1 INTRODUCCIÓN Intelligent distributed monitoring Architecture for Flexible Manufacturing Systems Based on Multi-agent Paradigm FM S Modeled as Society of cognitive agents Monitoring Heterogeneou s resources Sufficient Knowledge for the decision- making FMS Monitoring Functions Co-operate to achieve

2 INTRODUCCIÓN Need adaptability in production structures Continuosly increasing -Increase in competition -Decrease product life cycle Efectiveness of a production Cycle, due date inventory level Flexibility and reacticity to integrate the evolution of the market

3 INTRODUCCIÓN Multi Agents Systems (MAS) - More Suitable for dynamic monitoring - Build Complex and flexible advanced intelligent systems Monitoring Systems More - decentralized - emergent - concurrent Than - centralized - planned - sequential

4 Multi-agent Architecture for Distributed and Real Time Monitoring Monitoring systems Functions - Error detection - Error diagnosis - Error handling Diagnosis Detection Adquisition of sensors information Error handling Execution RESOURCES Sensors Effectors

5 Multi-agent Architecture for Distributed and Real Time Monitoring MAADM Replace a Centralised monitoring Network of cognitive agents “RESOURCE MONITORING AGENTS” (RMA) Each resource Under the responsability a RMA handle: Local rules and enviromental information to perform the monitoring functions Overall system perfomance It is not globally planned EMERGE Dynamic interaction of the agents in real-time CONTRACT NET PROTOCOL The architecture receives the tasks to execute in the FMS Task Manager Agent INTERFACE

6

7 MAA - The monitoring of the FMS is distribuited over the RMAs - Global Monitoring is achieved by passing appropiate messages between the RMAs - The Contract Net Protocol is used to co-ordinate and co- operate the RMAs, and ro reshedule the operations in failure

8 Resource Monitoring Agent Model Cognitive Agent Model - Knowledge - Control ( Reasoning ) - Co-operation - Communication Distinguishes Knowledge (internal and external activities) TYPES - Self-Knowledge: own information for monitoring its resource (resource state database, diagnosis production rules and error hendling production rules) - Other Knowledge: information about other RMAs, wich can operate whit him

9 Resource Monitoring Agent Model

10 Communication (synchronous and is done by sending messages) TYPES OF MESSAGES - Information request messages: part dimensions, part weight,presence or no presence of the part, resource status, tools status, etc - Failure Alert Messages: resource in failure, tool in failure, etc - Cooperation and co-ordination messages: contracts, bids, acknowledgements, etc Resource Monitoring Agent Model

11 Control (RMAs decision-making capability to acomplish the monitoring functions and to co-operate with others agents) SET OF METHODS - Detection: It interacts with the resources status database to detect the abnormal situations ocurring on each resource. - Diagnosis: It identifies and recognises faults, using a rule-based expert system. The knowledge bse is a set of production rules. If event 1 and event n then fault 1,........,fault m - Error Handling: It determinate the corrective actions that handle the faults. They are:resheduling, urgent halt, transit to degradaded mode, operator call and actionleap - Contract Net Protocol Resource Monitoring Agent Model

12 RESCHEDULING It reschedules locally operations that couldn´t be executed by the resource in failure using the Contract Net Protocol, developed to resolve the problem of dynamic scheduling in FMS. - Detection: It interacts with the resources status database to detect the abnormal situations ocurring on each resource. - Diagnosis: It identifies and recognises faults, using a rule-based expert system. The knowledge bse is a set of production rules. If event 1 and event n then fault 1,........,fault m - Error Handling: It determinate the corrective actions that handle the faults. They are:resheduling, urgent halt, transit to degradaded mode, operator call and actionleap - Contract Net Protocol Resource Monitoring Agent Model

13 Contract Net Protocol

14 Propagación de Requisito Agente Recurso Manager RA 21 RA 22 RA 41 RA 42 Requisito (OP 0 ) RA 11 RA 12 RA 31 RA 32 RA 51 RA 52 Requisito (OP 2 ) Requisito (OP 1 ) Requisito (OP 3 ) Requisito (OP 2 ) Requisito (OP 4 ) Requisito (OP 1 ) Requisito (OP 0 )

15 Realimentación de Petición Agente Recurso Manager RA 21 RA 22 RA 41 RA 52 Petición (OP 0 ) RA 11 RA 12 RA 31 RA 32 RA 51 RA 52 Petición (OP 2 ) Petición (OP 1 ) Petición (OP 3 ) Petición (OP 2 ) Petición (OP 4 ) Petición (OP 1 ) Petición (OP 0 )

16 Propagación de Contrato Agente Recurso Manager RA 21 RA 31 RA 41 RA 51 RA 11 Contrato (OP 0 ) Contrato (OP 2 ) Contrato (OP 1 ) Contrato (OP 3 ) Contrato (OP 4 )

17 Realimentación de Aceptación y Cancelación de petición Agente Recurso Manager RA 21 RA 42 RA 11 RA 32 RA 52 Admisión (OP 0 ) Cancelación (OP 0 ) Admisión (OP 2 ) RA 12 Cancelación (OP 2 ) Cancelación (OP 1 ) Cancelación (OP 1 ) Admisión (OP 3 ) RA 31 RA 51 Cancelación (OP 4 ) Admisión (OP 1 ) RA 22 RA 41 Cancelación (OP 3 ) Admisión (OP 4 )

18 Resolución de conflictos e incertidumbres RA 3 RA 2 RA 1 P1P1 P2P2 P1P1 P2P2 Tiempo Actual Ventana reservada para contratos anteriores Reparo Temporal RA 3 RA 2 RA 1 C1C1 C2C2 P1P1 P2P2 Tiempo Actual Ventana reservada para contratos anteriores Reparo Temporal RA 3 RA 2 RA 1 A1A1 P2P2 C1C1 P2P2 Tiempo Actual Ventana reservada para contratos anteriores Reparo Temporal RA 3 RA 2 RA 1 P1P1 P2P2 P 1 = P 2 Tiempo Actual Ventana reservada para contratos anteriores Reparo Temporal

19 Conclusiones Las principales ventajas de la arquitectura propuesta son flexibilidad en un ambiente variable, reconfiguración y adaptibilidad inherente, tolerancia a fallos rápida y eficiente, desarrollo modular y autonomía operacional. La implementación de la arquitectura multiagente fue llevada a cabo en Java por ser orientada a objetos, distribuida, segura, dinámica y programación de alto desempeño. El Java Expert System Shell fue integrado en el software, para el desarrollo del sistema experto basado en reglas, usado en el diagnóstico y manejo de error. Los resultados experimentales obtenidos en la simulación, demostraron que la arquitectura multiagente is una plataforma de trabajo bastante apropiada para monitorización distribuida y en tiempo real de FMS.