Here - Agents Lab - University of Nottingham

1 role Thermostat {23 task AchieveTemperature {4 targetTemp : double ;5 }67 task KeepTemperature {8 inputView : UserView ;9 }1011 task DoSelfTest { }1213 understands {14 newThreshold : double ;15 }16 }1 interface Conditioner {23 obsprop isHeating : boolean ;4 obsprop isCooling : boolean ;56 operation startHeating ( speed : double );7 operation startCooling ( speed : double );8 operation stop ();9 }1 interface Thermometer {2 obsprop currentTemp : double ;3 }1 interface View {2 obsprop desiredTemp : double ;3 obsprop thermStatus : ThermostatStatus ;4 }Fig. 2. Definition of an agent role (on the left) and artifact interfaces (on the right).The overall (dynamic) set of agents and artifacts can be organized in one ormultiple logical containers called workspaces, possibly in execution on differentnodes of the network. An agent actually can eventually use concurrently andtransparently artifacts located in different workspaces, not necessarily only thosethat belong to the workspace where the agent is running.The computational model/architecture adopted for simpAL agents is a simplifiedversion of the BDI one, implementing a sense-plan-act like executioncycle [15], but using OOP instead of logic programming to represent and manipulatedata structures. An agent has a belief base, as a long term private memorystoring information about: (i) the private state of an agent, (ii) the observablestate of the environment, and (iii) information communicated by other agents.In simpAL the belief base is composed by a set of beliefs represented by simplevariable-like information items, characterized by a name, a type, and a value—which could be any data object 3 . To perform tasks, an agent exploits the plansavailable in its plan library. Plans are modules of procedural knowledge specifyinghow to act and react to the events of the environment in order to accomplishsome specific task. The set of plans in the plan library depends on the scriptsloaded by the agent. As detailed later on, scripts are modules containing thedescription of set of plans, written by the agent programmers. An agent canhandle multiple tasks in execution at a time.3 It is worth remarking that in existing agent-oriented languages beliefs are typicallyrepresented by first-order logic literals, denoting information that can be used byreasoning engines. However the logic representation is not necessarily part of thebelief concept, as remarked by Rao and Georgeff in [12]:“[beliefs] can be viewed asthe informative component of the system state” and “[beliefs] may be implemented asa variable, a database, a set of logical expressions, or some other data structure”([12],p. 313).137