Personal tools
You are here: Home Member Resources info NIAD&R Info Multi-Agent teams’ coordination and simulation

Multi-Agent teams’ coordination and simulation

Coordinating teams of autonomous (or semi-autonomous) agents that perform in rich, dynamic, both cooperative and adversarial environments is a major aim of this work line. For this objective, we are exploring several research directions that can be seen as complementary: new coordination protocols; methodologies for analyzing team behavior; implementation of agent-based common framework suitable for controlling teams of cooperative robots for robosoccer; design of realistic multi-agent simulators (Coastal Ecosystems Simulator); generalizating these methods to other domains.

Researchers:

L. Reis, A. Pereira, F. Reinaldo, L. Mota, P. Faria, R. Braga, D. Silva, P. Abreu, P. Valente, scholarship holders.

Coordinator:

Luís Paulo Reis

Research direction:

Coordinating teams of autonomous (or semi-autonomous) agents that perform in rich, dynamic, both cooperative and adversarial environments is a major aim of this work line. For this objective, we are exploring several research directions that can be seen as complementary: new coordination protocols; methodologies for analyzing team behavior; implementation of agent-based common framework suitable for controlling teams of cooperative robots for robosoccer; design of realistic multi-agent simulators (Coastal Ecosystems Simulator); generalizating these methods to other domains.

Agent-based Simulation

Research goal:

 We have being pursuing the development of realistic agent-based simulators. In this context several simulators were developed or extended.

Recent work (2006):
  • A complete multi-agent simulation system for ecological environments - EcoDynamo - was developed. This system includes an ecological simulator, a 3D graphical visualizer, and an automatic calibration agent based on machine learning techniques capable of calibrating complex ecological models. It also includes different agents with learning, optimization and negotiation capabilities, representing the often-neglected intelligent entities present in this type of environment. These agents interact with the simulation software in an intelligent manner simulating the behaviour of humans and increase the simulation realism.
  • Agents interact with the system using ECOLANG, a language developed to describe ecological systems. The simulation system is now being applied to coastal ecological models, including Ria Formosa in Algarve and used for aquaculture optimization and environmental resource management, studying how correct management of environmental resources can make progress in the direction of sustainability.
  • New features for known simulators like SoccerServer 2D robotic soccer simulator and RoboCup Search and Rescue simulator were implemented. In this context we have adapted the SoccerServer simulation system to enable more realistic soccer simulation using models of real soccer players and strategies defined by real soccer coaches.
Current and future work:
  • Future work will be concerned with improving EcoDynamo calibration agent and developing new agents to interact with the simulators. These agents will enable the system to be used for aquaculture optimization and for predicting the long-term effect of human interaction with the ecosystem.

 

Multi-Agent Coordination

Research goals:

This research line aims at developing coordination protocols for teams of autonomous (or semi-autonomous) agents that perform in rich, dynamic, both competitive and adversarial environments. It also aims at developing methodologies for organizing and making available knowledge, languages and protocols that enable teamwork

Recent work:
  • We have developed several new coordination methods for teams of autonomous agents that can be applyed to realistic simulations like robotic soccer, search and rescue, ciber-mouse, traffic and ecological simulations. Based on our previous work on strategic modeling, we have developed a multi-purpose, adaptable, strategic coordination layer that allows management of heterogeneous teams, for both centralized and decentralized environments, with reduced use of communication. The model uses a multi-level hierarchical approach. In the first, lower level, the concept of roles is used to reflect the agent’s usual activities. The second level introduces a sub-tactic that aggregates agents with various roles to solve partial objectives. On top of the sub-tactics, formations are used to distribute available agents throughout the sub-tactics. A higher tactical level, then, uses a hybrid method to switch formations. This method is based on a combination of events, situations and precedence. On top of the previous levels, a strategic level is defined that allows commuting between tactics according to scenario conditions.
  • We have also developed a framework for high-level setplay definition and execution, applicable to any RoboCup cooperative league and similar domains. The framework is based in a standard, league-independent and flexible language that defines setplays which may be interpreted and executed at run-time. These methodologies were applied for both RoboCup Rescue and RoboCup Soccer and tested in real competitions and controlled experiments. It enabled us to win three leagues in RoboCup European championship and to be champions of RoboCup 2006 in the Simulation 3D league, showing the usefulness of the approach.
Current and future work:
  • Future work is concerned with the development of a graphical tool for designing team strategies with an associated source code generator. These developments will enable a more generalized use of our strategic layer in the context of RoboCup and in other cooperative domains.
  • We plan to use the strategical layer, through different instantiations, built by means of our graphical tool, in all our teams (simulation 2D, simulation 3D, small-size, middle-size, legged, simulation rescue and physical visualization) participating in European and world RoboCup competitions in 2007.

 

Intelligent Robotics

Research goals:

 Real/simulated robot teams for competitions

Recent work:
  • Together with IEETA/Univ.Aveiro and ISR-P we have developed seven real/simulated robotic teams (simulation 2D, simulation 3D, Coach, small-size, middle-size, 4legged and rescue) and we have participated in five different leagues in RoboCup 2006 European championship (held in Eindhoven, Holland) and in the World Robotic Soccer Championship, (held in Bremen, Germany). In the European championship we won three leagues: Simulation 3D, Simulation Rescue and Small-Size and achieved second place in the Simulation 2D league. Our 3D team was also World Champion of RoboCup, scoring 77 goals with conceding a single goal on this competition.
  • In this research line we are also developing an intelligent wheelchair for individuals with cerebral palsy and quadriplegia. The interaction with the chair is based on facial expressions recognition methodologies, appropriated for the detection of expressions on individuals with Quadriplegia and Cerebral Palsy Handicapped and its integration as the main interaction mechanism to control and configure the wheel chair. The wheel chair control module is based on a multi-agent system with learning capabilities and a user interface that allows interaction, using an appropriated high-level command language. The first results obtained using a simulator showed that our approach has potential to enable a wheelchair to be fully controlled through facial expressions.
Current and future work:
  • Future work includes the participation in RoboCup 2007 (Atlanta) and European RoboCup (Hanover) with our simulated and robotic teams in soccer and rescue competitions. Regarding our wheelchair, future work includes the improvement of the vision system, aiming to allow the system to better perform in outdoor environments or rough lighting conditions. We also intend to construct, from scratch, a real wheelchair prototype and test the methodologies previously developed using simulators in this wheelchair. Future work includes also the development of high-level planning methodologies to enable the user to control the wheelchair for performing complex tasks in the environment, using simple facial expressions.

Individual and Team Performance Analysis

Research goals:

 This research line aims at creating methodologies for analyzing individual and team performance during the accomplishment of complex tasks (like playing soccer, performing a search and rescue operation or an industrial collective labor), and using them for coaching teams of autonomous agents. We have developed prototypes referred below

Recent work:
  • A distributed surveillance system, based on autonomous surveillance agents capable of following a surveillance policy communicated by a central agent in a specific language.
  • A soccer intelligent analysis system based on surveillance agents (with 3D player and ball detection capabilities) and a central agent capable of creating match reports including player individual performance (passes, shoots, runs, dribbles, etc.) and team performance information (ball possession and circulation, playing style, space of effective play, etc.).
  • FCPx tool for high-level evaluation of RoboCup Rescue simulated team’s strategies.
Current and future work:
  • Future work will be concerned with developing methodologies for automatic game analysis and player model automatic creation and the improvement of our player detection and tracking system.
[Go back]

 

 

Document Actions