MAS AND SOA: A CASE STUDY EXPLORING PRINCIPLES AND TECHNOLOGIES TO SUPPORTSELF-PROPERTIES IN ASSEMBLY SYSTEMS

Luís Ribeiro (ldr@uninova.pt)

José Barata (jab@uninova.pt)

Armando Comlombo(awcolombo@ieee.org)

1

UNINOVAInstituto de Desenvolvimento de Novas Tecnologias

PRESENTATION OUTLINE Motivation

Systems, Paradigms and IT MAS and SOA System Componentization

Building Blocks NOVAFLEX Cell

Architecture A skill based approach Generic Services Device Level System Level

Implementation A generic communication Interface Interaction Patterns Test Scenario

Conclusions and Outlook2

MOTIVATION

3

Market Turbulence

Profitable but volatile opportunities

Mass Customization

Safety and Environment

Sustainable Development

…

Flexible, Bionic, Holonic and Reconfigurable Manufacturing Systems; Evolvable Assembly Systems, Evolvable Production Systems...

MAS

SOA Internet

Web Services

Agent Platforms

Web Standards

Neural Networks

Prospective Reasoning

…

SYSTEMS, PARADIGMS AND IT (1) Current approaches will soon be insufficient The Systems advocated by modern production

paradigms are increasingly powerful, full featured and COMPLEX

4

SYSTEMS, PARADIGMS AND IT (2)

Existing IT/AI are no adequate in the present Software Hardware

There is a lack of support tools Development Analysis

Is a new paradigm required? Understand existing paradigms and technologies Attempt to integrate/adequate them to fulfill emerging

requirements

5

MAS AND SOA: BRIEF COMPARATIVE ANALYSIS

6

Characteristics SOA MAS

Basic Unit Service Agent

Autonomy Both entities denote autonomy as the functionality provided is self-contained

Behaviour description

In SOA the focus is on detailing the public interface rather than describing execution details.

There are well established methods to describe the behaviour of an agent.

Social ability Social ability is not defined for SOA nevertheless the use of a service implies the acceptance of the rules defined in the interface description

The agents denote social ability regulated by internal or environmental rules

Complexity encapsulation

Again, the self-contained nature of the functionalities provided allows hiding the details. In SOA this encapsulation is explicit.

Communication infrastructure

SOA are supported by Web related technologies and can seamlessly run on the internet.

Most implementations are optimized for LAN use.

Support for dynamically reconfigurable run-time architectures

Reconfiguration often requires reprogramming

The adaptable nature of agents makes them reactive to changes in the environment.

Interoperability Assured by the use of general purpose web technologies.

Heavily dependent on compliance with FIPA-like standards.

Computational requirements

Lightweight implementations like the DPWS guarantee high performance without interoperability constraints

Most implementations have heavy computational requirements

COMPONENTIZATION - CONCEPT

Each device is provided an IT frontend

Harmonizes the devices with the infrastructure

Improves Flexibility and Scalability

Enables (re)composition of the system using building

blocks

Encapsulates complexity

7

Each type of service (CLS, MDS and SMI) is generic for

a given family of hardware.

Each service provides generic processing and

information flow

Each service has a typical interaction pattern

COMPONENTIZATION - CONCEPT

8

•Service to Machine Interface (SMI): wraps legacy devices harmonizing them with the remaining infrastructure

•Manufacturing Device Service (MDS): abstracts a manufacturing component: gripper, robot, conveyor, …

•Coalition Leader Service (CLS): aggregates and orchestrates MDSs

9

COMPONENTIZATION – BUILDING BLOCKS

COMPONENTIZATION – NOVAFLEX CELL

10

COMPONENTIZATION – NOVAFLEX CELL

Type of Service

Hardware

CLS Pallets, Station “Orchestrators”, Node “Orchestrators”

MDS Robots, Grippers, Tool warehouses, Conveyors, Conveyors, Routing Devices, Fixing Devices

SMI Bosch Controller SMI

11

ARCHITECTURE – A SKILL BASED APPROACH

A skill is a action a that a given service knows how

to execute and therefore is able to offer to the

remaining system.

MDS offers simple skills – atomic skills

CLS offers complex skills – skills that are composed

of other skills

12

ARCHITECTURE – SKILLS AVAILABLE IN NOVAFLEX’S GENERIC SERVICES

Type of Service

Device Skills

SMI Bosch Station “Interop” service

MDS Gripper grip, ungrip, get_lcp_list, get_lcp_item

Robot move, move linear, change speed, lockgripper, unlockgripper

Tool Warehouse store part, unstore part, get position to store part, get part position, confirm store, confirm unstore

Conveyor input part, output part

Fixing Device Fix pallet, unfix pallet

Routing Device Output part

CLS Coalition dependent

RobotMDS + gripperMDS -> Pick and PlaceRobotMDS + gripperMDS + Tool Warehouse -> Switch GripperConveyorMDS + RobotMDS + gripperMDS + Tool Warehouse -> produce watch…

13

ARCHITECTURE – DEVICE LEVEL

Each device performs self-analysis

14

ARCHITECTURE – SYSTEM LEVEL

Interaction and Orchestration

15

ARCHITECTURE – A PLUG AND PLAY RECONFIGURABLE ENVIRONMENT

To accomodate changing requirements the building blocks (SMI, MDS and CLS) can be (re)combined.

16

IMPLEMENTATION - A GENERIC COMMUNICATION INTERFACE FOR WEB SERVICES (1)

Harmonizes the IT infrastructure

Supports Generic Ochestration

Supports Structured Interaction

Supports Runtime Changes

Minimizes Code Explosion and Complexity

17

It is essential as services take diferent roles during functionning

IMPLEMENTATION - A GENERIC COMMUNICATION INTERFACE FOR WEB SERVICES (2)

18

• Implements:▫Generic message structure▫Generic types of interaction

IMPLEMENTATION - A GENERIC COMMUNICATION INTERFACE FOR WEB SERVICES (3)

19

IMPLEMENTATION – SERVICE TO MACHINE INTERFACE (SMI)

Interaction Pattern NOVAFLEX

Interoperates the Bosch SCARA controller privding acess to the robot, grippers and tool warehous

20

IMPLEMENTATION – MANUFACTURING DEVICE SERVICE (MDS)

Interaction Patterns with/out SMI NOVAFLEX

Abstracts each device harmonizing them with the infrastructure

21

IMPLEMENTATION – COALITION LEADER SERVICE

NOVAFLEX

•The pallet is the main Coalition Leader and ochestrates its way around the system•Each station has its own ochestrator (CLS) that according to the devices under the coalition offers diferent complex skils. For instance Stattion 2 provides pick and place and swicth gripper operations under the presence of the following MDS: robot, gripper and toolwarehouse

22

TEST SCENARIO – RE-ORCHESTRATION AS A RECOVERY ACTION

Failure Scenario:

The pallet orchestrates its way until Station 2 where it requests the CLS a

pick and place operation.

When attempting to open the active gripper detects that is unable to perform

the operation. Upon diagnosing the fault and acknowledging that there is no

possible recovery within its scope, the gripper MDS reports it to the CLS.

The CLS checks in its diagnosis and recovery knowledge base if there a

specific handler for that fault and eventually identifies one.

For that specific gripper, a fault in step 2 of a pick and place operation can be

recovered by switching gripper. A conversation is initiated with the tool

warehouse.

The Gripper is replaced

The Pick and Place Restarts 23

TEST SCENARIO – GRIPPER REPORTS THE FAILURE

24

25

Despite the theoretical prospects and expectations, in practice technology requires further improvements to meet the IT requirements of modern production systems

Computational power is still a constraint in embedded tiny devices

SOA and MAS paradigms individually provide an incomplete solution

Experience shows that there are benefits in merging both concepts and technologies

CONCLUSIONS

OUTLOOKCurrently

DPWS Service

DPWS Service

DPWS Service

DPWS Service

DPWS Service

DPWS Service

DPWS Service

NOVAFLEX

DPWS Service

DPWS Service

DPWS Service

DPWS Service

DPWS Service

DPWS Service

DPWS Service

Future

26

Thanks for your attention

27