SmartSociety – A Platform for Collaborative People-Machine Computation

SmartSociety – A Platform for

Collaborative People-Machine

Computation

Hong-Linh Truong

withOgnjen Scekic, Daniele Miorandi, Tommaso Schiavinotto,

Dimitrios I. Diochnos, Alethia Hume, Ronald Chenu-Abente, Michael Rovatsos, Iacopo Carreras, Schahram Dustdar, Fausto

Giunchiglia

Distributed Systems Group, TU Wien

[email protected]

http://dsg.tuwien.ac.at/staff/truong/

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 1

Outline

Background

Motivation

The SmartSociety service platform

Prototype and Examples

Conclusions and Future work

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 2

Background –

Computing Elements

Machine-based

Computing

Human-based

Computing

Things-based

computing

Grid

Pro

cessin

g

Unit

Pro

cessin

g

Unit

Arc

hitectu

reA

rchitectu

reC

om

m.

Com

m.

SMP

Ad hoc networks Web of things

38th IEEE SOCA, 20 Oct, 2015, Rome, Italy

Human and machine for solving

complex problems

Collaborative working environments

provide a lot of services but it is up to the user to

use them (e.g., in FP6 projects)

Crowdsourcing

try to “program”/harness human capabilities to solve

problems that machines/software cannot do

Cognitive computing

(want fully automatically to) replace human work by

machines/software, e.g., IBM Watson

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 4

We want machines/humans to coexist – have a fair sharing job

distribution

We want machines/humans to coexist – have a fair sharing job

distribution

Our approach: Collective Adaptive

Systems and Hybridity

Hybridity

People AND software complement each other

Collectiveness

Socio-technical systems/social machines

In future also ‘things’ (sensors, actuators)

Collectives as the first class entity

Adaptiveness

Respond/adapt to ad-hoc situations

Favor collaboration patterns instead of predefined workflows

Human-driven adaptation

Embrace uncertainty

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 5

Motivation (1)

Users/Customers:

need complex collaborative human-machine tasks

(e.g., in predictive maintenance and complex data

analytics)

Peers

software or human services, individually or

collectively providing solutions for solving complex

problems

Human-machine intelligences but “collaboration” is

the main aspect to be supported

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 6

Motivation (2)

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 7

How to support users with task execution and workforce management

functionality?

How to allow peers to collaborate with a fair working condition?

Enable collaboration and adaptiveness

Not purely crowdsourcing(e.g., workflows of human tasks) or

cognitive algorithms (replace human work by machines)

peers & collectives

User-specfic business logic

(arbitrary code)

get peers and plans

Composition

Manager

collective-based task

negotiate a planNegotiator

execute a plan & monitor

ExecutorCommunication

aggregate resultsUser-specfic

arbitrary code

etc.

( . . Peer Manager )

. .

Incentive Manager

)

User ApplicationTask Request

peers & collectives

Application

Motivation (3): Complex services and

interactions

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy

AP

IsA

PIs

8

SmartSociety Platform - core

functions

Enabling platform services

Collectives Management

Information and Collective lifecycle management

Communication

Composition and Execution

Monitoring, Incentive Management, and Provenance

Supporting diverse types (and models) of

applications to be developed

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 9

Not a tightly coupled integrated system for a single purpose

enable the development and programming of applications

Not a tightly coupled integrated system for a single purpose

enable the development and programming of applications

SmartSociety Platform - Key

service engineering issues

System scalability and extensibility

Decouple functions with REST-based open APIs

Loose couple and scalable communication among

services and humans

Asynchronous interactions for humans

Extensive connectors/plug-ins

Human-machine diversity and hybridity

Need many algorithms and protocols and interfaces

Software platforms enable interactions among

people and software services (machines)

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 10

SmartSociety Platform Architecture

crowd of human andmachine peers

privacy model

virtualization & communication

orchestration

collective formation

virtualization & communication

runtime control

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 11

Peer Abstraction & Management

Maintaining information about humans, software

services and their collectives

Utilizing the entity-centric semantic enhanced model

Incorporating several search algorithms from the

Information Retrieval domain

Trading off privacy: disclosing private data needed for

decision making vs. restricted functionality

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 12

Communication Middleware

Virtualization — supporting heterogeneous human and machine

service units as uniformly addressable entities; Supporting ‘collective’ as

a first-class, dynamically-defined entity.

Heterogeneity — supporting various types of communications

channels/protocols/formats between the platform and service units as well

as among service units/collectives, transparently to the platform user.

Communication — providing functionalities of message

transformation, routing, delivery with configurable delivery and privacy

options

Persistence — message persistence and querying.

Scalability — ability to handle large number of intermittently available

service units.

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 13

SMARTCOM Middleware

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy

Philipp Zeppezauer, Ognjen Scekic, Hong Linh Truong, Schahram Dustdar:

Virtualizing Communication for Hybrid and Diversity-Aware Collective Adaptive Systems. ICSOC Workshops 2014: 56-67

Philipp Zeppezauer, Ognjen Scekic, Hong Linh Truong, Schahram Dustdar:

Virtualizing Communication for Hybrid and Diversity-Aware Collective Adaptive Systems. ICSOC Workshops 2014: 56-67

14

Orchestration Manager

Composition

Runtime, various different algorithms: Agent-based,

Ant Colony, FCFS, Greedy

Based on different criteria for different types of tasks

Negotiation

Guided by different protocols but driven by humans

Execution

self-enacted by human peers or orchestrated by

software peer(s)

• monitored for constraint and QoR satisfaction

• elastic management of collectives at runtime

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 15

SmartSociety - current implementation

Peer management: PostgreSSQL and NodeJS

Communication

MongoDB for persisting messages, Apache ActiveMQ for

message queuing

Adapters for email, Dropbox, REST, Facebook, etc.

Orchestration: several algorithms, NodeJS

Provenance and Incentive: Python, Django

REST services with cloud deployment using VMs and

Docker.

But not everything has been integrated

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy

https://gitlab.com/smartsociety/

http://www.smart-society-project.eu/publications/deliverables/

16

SmartShare

17

• Open call: generate rich plans

• Services: recommendation, reputation

and orchestration negotiation

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy

• Experiments: Israel, involving

150 students and staff of the

Ben Gurion University.

• Plan for Italy scenarios

AskSmartSociety!

Machine peers:

Answering collectives from Google and Twitter

Human peers

Provide answers or filter results

AskSmartSociety! is able to actively (on-demand) locate

and engage those individuals or software services

provide answers of required quality.

Quality criteria determine the composition of the

collective, i.e., different service matching and ranking

Machine-centric versus human-centric

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 18

https://www.youtube.com/watch?v=Jr9z2Coqc6M

Performance - Communication

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 19

Windows 7

64-bit, Intel Core2 Duo @ 2.53 GHz, 4.00 GB RAM.

Conclusions and Future work

• Hybrid collective adaptive systems

• Need scalable service platforms to enable the development and

execution of human-machine collaborative complex tasks

• The SmartSociety platform

• Provide various services, protocols and algorithms to enable the

development of human-machine collaborative tasks

• But an open system requires an continuous design and

engineering plan

• Future work

• Orchestration and execution management algorithms and

external services integration

• Platform evaluation in large-scale settings

• Programming models

8th IEEE SOCA, 20 Oct, 2015, Rome, Italy 20

Thanks for your attention!

Hong-Linh Truong

Distributed Systems GroupTU Wien

[email protected]

dsg.tuwien.ac.at/staff/truong

218th IEEE SOCA, 20 Oct, 2015, Rome, Italy