A First-Order Theory of Agent Models with a Layered Architecture

For the efficient design and develotmwnt of multiagent systems, it is iml~ortam to clarify a relationship of agent theories, architecturvs and languagvs. From this viewpoint, we propose a first-order thool’y of agent models hy syntactic fl)rmulatit)n.Thi.fi)rmalization gives a ullified model of intelligent ;.tgVllt.’, [’l)l" t)lamting, ex@cutiug, sensing, and rel,lantfing. "Carious COllt.’e])ts of agelltS" Vllviroll.lllOllI arid itlplit;.d attitudes, pxpressed by modM logi,-s, arp rvl,uih into three-layered first-order theories (Figure 1): thobfi ct-b velt heory of &gents" environmem such as t inw. ;.rction. ,’onditiomd and causality, tllo knou,ltdy~-lt v, [ tl,eory of information about his environnlel,t such ;.~s a.~sumptiolt, verificatilm, plannilig alid m’titm ,,x,,culion. and an mttnti,n-hvd theory guiding his a,’ti,n.including goal achievement. A descril~tion of an ag0,nt on ~a~.’h level is formalizvd a.~ a s,,I of fifsl-ordpr Jogival formula.s which expres.,, the stalos and lily st;try t l’ansitiolts #’orrosltolldillg to t lip con0-opts ~Jli eavh h.v,’l listed abow.. The rvlationshil~.betw0,,.u lhl’pp level.are defined ba..,od on "reflertio’a {Attardi & Sin,i 19ql I.