190 likes | 273 Views
Modeling Component-Based Intelligent Agents in an HLA-Environment using an Agent Development Tool. Jennie Byström Magnus Johansson Staffan Löf. Pitch AB. Outline. Problem Identification Project Focus Definitions – agents, components Characteristics of Agent Factory Summary.
E N D
Modeling Component-Based Intelligent Agents in an HLA-Environment using an Agent Development Tool Jennie Byström Magnus Johansson Staffan Löf Pitch AB
Outline • Problem Identification • Project Focus • Definitions – agents, components • Characteristics of Agent Factory • Summary
Problem Definition • Increasing demand on realistic agent behavior • Hard and time-consuming to develop agents • Development of agent environment takes time • Difficult to change ”coded” behavior as opposed to modeled behavior • There is a desire to reuse existing physical models and add agent behavior
Project Focus • Develop an approach to make the task of creating agent behavior more efficient • Focus on the SME – support the knowledge acquisition process • Consider the complete chain of activities • Behavior modeling • ”Situatedness” • Code generation • Executing/verifying/validation behavior • Use a component-based approach for agent composition
Intelligent Agents - Definition An entity that resides in environments where it interprets ”sensor” data that reflect events in the environment and executes ”motor” commands that produce effects in the environment – FIPA’97
Outside World response stimuli Sensory System Motor System KB Procedural Knowledge Fact Knowledge Cognition Agent Intelligent Agents - Concept
Components • A piece of self-contained software designed for a specific purpose • Exposed through an interface • Designed to be combined and collaborate with other components • Easily interchangeable • Federates are components in an HLA-federation
Characteristics of Agent Factory • Goal-oriented knowledge representation • Rule-based structure for behavior representation • Provides a graphical modeling approach to agents • Provides a predefined, standardized, component-based framework for agents • Domain-independent • Generation of Java-agents
Knowledge Architecture • Goal-oriented behavior • Affects the outside world • Goal-plan-phase structure • Reactive knowledge • Affects the outside world • Reacts on changes in the outside world • Common knowledge • Expands the fact knowledge • Reacts on changes in the outside world • Fact knowledge • The agent beliefs
Goal Sub-goal-1 Sub-goal-2 Plan-1 Plan-2 Sub-goal-3 Sub-goal-4 Phase-1 Phase-3 Phase-2 Phase-4 Phase-5 Plan-3 Plan-4 ... ... Phase-6 Phase-8 Phase-7 Phase-9 Phase-10 Goal-oriented Behavior
Modeling Agent Behavior • Graphical representation of the reasoning behavior
IF targetWithinFireRange THEN fireMissileAtTarget int dx = target.getX() - pilot.getVehicle().getX(); int dy = target.getY() - pilot.getVehicle().getY(); int distance = Math.sqrt(dx * dx + dy * dy); return (distance < 55); Rule Components • Abstraction of conditions and actions • SME-level of the knowledge • Increases reusability and readability of the behavior
Creating and using Rule Components • Graphical construction of rules • Drag and drop to used specified rule components in rules
Agent Sensory system Motor system Sensory system Motor system pRTI Behavior system HLA as Agent Framework • pRTI is used as infrastructure within the agent to connect the behavior component with sensory and motor systems • Component-based approach • Sensory and motor systems becomes more easily exchangeable
Federate Federate pRTI Sensory system Motor system Sensory system Motor system pRTI Behavior system Agent federate Agent Federate • The agent federate hosts an internal federation and participates in an external federation • The sensory and motor systems act as filter between the two
SME- knowledge AgentFactory Behavior Modeling HLA- compliance Intermediate Format Conceptual model SOM FOM Knowledge Representation Back-end Front-end Code Generation • Generates code to integrate the specified motor and sensory systems with the behavior component in an agent federate • Two-phase generation
Validating Behavior • Face Validation • Transparent Agents
Agent Factory Internal Knowledge Space Repository Behavior Editor Agent Composition Code Generator Agent Factory – Overall Architecture
Summary • Agent Factory is an initial approach to support effective development of agent behavior • Graphical tool to support symbolic modeling of behavior knowledge • HLA is used as technical framework • Code generation techniques to automatically generate HLA-compliant federates hosting agents