPT endpoint and asks for a short grading out of 100 formatted in a certain way, and a brief description of how well the ATC communicated to the pilot. This is displayed to the user, and the grading is used for a score within the game