Artificial Intelligence Introduction to the Intelligent Agents Technology

Transcription

1 Artificial Intelligence Introduction to the Intelligent Agents Technology Maurizio Martelli, Viviana Mascardi {martelli, University of Genoa Department of Computer and Information Science AI, University of Genoa, DISI Agents and MASs 1 / 92

2 Disclaimer This presentation may contain material protected by copyright laws. In particular, many lessons are based on the material that Russel and Norvig made available on the Web, as part of their book Artificial Intelligence: A Modern Approach We made this material available on the Web only to ensure timely dissemination among the students of the AI Course at the Computer Science Department of the University of Genova, and is meant only for students personal use. Any use different from the students personal use is prohibited. AI, University of Genoa, DISI Agents and MASs 2 / 92

3 Outline 1 What did it happen in Genova, one year ago? 2 What is an agent? 3 Short history of agents and MASs 4 Agent theories 5 Agent architectures 6 Agent languages 7 Applications of agents and MASs AI, University of Genoa, DISI Agents and MASs 3 / 92

4 WOA 2007 What did it happen in Genova, one year ago? AI, University of Genoa, DISI Agents and MASs 4 / 92

5 What did it happen in Genova, one year ago? WOA 2007: Scientific Committee Matteo Baldoni (U. Torino) Antonio Boccalatte (U. Genova) Flavio De Paoli (U. Milano - Bicocca) Maurizio Martelli (U. Genova) Viviana Mascardi (U. Genova) AI, University of Genoa, DISI Agents and MASs 5 / 92

6 What did it happen in Genova, one year ago? The school before WOA AI, University of Genoa, DISI Agents and MASs 6 / 92

7 What did it happen in Genova, one year ago? The first day of WOA workshop AI, University of Genoa, DISI Agents and MASs 7 / 92

8 What did it happen in Genova, one year ago? The first day of WOA workshop AI, University of Genoa, DISI Agents and MASs 8 / 92

9 What did it happen in Genova, one year ago? The panel with academic and industrial partners AI, University of Genoa, DISI Agents and MASs 9 / 92

10 What did it happen in Genova, one year ago? Nice atmosphere among those working with agents... AI, University of Genoa, DISI Agents and MASs 10 / 92

11 What did it happen in Genova, one year ago? Why I was not here last week...? AI, University of Genoa, DISI Agents and MASs 11 / 92

12 What is an agent? Some opinions about agents N. Negroponte, Double Agents, WIRED Columns, 1995 a society of electronic agents will be able to communicate far more efficiently than a collection of human cooks, maids, chauffeurs, and butlers. Rumors become facts and travel at the speed of light. Since I constantly argue in articles and lectures that intelligent agents are the unequivocal future of computing. AI, University of Genoa, DISI Agents and MASs 12 / 92

13 What is an agent? Some opinions about agents H. S. Nwana, KE Review, 11(3), 1996 The range of firms and universities actively pursuing agent technology is quite broad and the list is ever growing. It includes small non-household names (e.g. Icon, Edify and Verity), medium-size organisations (e.g. Carnegie Mellon University (CMU), General Magic, Massachusetts Institute of Technology (MIT), the University of London) and the real big multinationals (e.g. Alcatel, Apple, AT&T, BT, Daimler-Benz, DEC, HP, IBM, Lotus, Microsoft, Oracle, Sharp). [...] agents are here to stay, not least because of their diversity, their wide range of applicability and the broad spectrum of companies investing in them. As we move further and further into the information age, any information-based organisation which does not invest in agent technology may be committing commercial hara-kiri. AI, University of Genoa, DISI Agents and MASs 13 / 92

14 What is an agent? Some opinions about agents M. Luck, JAAMAS, 9(3), 2004 Agent-based systems are one of the most vibrant and important areas of research and development to have emerged in information technology in the 1990s, and underpin many aspects of modern computing infrastructures and applications. [...] Agent technology is not restricted to one specific computing or communications domain. Rather, it is likely to play a key role in many aspects of computing. Critical areas of interest address issues of complex problem-solving in science, society, industry and businesses. AI, University of Genoa, DISI Agents and MASs 14 / 92

15 What is an agent? What is an agent, anyway? N. Jennings, K. Sycara, M. Wooldridge, JAAMAS 1(1), 1998 An agent is an hardware or software system situated autonomous flexible reactive proactive social AI, University of Genoa, DISI Agents and MASs 15 / 92

16 What is an agent? What is an agent, anyway? Besides being characterised by the notions identified by N. Jennings, K. Sycara, M. Wooldridge ( weak definition), an agent may be conceptualised following an antropomorphic approach ( strong definition). Y. Shoham, Artificial Intelligence, 60(1), 1993; A. S. Rao, M. P. Georgeff, Proc. of KR&R-92 Mentalistic notions: beliefs, desires, intentions, commitments,... J. Bates, Communications of the ACM, 37(7), 1994 Emotional notions: friendlyness, trust, untrust; agents must be believable AI, University of Genoa, DISI Agents and MASs 16 / 92

17 What is a MAS? What is an agent? AI, University of Genoa, DISI Agents and MASs 17 / 92

18 What is an agent? What is a MAS? F. Zambonelli, A. Omicini, JAAMAS, 9(3), today s software engineering approaches are increasingly adopting abstractions approaching that of agent-based computing. This trend can be better understood by recognising that the vast majority of modern distributed systems scenarios are intrinsically prone to be developed in terms of MASs, and that modern distributed systems are already de facto MASs, i.e., they are indeed composed of autonomous, situated, and social components AI, University of Genoa, DISI Agents and MASs 18 / 92

19 What is a MAS? What is an agent? Franco Zambonelli Andrea Omicini AI, University of Genoa, DISI Agents and MASs 19 / 92

20 What is an agent? What is a MAS? Each agent has incomplete information, or capabilities for solving the problem, thus each agent has a limited viewpoint; there is no global system control; data is decentralized; and computation is asynchronous AI, University of Genoa, DISI Agents and MASs 20 / 92

21 Short history of agents and MASs Agents history Current interest in autonomous agents did not emerge from a vacuum. Researchers and developers from many different disciplines have been talking about closely related issues for some time. The main contributors are: artificial intelligence object-oriented programming and concurrent object-based systems human-computer interface design AI, University of Genoa, DISI Agents and MASs 21 / 92

22 Short history of agents and MASs Artificial Intelligence Among the oldest areas of research, the activity most closely connected with that of autonomous agents was AI planning (A. Newell e H. A. Simon, Lernende Automaten, 1961; R. E. Fikes e N. Nilsson, Artificial Intelligence, 5(2), 1971). Until the 1980s comparatively little effort within the AI community was directed to the study of intelligent agents. The primary reason for this state of affairs was that AI researchers had historically tended to focus on the various different components of intelligent behaviour (learning, reasoning, problem solving, vision understanding and so on) in isolation. Expert systems: disembodied intelligence, no real-time capabilities, no social ability. AI, University of Genoa, DISI Agents and MASs 22 / 92

23 Short history of agents and MASs Artificial Intelligence First delusions with planning: first-principle planning (meaning that, in order to satisfy a goal, the agent has to formulate an entirely new plan for that goal) does not allow the implementation of reactive agents ( calculative rationality ) and it is undecidable in many cases (D. Chapman, Artificial Intelligence, 32, 1987) Traditional, symbolic approaches to agency and AI leave room to the behavioural AI, or reactive AI (R. Brooks, Artificial Intelligence, 47, 1991). AI, University of Genoa, DISI Agents and MASs 23 / 92

24 Short history of agents and MASs Artificial Intelligence From the 90s, researchers start to recognise that both purely reactive architectures, à la Brooks, and purely deliberative architectures, à la first-principle planning, have both advantages and disvantages ==> birth of hybrid architectures, usually stratified (TouringMachines by I. A. Ferguson, 1992, InteRRaP by J. P. Muller e M. Pischel, 1994). Practical reasoning agents: agents whose architecture is modelled on or inspired by a theory of practical reasoning in humans. By practical reasoning, we mean the kind of pragmatic reasoning that we use to decide what to do. The most widely known architecture for practical reasoning is the Belief-Desire-Intention one (M. P. Georgeff e A. L. Lansky, Proc. di AAAI-87, 1987). AI, University of Genoa, DISI Agents and MASs 24 / 92

25 Short history of agents and MASs Artificial Intelligence 1993: researchers start to understand and accept that intelligent agents are not just a branch of AI, but they are an autonomous technological and methodological approach. Y. Shoham, with his paper Agent-oriented programming, Artificial Intelligence 60(1), is considered the father of agent oriented programming. AI, University of Genoa, DISI Agents and MASs 25 / 92

26 Short history of agents and MASs Object-oriented programming Objects are defined as computational entities that encapsulate some state, are able to perform actions, or methods on this state, and communicate by message passing. So, what is the difference between an agent and an object? AI, University of Genoa, DISI Agents and MASs 26 / 92

27 Short history of agents and MASs Object-oriented programming Differences between agents and objects: Autonomy: an object can be thought of as exhibiting autonomy over its state: it has control over it. But an object does not exhibit control over it s behaviour. Objects do it for free; agents do it for money. Flexibility: the standard object model has nothing whatsoever to say about how to build systems that integrate proactive, reactive and social behaviour. One could argue that we can build object-oriented programs that do integrate these types of behaviour. But this argument misses the point, which is that the standard object-oriented programming model has nothing to do with these types of behaviour. Concurrency: in the standard object model, there is a single thread of control in the system. AI, University of Genoa, DISI Agents and MASs 27 / 92

28 Short history of agents and MASs Human-computer interface design With respect to the interaction paradigm known as direct manipulation, we would like to have programs that behave like digital butlers : N. Negroponte, Being Digital, 1995 The agent answers the phone, recognizes the callers, disturbs you when appropriate, and may even tell a white lie on your behalf. If you have somebody who knows you well and shares much of your information, that person can act on your behalf very effectively. If your secretary falls ill, it would make no difference if the temping agency could send you Albert Einstein. This issue is not about IQ. It is shared knowledge and the practice of using it in your best interests. Like an army commander sending a scout ahead... you will dispatch agents to collect information on your behalf. AI, University of Genoa, DISI Agents and MASs 28 / 92

29 Short history of agents and MASs History of MASs: Actors Actors (C. Hewitt, P. Bishop and R. Steiger. A Universal Modular Actor Formalism for Artificial Intelligence IJCAI 1973; G. Agha. ACTORS: A Model of Concurrent Computation in Distributed Systems. The MIT Press: Cambridge, MA, 1986) were proposed as universal primitives of concurrent computation. Actors are self-contained, interactive autonomous components of a computing system that communicate by asynchronous message passing. The basic actor primitives are: create: creating an actor from a behavior description and a set of parameters, possibly including existing actors; send: sending a message to an actor; become: changing an actor s local state. AI, University of Genoa, DISI Agents and MASs 29 / 92

30 Short history of agents and MASs History of MASs: Actors Carl Hewitt... AI, University of Genoa, DISI Agents and MASs 30 / 92

31 Short history of agents and MASs History of MASs: Actors This is me with Carl Hewitt!!! AI, University of Genoa, DISI Agents and MASs 31 / 92

32 Short history of agents and MASs History of MASs: the Contract Net Protocol In the Contract Net Protocol (Davis, R. and R. G. Smith. Negotiation as a metaphor for distributed problem solving. Artificial Intelligence, vol. 20 pp , 1983), agents can dynamically take two roles: manager or contractor. Given a task to perform, an agent first determines whether it can break it into subtasks that could be performed concurrently. It employs the Contract Net Protocol to announce the tasks that could be transferred, and requests bids from nodes that could perform any of these tasks. A node that receives a task announcement replies with a bid for that task, indicating how well it thinks it can perform the task. The contractor collects the bids and awards the task to the best bidder. AI, University of Genoa, DISI Agents and MASs 32 / 92

33 Agent theories Agent theories Online reference: http: // mjw/pubs/ker95/section3_2.html (from M. Wooldridge and N. R. Jennings. Intelligent Agents: Theory and Practice. In Knowledge Engineering Review 10(2), 1995) An agent theory is a specification for an agent; agent theorists develop formalisms for representing the properties of agents, and using these formalisms, try to develop theories that capture desirable properies of agents. Our starting point is the notion of an agent as an entity which appears to be the subject of beliefs, desires, etc. (N. Seel. Agent Theories and Architectures. PhD thesis, Surrey University, Guildford, UK, 1989). The philosopher Dennett has coined the term intentional system to denote such systems (D. C. Dennett. The Intentional Stance. The MIT Press: Cambridge, MA, 1987). AI, University of Genoa, DISI Agents and MASs 33 / 92

34 Agent theories Agents as Intentional Systems When explaining human activity, it is often useful to make statements such as the following: Janine took her umbrella because she believed it was going to rain. Michael worked hard because he wanted to possess a PhD. These statements make use of a folk psychology, by which human behaviour is predicted and explained through the attribution of attitudes, such as believing, wanting, hoping, fearing,... The attitudes employed in such folk psychological descriptions are called the intentional notions. Intentional system = system made up of entities whose behaviour can be predicted by the method of attributing belief, desires and rational acumen. AI, University of Genoa, DISI Agents and MASs 34 / 92

35 Agent theories Agents as Intentional Systems To ascribe beliefs, free will, intentions, consciousness, abilities, or wants to a machine or to a piece of software is legitimate when such an ascription expresses the same information about the machine that it expresses about a person. It is useful when the ascription helps us understand the structure of the machine, its past or future behaviour, or how to repair or improve it. Ascription of mental qualities is most straightforward for machines of known structure such as thermostats and computer operating systems, but is most useful when applied to entities whose structure is incompletely known. J. McCarthy. Ascribing mental qualities to machines. Technical report, Stanford University AI Lab, AI, University of Genoa, DISI Agents and MASs 35 / 92

36 Agent theories Agents as Intentional Systems What objects can be described by the intentional stance? AI, University of Genoa, DISI Agents and MASs 36 / 92

37 Agent theories Agents as Intentional Systems The light switch example Y. Shoham. Agent-oriented programming, Artificial Intelligence 60(1), It is perfectly coherent to treat a light switch as a (very cooperative) agent with the capability of transmitting current at will, who invariably transmits current when it believes that we want it transmitted and not otherwise; flicking the switch is simply our way of communicating our desires. AI, University of Genoa, DISI Agents and MASs 37 / 92

38 Agent theories Agents as Intentional Systems AI, University of Genoa, DISI Agents and MASs 38 / 92

39 Agent theories Agents as Intentional Systems The representation of intentional notions raises a set of delicate technical questions, both on the syntactic and the semantic side. Facing these questions requires a background that you do not have, thus we cannot discuss about them in this course. AI, University of Genoa, DISI Agents and MASs 39 / 92

40 Agent architectures Agent architectures Online reference: http: // mjw/pubs/ker95/section3_3.html (from M. Wooldridge and N. R. Jennings. Intelligent Agents: Theory and Practice. In Knowledge Engineering Review 10(2), 1995) L. P. Kaelbling. A situated automata approach to the design of embedded agents. SIGART Bulletin, 2(4):85-88, 1991 An agent architecture is a specific collection of software (or hardware) modules, typically designated by boxes with arrows indicating the data and control flow among the modules. A more abstract view of an architecture is as a general methodology for designing particular modular decompositions for particular tasks. AI, University of Genoa, DISI Agents and MASs 40 / 92

41 Agent architectures Deliberative architectures A deliberative agent or agent architecture contains an explicitly represented, symbolic model of the world, and in which decisions (for example about what actions to perform) are made via logical (or at least pseudo-logical) reasoning, based on pattern matching and symbolic manipulation. AI, University of Genoa, DISI Agents and MASs 41 / 92

42 Agent architectures Deliberative architectures: the BDI one AI, University of Genoa, DISI Agents and MASs 42 / 92

43 Agent architectures Deliberative architectures: the BDI one Data structures: Beliefs represent the knowledge of the agents. Goals are beliefs, or conjunctions and disjunctions of beliefs, which must be achieved or tested in the current state. Plans contain the procedural knowledge of agents. They are characterized by a trigger; a context; a body; a maintenance condition; a set of success actions ; a set of failure actions. Intentions are partially instantiated plans. AI, University of Genoa, DISI Agents and MASs 43 / 92

44 Agent architectures Deliberative architectures: the BDI one Interpreter: 1 observe the world and the agent s internal state, and update the event queue consequently; 2 generate possible new desires (tasks) by finding plans whose trigger event matches an event in the event queue; 3 select one from this set of matching plans for execution; 4 push the selected plan onto an existing or new intention stack, according to whether or not the event is a (sub)goal; 5 select an intention stack, take the topmost plan and execute the next step of this current plan: if the step is an action, perform it, otherwise, if it is a subgoal, post it on the event queue. AI, University of Genoa, DISI Agents and MASs 44 / 92

45 Agent architectures Problems with the deliberative architectures 1 The transduction problem: that of translating the real world into an accurate, adequate symbolic description, in time for that description to be useful. 2 The representation/reasoning problem: that of how to symbolically represent information about complex real-world entities and processes, and how to get agents to reason with this information in time for the results to be useful. AI, University of Genoa, DISI Agents and MASs 45 / 92

46 Agent architectures Reactive architectures There are many unsolved (some would say insoluble) problems associated with symbolic AI. These problems have led some researchers to question the viability of the whole paradigm, and to the development of what are generally know as reactive architectures. A reactive architecture is one that does not include any kind of central symbolic world model, and does not use complex symbolic reasoning. AI, University of Genoa, DISI Agents and MASs 46 / 92

47 Agent architectures Reactive architectures: the Subsumption one The main critic of symbolic AI has been Rodney Brooks, a researcher at MIT who apparently became frustrated by AI approaches to building control mechanisms for autonomous mobile robots. In a 1985 paper (R. A. Brooks. A robust layered control system for a mobile robot. IEEE Journal of Robotics and Automation, 2(1):14-23, 1986), he outlined an alternative architecture for building agents, the so called subsumption architecture, based on three key theses: Intelligent behaviour can be generated without explicit representations of the kind that symbolic AI proposes. Intelligent behaviour can be generated without explicit abstract reasoning of the kind that symbolic AI proposes. Intelligence is an emergent property of certain complex systems. AI, University of Genoa, DISI Agents and MASs 47 / 92

48 Agent architectures Reactive architectures: the Subsumption one Brooks identifies two key ideas that have informed his research: Situatedness and embodiment: Real intelligence is situated in the world, not in disembodied systems such as theorem provers or expert systems. Intelligence and emergence: Intelligent behaviour arises as a result of an agent s interaction with its environment. Also, intelligence is in the eye of the beholder; it is not an innate, isolated property. AI, University of Genoa, DISI Agents and MASs 48 / 92

49 Agent architectures Reactive architectures: the Subsumption one AI, University of Genoa, DISI Agents and MASs 49 / 92

50 Hybrid architectures Agent architectures Many researchers have suggested that neither a completely deliberative nor completely reactive approach is suitable for building agents. They have argued the case for hybrid systems, which attempt to marry classical and alternative approaches. AI, University of Genoa, DISI Agents and MASs 50 / 92

51 Agent architectures Hybrid architectures: TOURINGMACHINES The TOURINGMACHINES architecture proposed by Ferguson (I. A. Ferguson. TouringMachines: An Architecture for Dynamic, Rational, Mobile Agents. PhD thesis, Clare Hall, University of Cambridge, UK, 1992) consists of perception and action subsystems, which interface directly with the agent s environment, and three control layers. The reactive layer generates potential courses of action in response to events that happen too quickly for other layers to deal with. It is implemented as a set of situation-action rules, in the style of Brooks subsumption architecture. AI, University of Genoa, DISI Agents and MASs 51 / 92

52 Agent architectures Hybrid architectures: TOURINGMACHINES The planning layer constructs plans and selects actions to execute in order to achieve the agent s goals. The modelling layer contains symbolic representations of the cognitive state of other entities in the agent s environment. These models are manipulated in order to identify and resolve goal conflicts - situations where an agent can no longer achieve its goals, as a result of unexpected interference. The three layers are able to communicate with each other (via message passing), and are embedded in a control framework dealing with conflicting action proposals from the different layers. AI, University of Genoa, DISI Agents and MASs 52 / 92

53 Agent architectures Hybrid architectures: TOURINGMACHINES AI, University of Genoa, DISI Agents and MASs 53 / 92

54 Agent languages Agent languages An agent language is a system that allows one to program hardware or software computer systems in terms of some of the concepts developed by agent theorists. At the very least, we expect such a language to include some structure corresponding to an agent. However, we might also expect to see some other attributes of agency (beliefs, goals, or other mentalistic notions) used to program agents. AI, University of Genoa, DISI Agents and MASs 54 / 92

55 Agent languages AGENT-0 In the already cited paper Agent-oriented programming, Artificial Intelligence 60(1), 1993, Shoham proposes that a fully developed AOP ( Agent-Oriented Programming ) system will have three components: 1 a logical system for defining the mental state of agents; 2 an interpreted programming language for programming agents; 3 an agentification process, for compiling agent programs into low-level executable systems. However, he only describes the first two components. AI, University of Genoa, DISI Agents and MASs 55 / 92

56 AGENT-0: syntax Agent languages AI, University of Genoa, DISI Agents and MASs 56 / 92

57 AGENT-0: syntax Agent languages AI, University of Genoa, DISI Agents and MASs 57 / 92

58 AGENT-0: syntax Agent languages Facts. Specify both the contents of actions and conditions for their execution. They are the atomic objective sentences of a simple temporal language: (t atom) and (NOT (t atom)) Example: (0 (stored orange 1000)). Private actions. (DO t p-action) where t is a time instant, and p-action is the name of a private action. AI, University of Genoa, DISI Agents and MASs 58 / 92

59 AGENT-0: syntax Agent languages Communicative actions. (INFORM t a fact) where t is the time point when the communicative act will take place, a is the receiver name, and fact is a fact statement. (REQUEST t a action) where t is the time point when the communicative act will take place, a is the receiver name, and action is an action statement (that might be both a private or a communicative action). (UNREQUEST t a action) where t is the time point when the communicative act will take place, a is the receiver name, and action is an action statement (that might be both a private or a communicative action). AI, University of Genoa, DISI Agents and MASs 59 / 92

60 Agent languages AGENT-0: syntax Non-action. Prevents commitment to a particular action: (REFRAIN action) Mental conditions. A mental condition is a logical combination of mental patterns of the form (B fact) meaning that the agent believes that fact is true, or ((CMT a) action) where CMT means commitment : the agent a is commited to perform action action. AI, University of Genoa, DISI Agents and MASs 60 / 92

61 AGENT-0: syntax Agent languages Capabilities. (action mentalcondition) the agent is capable of performing action if mentalcondition is true. Conditional actions. (IF mentalcondition action) action can be performed only if mentalcondition holds. AI, University of Genoa, DISI Agents and MASs 61 / 92

62 Agent languages AGENT-0: syntax Message conditions. A message condition is a logical combination of message patterns that are triples of the form (From Type Content) where From is the sender s name, Type is either INFORM, REQUEST or UNREQUEST, and Content is either a fact or an action. Commitment rules. A commitment rule has the following form: (COMMIT messagecondition mentalcondition (agent action)*) where messagecondition and mentalcondition are message and mental conditions, resp., agent is the name of the agent toward which the commitment is taken, action is an action and * means zero or more. AI, University of Genoa, DISI Agents and MASs 62 / 92

63 AGENT-0: interpreter Agent languages AI, University of Genoa, DISI Agents and MASs 63 / 92

64 Agent languages The contract proposal example buyerrole contractproposal sellerrole accept refuse contractproposal acknowledge accept refuse contractproposal acknowledge AI, University of Genoa, DISI Agents and MASs 64 / 92

65 Agent languages The contract proposal example When the seller agent receives a contractproposal(stuff, amount, price) message, if there is enough stuff and the proposed price is max, the seller sends an accept message to the buyer and concurrently ships the required goods; if there is not enough stuff or the price is min, the seller agent sends a refuse message to the buyer; if there is enough stuff and min price max, the seller sends a contractproposal message back to the buyer. AI, University of Genoa, DISI Agents and MASs 65 / 92

66 Agent languages AGENT-0: the seller agent example Existentially quantified variables are preceeded by?. Universally quantified ones are preceeded by?!. Timegrain Min Capabilities ((DO?time (ship?!buyer?stuff?required-amount?!price)) (AND (B (?time (stored?stuff?stored-amount))) (>=?stored-amount?required-amount))) Initial beliefs (0 (stored orange 1000)) (?!time (min-price orange 1)) (?!time (max-price orange 2)) AI, University of Genoa, DISI Agents and MASs 66 / 92

67 Agent languages AGENT-0: the seller agent example Commitment Rules (COMMIT (?buyer REQUEST (DO now+1 (ship?buyer?stuff?req-amnt?price))) (AND (B (now (stored?stuff?stored-amount))) (>=?stored-amount?req-amnt) (B (?!time (max-price?stuff?max))) (>=?price?max)) (?buyer (DO now+1 (ship?buyer?stuff?req-amnt?price))) (myself (INFORM now+1?buyer (accepted?stuff?req-amnt?price))) (myself (DO now+1 (update-stuff?stuff?req-amnt))) ) AI, University of Genoa, DISI Agents and MASs 67 / 92

68 Agent languages AGENT-0: the seller agent example (COMMIT (?buyer REQUEST (DO now+1 (ship?buyer?stuff?req-amnt?price))) (OR (AND (B (now (stored?stuff?stored-amount))) (<?stored-amount?req-amnt)) (AND (B (?!time (min-price?stuff?min))) (<=?price?min))) (myself (INFORM now+1?buyer (refused?stuff?req-amnt?price))) ) AI, University of Genoa, DISI Agents and MASs 68 / 92

69 Agent languages AGENT-0: the seller agent example (COMMIT (?buyer REQUEST (DO now+1 (ship?buyer?stuff?req-amnt?price))) (AND (B (now (stored?stuff?stored-amount))) (>=?stored-amount?req-amnt) (B (?!time (max-price?stuff?max))) (<?price?max) (B (?!time (min-price?stuff?min))) (>?price?min)) (myself (DO now+1 (eval-mean?max?price?mean-price))) (myself (REQUEST now+1?buyer (eval-counter-proposal?stuff?req-amnt?mean-price))) ) AI, University of Genoa, DISI Agents and MASs 69 / 92

70 Agent languages Concurrent METATEM M. Fisher and H. Barringer. Concurrent METATEM Processes A Language for Distributed AI, in Proc. of the European Simulation Multiconference, 1991 Formal approach: temporal logic. Time: managed by the since, until, in the next state, in the last state, sometimes in the past, sometimes in the future, always in the past and always in the future operators. Sensing: no explicit operators provided. Communication: each agent has a communicative interface the other agents in the system must know in order to exchange information. Concurrent METATEM does not provide a set of speech acts that all the agents recognize. AI, University of Genoa, DISI Agents and MASs 70 / 92

71 Agent languages Concurrent METATEM Concurrency: a Concurrent METATEM specification defines a set of concurrently executing agents. Nondeterminism: mainly due to nondeterministic temporal operators such as sometimes in the past, sometimes in the future. Modularity: no support to modularity provided. Semantics: Kripke-style semantics given by the = relation that assigns the truth value of a formula in a model M at a particular moment in time i and w.r.t. a variable assignment. AI, University of Genoa, DISI Agents and MASs 71 / 92

72 Agent languages Concurrent METATEM: syntax ψ Uφ : ψ will be true until φ will become true primitive ψ Sφ : ψ was true until φ became true primitive φ : φ is true in the next state [false Uφ] φ : there was a last state and φ was true in it [false Sφ] φ : if there was a last state, φ was true in it [ φ] φ : φ will be true in some future state [true Uφ] φ : φ was true in some past state [true Sφ] φ : φ will be true in all future states [ φ] φ : φ was true in all past states [ φ] AI, University of Genoa, DISI Agents and MASs 72 / 92

73 Agent languages Concurrent METATEM: the seller agent example Interface of the seller agent seller(contractproposal)[accept, refuse, contractproposal, ship] Rigid predicates (predicates whose value never changes) min-price(orange, 1). max-price(orange, 2). Flexible predicates (predicates whose value changes over time) storing(orange, 1000). AI, University of Genoa, DISI Agents and MASs 73 / 92

74 Agent languages Concurrent METATEM: the seller agent example Program rules Buyer. Stuff. Req Amnt. Price. [contractproposal(buyer, seller, Stuff, Req Amnt, Price) storing(stuff, Old Amount) Old Amount >= Req Amnt max-price(stuff, Max) Price >= Max] = [ship(buyer, Stuff, Req Amnt, Price) accept(seller, Buyer, Stuff, Req Amnt, Price)] AI, University of Genoa, DISI Agents and MASs 74 / 92

75 Agent languages Concurrent METATEM: the seller agent example Buyer. Stuff. Req Amnt. Price. [contractproposal(buyer, seller, Stuff, Req Amnt, Price) storing(stuff, Old Amount) min-price(stuff, Min) Old Amount < Req Amnt Price <= Min] = [refuse(seller, Buyer, Stuff, Req Amnt, Price)] AI, University of Genoa, DISI Agents and MASs 75 / 92

76 Agent languages Concurrent METATEM: the seller agent example Buyer. Stuff. Req Amnt. Price. [contractproposal(buyer, seller, Stuff, Req Amnt, Price) storing(stuff, Old Amount) min-price(stuff, Min) max-price(stuff, Max) Old Amount >= Req Amnt Price > Min Price < Max New Price = (Max + Price) / 2] = [contractproposal(seller, Buyer, Stuff, Req Amnt, New Price)] AI, University of Genoa, DISI Agents and MASs 76 / 92

77 Agent languages AgentSpeak(L) AgentSpeak(L) (A. S. Rao. AgentSpeak(L): BDI agents speak out in a logical computable language, in Proc. of MAAMAW 96, 1996) takes as its starting point the procedural reasoning system PRS and its dmars implementation. AgentSpeak(L) is based on a restricted first-order language with events and actions. Beliefs, desires and intentions of the agent are not represented as modal formulas, but they are ascribed to agents, in an implicit way, at design time. The current state of the agent can be viewed as its current belief base; states that the agent wants to bring about can be viewed as desires; and the adoption of programs to satisfy such stimuli can be viewed as intentions. AI, University of Genoa, DISI Agents and MASs 77 / 92

78 Agent languages AgentSpeak(L): syntax AI, University of Genoa, DISI Agents and MASs 78 / 92

79 Agent languages AgentSpeak(L): interpreter AI, University of Genoa, DISI Agents and MASs 79 / 92

80 Agent languages AgentSpeak(L): a waste collector example Initial beliefs AI, University of Genoa, DISI Agents and MASs 80 / 92

81 Agent languages AgentSpeak(L): a waste collector example Plans AI, University of Genoa, DISI Agents and MASs 81 / 92

82 Agent languages AgentSpeak(L): a waste collector example AI, University of Genoa, DISI Agents and MASs 82 / 92

83 Agent languages AgentSpeak(L): a waste collector example AI, University of Genoa, DISI Agents and MASs 83 / 92

84 Agent languages AgentSpeak(L): a waste collector example Triggering event AI, University of Genoa, DISI Agents and MASs 84 / 92

85 Agent languages AgentSpeak(L): a waste collector example Intention (instantiated plan) AI, University of Genoa, DISI Agents and MASs 85 / 92

86 Applications of agents and MASs Applications of agents and MASs Potential applications of agent-based systems can be divided into three broad categories: 1 Assistant agents, such as agents engaged in gathering information or executing transactions on behalf of their human principals on the Internet. 2 Multi-agent decision systems, where the agents participating in the system must together make some joint decisions. 3 Multi-agent simulation systems, where the multi-agent system is used as a model to simulate some real-world domain. AI, University of Genoa, DISI Agents and MASs 86 / 92

87 Assistant agents Applications of agents and MASs The Trading Agent Competition (M. P. Wellman, A. R. Greenwald, P. Stone, P. R. Wurman, Electron. Markets 13(1), 2002), where agents seek to book hotels and make travel arrangements for their principals, provides an example of this type of application. AI, University of Genoa, DISI Agents and MASs 87 / 92

88 Applications of agents and MASs Multi-agent decision systems Agents representing the various components of a telecommunications network may jointly seek to allocate scarce resources across the network, such as call-connections, and thereby manage the operation of the network. The joint decision-making mechanism used by the agents involved may be an economic mechanism, such as an auction, or an alternative mechanism, such as one based on argumentation. The european projects ACTS and EURESCOM involved the most important european telecommunication providers, and are good examples of a multi-agent decision system. AI, University of Genoa, DISI Agents and MASs 88 / 92

89 Applications of agents and MASs Multi-agent simulation systems Typically, multi-agent models are used for domains with many different components, interacting in diverse and complex ways, and where the system-level properties are not readily inferred from the properties of the components. Examples of such domains include: human economies, human and animal societies, biological populations, road-traffic systems, computer networks, and games (such as the agent-based Creatures, S. Grand, D. Cliff, A. Malhotra, in Proc. of ICAA 97). AI, University of Genoa, DISI Agents and MASs 89 / 92

90 Applications of agents and MASs Industrial and commercial applications Process control: Monitoring agents for the space shuttle (G. S. Semmel, S. R. Davis, K. W. Leucht, D. A. Rowe, K. E. Smith, and L. Boloni. Space shuttle ground processing with monitoring agents. IEEE Intelligent Systems, 21(1):68-73, 2006). Supply Chain Management: YAMS (Yet Another Manufacturing System), by Parunak, 1987, applies the Contract Net Protocol to manufacturing control. More recently, Lost Wax and Cap Gemini have developed an agent-based demonstrator in which aircraft are serviced, covering routine and emergency demands for mobile service engineers. AI, University of Genoa, DISI Agents and MASs 90 / 92

91 Applications of agents and MASs Industrial and commercial applications Electronic commerce: Kasbah, by A. Chavez and P. Maes, is a simple electronic marketplace in which agents buy and sell goods. Air traffic control: in OASIS, by M. Ljunberg and A. Lucas, agents are used to represent both aircraft and the various air-traffic control systems in operation. AI, University of Genoa, DISI Agents and MASs 91 / 92

92 Applications of agents and MASs Industrial and commercial applications Medical Applications: GUARDIAN, by B. Hayes-Roth, helps manage patient care in the Surgical Intensive Care Unit (SICU). Entertainment: The second in the Lord of the Rings film trilogy, The Two Towers, achieved visually impressive battle-scenes by using the Massive agent system. Although the battle scene was broadly predetermined, the movement and action of each individual character is controlled by perceiving and responding to the artificial environment and to other characters. AI, University of Genoa, DISI Agents and MASs 92 / 92