An Example of Dynamical Physical Agents Josep Lluís de la Rosa*, Bianca Innocenti 1 *, Albert Oller**, Albert Figueras* Josep Antoni Ramon*, Israel Muñoz* and Miquel Montaner* *Institut d’Informàtica i Aplicacions (IiiA) Universitat de Girona & LEA – SICA Av. Lluís Santaló, s/n E-17071 Girona, Catalunya +34 972 41 84 78 {bianca, peplluis, figueras, jar, imunoz, mmontane}@eia.udg.es ** Dept. d'Enginyeria Elèctrica, Electrònica i Automàtica Universitat Rovira i Virgili (URV) Autovia de Salou s/n. E-43006 Tarragona, Catalunya +34 977 55 97 04 aoller@etse.urv.es ABSTRACT This paper shows the benefits obtained when the dynamic behaviour of the agent’s physical body is taken into account. The agent oriented language Agent0, highlighted the need of declaring the capacities of agents in their reasoning. An example of convoying two controlled autonomous mobile robots as agents is shown. The responsibility of avoiding collisions is for the rear agent, but the reliability of sure decisions based on dynamics is of both of them. The deliberative co-operative decisions based on dynamics provide the controllers with safer set points. Finally, some experimental results using the RoboCup real robots are shown. Keywords Modelling the behaviour of other agents, autonomous robots, designing agent systems. 1. INTRODUCTION A real challenge to AI is to come up with solutions to the problems that are solved routinely by humans without any measurements or any computations in a co-operative way. Let us consider a range of driving automation problems such as: (1) freeway driving with no traffic; (2) freeway driving with light traffic; (3) freeway driving with moderate traffic; (4) freeway driving with heavy traffic; (5) city driving in Helsinki; (6) idem in London; (7) idem in Rome; (8) idem in Istanbul. The current developments, according to L. Zadeh’s opinion, show that automation of (1) is achievable; (2) might be possible, with some qualifications; (3) is not possible today but might be in the future. Beyond (3), the problems are intractable, with no solution in sight. This paper tries to do a step forwards approaches of higher degree of complexity than (2) by using the football robots technology of RoboCup. It contains the problems of driving or manoeuvring one car, and its non-straightforward extension to multiple cars, problems (2) to (8). The fact is that not only feedback control is necessary for solving these problems, but also the co-operative aspects of AI have to be integrated. In this paper, small robots that have clear dynamic movements will emulate the cars. The robots were developed for MIROSOT (Micro Robot Soccer Tournament) and RoboCup events from 1996 [2] and [4]. There is no lack of generality in this approach since we will stress on the co-operative decisions among autonomous mobile robots by considering the dynamics of emulated vehicles [3] and [9]. Techniques applied to Cupertino use the agent oriented analysis that has to be finally implemented on mobile robots. This paper in section 2 introduces concepts of physical agents that pretend to represent the situation of embodying one software agent in an autonomous robot. The section 3 completes the notion of physical agent with dynamical knowledge of autonomous vehicles emulated by autonomous robots. Section 4 shows an example of the advantage of using some robot dynamics’ knowledge in a case of convoying two autonomous vehicles. Finally, in section 5 some conclusions show open research on the formulation of knowledge about dynamics. 2. PHYSICAL AGENTS Previous to the physical agents’ definition, software agents will be introduced. Definition 1: Software agents. This term denotes a software-based computer system that has several properties [13] as autonomy, social ability, reactivity, pro-activeness, mobility, rationality, etc. Physical Agents are software agents that contain the N/S (Numerical/Symbolical) and S/N (Symbolical / Numerical) interface that is typical of real systems, which according to [1] and [8] are constrained by imprecision, uncertainty, changing through time, and others. One typical implementation of physical agents (but not the unique) is mobile robots, that in current research are progressively more and more autonomous and co-operative. The traditional AI has focused on symbolic paradigms (toy problems) and has not 1 Bianca Innocenti is a full time PhD student