ACORD - Adaptative Coordination of Robotic Teams
This project aims at developing coordination and communication methodologies to enable teams of heterogeneous robots, with different skills, and made by different manufacturers, to accomplish complex collective tasks, with emphasis on playing robosoccer games.
Project Title: ACORD - Adaptative Coordination of Robotic Teams (PTDC/EIA/70695/2006)
Duration: 24 Months (January 2007 - December 2009)
Funding Entity: FCT/POSI
Partners: University of Aveiro, FEUP, IEETA
Researchers: Luis Paulo Reis (Principal Investigator), Nuno Lau, António Paulo Moreira, José Luís Azevedo, André Conceição, Armando Sousa, Artur Pereira, José Manuel Vieira, Luis Miguel Almeida, Luís Mota, Luís Seabra Lopes, Manuel Bernardo Cunha, Paulo Costa, Paulo Pedreiras, Pedro Faria, Pedro Fonseca, Rodrigo Braga
This project aims at developing coordination and communication methodologies to enable teams of heterogeneous robots, with different skills, and made by different manufacturers, to accomplish complex collective tasks, with emphasis on playing robosoccer games. The main result of ACORD project will be a new approach to cooperative robotics in general and to building RoboSoccer teams in particular.
The project main goals are twofold: (i) to design a software common framework suitable for the implementation of a agent-based software that may control different robots to be used for several cooperative robotic tasks; (ii) to develop coordination and communication methodologies that enable teams of heterogeneous robots, built and programmed by different entities, to accomplish complex collective tasks. In particular, these methodologies will be applicable in any Robotic Soccer league (simulation 2D, simulation 3D, small-size and middle-size).
The agent-based software may control each one of the robots of the team, regarding its action and perception capabilities and the cooperative task to perform, enabling different robots to directly cooperate to fulfill any collective task. However, other robots with different software and hardware architecture will also be able to join the team by complying with the high-level communication and coordination protocols defined in the project.
In order to control different robots, the common framework software agent needs specific parts that must be built to deal with special constraints related with the agent’s perception and action capabilities. Low-level skills will be designed for each type of robot, although their specific high-level decisions will be derived through the common framework. Dynamical models adjusted to each type of robot and actuator will be built along with efficient, precise and robust control systems for robot safe mobility. A general language of action has also to be developed to enable the specific action modules to understand high-level decision making. The project also includes the construction of specific perception modules that deal with image processing and analysis. Novel types of sources of perception, such as range-finders or compasses may be evaluated during the course of this project. A special perception language will enable the communication of perception information from these and other modules to the common framework agent. Each agent also includes a dynamic world state model that will be updated through perception, robot communication and action prediction.
In ACORD project, both inter-agent cooperation and team coordination policies will be investigated. This will include cooperation methods like dynamic positioning and role exchange and strategic positioning. Also intelligent communication protocols and methodologies to communicate coordination-related knowledge will be researched. Special supervisor agents will be developed for each RoboCup league or new set of cooperative tasks to perform, capable to communicate with the common framework agents via a supervisor language based on high-level concepts like tactics, formations, situations, roles, etc.
The very idea of ACORD project is then to design an agent-based common framework and coordination and communication mechanisms applicable to fulfill different cooperative robotic tasks, including playing in all major RoboCup soccer leagues. In order to test our approach, we will develop robotic/software platforms that conform to RoboCup different leagues’ rules and standard environments for each RoboCup league, including special sensors to enable better robot performance analysis, and test our approach in RoboCup 2007 and 2008 international competitions. This way, we plan to give continuity to successful Portuguese participations (robotic teams developed by this proposal participants) in RoboCup. In recent competitions, FC Portugal (Univ.Aveiro/FEUP) became European and World champion of RoboCup 2006 – Simulation 3D league and 5DPO (FEUP) became European champion and World vice-champion in RoboCup 2006 - Small-Size league. [More Info]
The objective of ACORD is to develop a common agent-based framework for controlling cooperative teams of mobile robots and coordination mechanisms to enable teams of heterogeneous robots to accomplish complex tasks in dynamic environments. Specific objectives are, to develop:
- Knowledge representation structures for cooperative knowledge in all robocup leagues;
- Communication languages between the common framework and specific action and perception modules;
- High-level communication and supervision languages for robotic teams performing complex tasks;
- Cooperative decision-making mechanisms both at individual and team level;
- Agents capable of supervising different teams to fulfill different cooperative tasks.
ACORD main contributions will be twofold: (i) a configurable framework flexible enough to deal with perception, decision-making and action for different (both real and virtual) cooperative teams of robots; (ii) Coordination and communication mechanism enabling heterogeneous robots built and programmed by different entities to work together as a team. [More Info]