Journal of Physical Agents

Abstract

This thesis proposes a framework to decision support suitable for supporting the distributed performing of cooperative actions in dynamic and complex multi-agent environments. Decision support is a process aiming to improve the decision-making performance in cooperative scenarios. Simply stated, decision-making is the process of selecting a specific action out of multiple alternatives. This process occurs continuously in daily life. Humans, for instance, have to take decisions about what cloths to wear, what food to eat, etc. In this sense, an agent is defined as anything that is situated in an environment and acts, based on its observation, its interpretation and its knowledge about its situation on such environment to fulfil a particular action. Therefore, to take decisions, agents must get knowledge that allows them to be aware on the actions that can or cannot perform. Here, such process takes three information parameters trying to embody an agent in a typically physical world. This set of information is known as decision axes, which any agent must take into account to decide if it can perform correctly the task proposed by other agent or human. Agents can make better decisions by considering and representing properly such information. Decision axes, mainly based on the agents’ environmental condition, the agents’ physical features and the agents’ social ability, provide to the multi-agent systems a reliable reasoning for achieving feasible and successful cooperative performance.

Currently, many researches tend to generate news advances in agent technology to increase the intelligence, autonomy, communication and self-adaptation in open and distributed agent scenarios. In this sense, this research aims to contribute to the development of a new path to impact on both individual and cooperative decisions in multi-agent environments. In this light, the thesis was used to implement the concrete actions involved in the robot soccer both in simulated as in real scenarios. It emulates a soccer game where agents must communicate, interact and cooperate among them to perform complex actions within a dynamic and competitive scenario, both to drive the design of the involved actions’ requirements as to demonstrate its effectiveness in cooperative jobs. Therefore, the thesis has obtained results, both on simulation and on real experimentations, showing that the framework to decision support for situated agents presented is capable to improve the interaction and the communication, reflecting in a suitable and reliable agent’s team-work within an unpredictable, dynamic and competitive environment.

The experimentation also showed that the selected information to generate the decision axes to situate agents are useful when these agents must perform the proper action or must make sure commitments at each moment in order to reach successfully a goal. Conclusions emphasizing the advantages and usefulness of the introduced approach, in the improvement of multi-agent performance in coordinated task and task allocation problems are presented.