Research

Research Projects

  • ExpoLIS - Assessment of Human Exposure to Air Pollution to Change the Way People Move in Cities, LISBOA-01-0145-FEDER-032088, Nov/2018 to Oct/2021
  • OPTOX - Development and validation of bio-optical ecotoxicological tests in marine phototrophs, PTDC/CTA-AMB/30056/2017, Nov/2018 to Oct/2021
  • INTERPHENO - Fostering High Throughput Plant Phenotyping by an Interdisciplinary Approach, PTDC/ASP-PLA/28726/2017, Sep/2019 to Oct/2021
  • ASSISIbf - Animal and robot Societies Self-organise and Integrate by Social Interaction (bees and fish), EU - FOCAS (Feb/2013 to Jan/2018)
  • EEII - Evolution and Ecology of Interacting Infohabitants, V Framework, EU (from May/2000 to Apr/2003

Partner Selection and Evolution of Cooperation

I'm interested in the evolution of cooperation in a population of interacting agents. Specifically, I would like to a minimal agent and environmental model that explains the prevalence of cooperation despite Game Theory predictions of free-riders. I've been working in a general algorithm for partner selection that can be applied in any n-player game.

Game Strategies in Multi-player Games

A game such as Diplomacy poses interesting challenges for Computer Science. From the standpoint of strategical analysis, the branching factor of its tree size is over 1031. The other factor is cooperation, since no single player is capable of winning the game per se, therefore communication, reputation, alliances and action coordination are aspects that must be addressed in the development of an Artificial Player.

João Ribeiro was a MsC student that developed a diplomat. His work addressed the strategical aspects by using potential fields to calculate its moves. His work has resulted in the following publications:

  • João Ribeiro, Pedro Mariano, and Luís Seabra Lopes. Darkblade: a program that plays diplomacy. In Luís Seabra Lopes Nuno Lau Pedro Mariano Luís Rocha, editor, Progress in Artificial Intelligence, 14th Portuguese Conference on Artificial Intelligence, EPIA 2009, volume 5816 of LNAI, pages 485–496 Springer, 2009.
  • João Ribeiro. Darkblade - an artificial inteligent agent for diplomacy. Master's thesis, University of Aveiro, 2008.

Rui Deyllot in his MsC developed a database that stores for specific board positions the strategy that guarantees controlling a specific province. We only consider two players. We start by enumerating board positions within some search horizon. Then we apply a planning algorithm to find who can control a specific province. His work is available in the following publication:

Nuno Fonseca investigated the classification of player relations.

Gustavo Pires during his MsC developed an agent using the work of Rui Deyllot. The initial goal of using a database to perform strategic analysis was not achieved. Instead, Gustavo found out that the database was most usefull during opening moves. His work is available in the following publication:

As of 2006 University of Aveiro adapted its courses to the Bologna Process. The above Master students correspond to the new 2nd cycle.

PhD Thesis

I've a PhD in the area of multi-agent systems. The title of my thesis is Distributed Control of Non-Cooperative Agents-Exploiters and Traitors. The thesis is written in Portuguese. The goal of my thesis is to prevent social behaviours such as treason, exploitation and manipulation to appear in a population of interacting agents.

I take an utilitarian approach to agent interaction meaning that I model interaction as a n-player game. The game is defined by a set of strategies and a set of utility functions. My goal is to restrict the agents to use the cooperative strategies, that is to say, those that increase the social gain of the population.

I intend to obtain a distributed model meaning that there is no central authority where the behaviour of agents is stored.

Give-Take Game

During the course of my PhD I looked at some problems refereed in the Cooperation literature and found many references to the Iterated Prisoner's Dilemma. I disliked its lack of agreement between the agents although I acknowledge that agents may choose to cooperate instead of defecting.

I've developed a game that has multiple strategies to cooperate and thus requires an agreement between the agents so that they can increase the social gain.

This is discussed in more detail in this link.

SimBatch

Most of my experimental work is running simulations of multi-agent systems interacting through a game. Since I have to test a lot of parameters, I've developed a program that takes care of setting the simulation, running it and collecting the files it generates. More information about this program can be found in this link.