Teamwork Scenario Editor

Recording the teamwork of embodied agents results in relatively large data collections. Identifying and extracting representative examples from such large database can be a tedious and time consuming knowledge engineering task. For instance, one has to identify which team members contribute to the team behavior as well as where the representative example of interest starts and ends in time. In addition, the same teamwork behaviors can be performed by different team constellations further complicating the identification process. We’ve developed an interactive application for the editing and manipulation of recorded observations which simplifies the identification and extraction of representative examples.

The workflow of the application is inspired from video editing applications, but instead of a single stream of video information, it allows the editing of the observation streams coming from a large number of agents. As the geographical area covered by the observations can be very large (about 100 square kilometers in some of our datasets), the applications allows us to zoom in to a specific area, select the agents of interest and replay specific actions at various speeds or step by step. The application allows the tracking of the trajectories of the selected agents and export the movement traces. Using the application we can find and isolate the representative examples, which can be exported for direct input to training algorithms.



Teamwork Scenario Editor (Screen-shot) Teamwork Scenario Editor (Screen-shot)

Download current development version:

Previous version: