Multi-Agent Debugging and Monitoring Framework (original) (raw)

In this paper we present a framework developed for the CAMBADA Middle-sized league robotic team, which allows human developers to better understand the robots actions during a game. Robotic soccer teams are in their nature dynamic multi-process and multi-agent systems, and knowing what is happening in all processes running on the agents at the same time is a hard task. To accomplish this task we developed a framework to create log files, one per process, and to interlace them later. The logs represent robot's knowledge. The framework allows the synchronization and visualization of logs and videos. Videos give the actual real behaviors. This will allow us to understand the robot's reasoning. A GUI utility to navigate and search inside log files was also developed.