Abstract

RoboCup Rescue is one of the research test beds proposed in the context of the RoboCup international initiative. The challenge is to develop teams of heterogeneous agents that limit building damage and rescue people in disaster scenarios like a post-earthquake situation. The social relevance of this challenge is evident.

The RoboCup Rescue initiative supports four major projects: simulation; robotic and infrastructure; integration; operation. The RoboCup Rescue Simulation is focused on the global strategy of an heterogeneous rescue team to limit the damage in a urban area after a major disaster. This league searches the best rescue team coordination strategies.

The simulator architecture offers a dynamic environment, such as: spread of fire, house and building collapse, civil movement and traffic, disruption of roads and physical status of victims. This project is exactly intended to develop methodologies enabling to build a Simulation RoboCup Rescue Team.

The RoboCup Rescue Simulator implements a virtual urban scenario were agents of several types execute their tasks. Platoon agents are composed of: fire brigades that may extinguish fires, police forces that may clear roads, and ambulance teams that may carry civilians to safe places. Center agents are hierarchically superior to platoon agents and include fire stations, police offices and ambulance stations. After an earthquake several roads are blocked by debris, there are civilians wounded and fires are starting to appear everywhere around the city. The task of the rescue team is to limit city damage and save as many lives as possible.

In order to develop a Rescue agent team several different problems must be addressed. The agent architecture, its basic skills, its decision mechanisms, the communication protocols it supports, the integration of autonomous decision and an hierarchical line of command, etc. In this project we propose to develop a team of RoboCup Rescue agents that has innovative coordination methodologies and is able to integrate learning techniques in its reasoning procedures.

The coordination methodologies that will be used in this project are adaptations and extensions of previously researched coordination framework coming from our experience in FC Portugal team of RoboCup Soccer Simulation League (world champion in RoboCup 2000).

The application of Machine Learning to the RoboCup Rescue domain is one of the objectives of this project. Agents will be based on a hybrid, hierarchical behavior based agent architecture that allows for the integration of learning. This new architecture provides flexible adaptation of agents in unknown and dynamic environments, enabling agents to build the most intelligent strategic plan to rescue citizens, for example.

Project Tasks

1. Formalization of a Rescue Team’s Strategy

2. Agents’ Architecture and Knowledge

3. Homogeneous Agents Teams’ Strategy

4. Agents’ Perception and Action

5. Coordination Infrastructure

6. Advanced Communication Methodologies

7. Center Supervision Methodologies

8. Application of Machine Learning

9. Integration and Evaluation

Results and Repercussions

Research conducted in FC Portugal Rescue project may be partially applied to several very useful social areas, like fire combat, battlefield scenarios, mine clearance, land exploration, control of hospital robots, public transport coordination, satellite control, cleanup of radioactive and toxic contamination and other problems that imply team coordination.

Expected results of FC Portugal include:

These results will be made available through: