This document proposes a conceptual map and classification system for ensembles of autonomic components based on their awareness and organization. It outlines trees of concepts related to the single component level (communication, internal architecture, awareness capabilities) and the ensemble level (organization, architecture, team features). The goal is to enhance reuse and adaptation of components by defining their relationships and required features for different coordination patterns. Future work includes formalizing the concept of distances between patterns and implementing test scenarios to validate the classifications.

1. CONCEPTUAL MAP AND CLASSIFICATION
IN ENSEMBLES OF
AUTONOMIC COMPONENTS: FROM
AWARENESS TO
ORGANIZATION
Nicola Capodieci and Giacomo Cabri
University of Modena and Reggio Emilia

2. CONTENTS
 Background
 Motivation
 Trees of concepts
 Ensemble as whole
 Single components features
 Communication
 Adaptation
 Awareness
 Conclusions and future work

3. BACKGROUND
 Ensembles of autonomic components have to
 Be independent
 Have little or no human interaction
 Be reliable
 Be adaptive
 Self-organize
 Self-express
 Self-aware (*-awareness)
 ….
 FOCUS ON:
DARS and
self-expression

4. MOTIVATIONS
 Classification as an instrument for:
 Design process
 Enhancing reutilization of code, projects etc…
 Older classification of autonomic components
 Sometimes outdated
 No focus on adaptive capabilities, awareness
 Building a general framework for self-expression

5. CLASSIFYING COORDINATION PATTERNS
 Previous focus on:
 the ensemble as a whole
 Hardware capabilities of the single robot
 Current trend on:
 What makes a single component adaptive?
 New approaches for ensemble coordination
 *-Aware components
 Our approach:
 Try to consider, integrate and discuss all the previous
topics.

6. THE TREE(S) OF CONCEPTS
 Two seeds:
 Single component
 Communication
 Internal architecture
 Awareness capabilities
 Ensemble as a whole
 Organization
 Global architecture
 Team features

7. ENSEMBLE AS A WHOLE 1/3
 Organization
Unstructured
Structured
Swarm
Leader Peers
Role Based

8. ENSEMBLE AS A WHOLE 2/3
 Architecture
Deliberative Reactive (!)

9. ENSEMBLE AS A WHOLE 3/3
 Team Features
Composition Size
Homogeneous Large Small
Heterogeneous

10. SINGLE COMPONENT TREE
 Communication:
 Direct:
 Well defined protocols
 Data transmission
Communication can be used for raising awareness.
 Direct
communication are characterized by
 Bandwidth and Range.

11. LESS DIRECT WAYS FOR COMMUNICATING
 Stigmergic approach:
 Computational Fields
 Virtual Pheromones
 …
 Sensing approach:
 “Sensing” changes in the environment
 Giving them different meanings according to the state in
which the unit is located

12. COMMUNICATION AND AWARENESS
 Communication is used for raising awareness of:
 Events
 Presence of team mates
And managing COORDINATION
 What about non-communicative coordination?
Strictly related to Awareness
 Social rules, conventions, common pre-shared
knowledge
See: Common Expected Payoff in [1]
[1] J.R. Kok, M.T.J. Spaan and N. Vlassis Non-communicative multi-robot
coordination in dynamic environments, in Robotics and autonomous
System, Multi-Robots in dynamic Environments, Vol. 50, Issues 2-3, 28
Feb. 2005, p. 99 - 114

13. CONVENTIONS, SOCIAL RULES … AWARENESS
“Imagine that you and a friend need to meet today.
You both arrived in Paris yesterday but you were
unable to get in touch to set a time and place.
Nevertheless, it is essential that you meet today.
Where will you go, and when?”
Vohra 1995 AAAI
Symposium on active learning

14. INTERNAL ARCHITECTURE 1/2
[2] G. Cabri, M. Puviani, and F. Zambonelli. Towards a taxonomy of adaptive
agent-based collaboration patterns for autonomic service ensembles.
2011 Collaborative Technologies and Systems, Philadelphia (USA), May 2011.

15. INTERNAL ARCHITECTURE 2/2
 Connection with cognitive heuristics self-aware [3]
agents:
 Feedback
 Reasoning
 Learning
 Planning
 Goal driven
 …
[3] A. Guazzini A Cognitive Heuristic model for Local Community
Recognition Lecture at AWASS 2012, June 2012 Edinburgh, Scotland,
U.K

16. THE CHALLENGE OF DEFINING AWARENESS
 Many previous works and classification
 Many fields/case studies
 Many open issues:
 Evaluating
 Assessing
 Definition
 …
 LOT OF CONFUSION!

17. SOME EXAMPLES IN PHILOSOPHY AND ART
R. Descartes: “Cogito ergo
Sum”
V. Van Gogh
“People say - and I’m quite willing to believe it -
that it’s difficult to know oneself ...”

18. … AND NEUROSCIENCE
Assessing visual
Self-awareness
Picture taken from: http://hellbox.org/squeezebox/archives/cat_sketchbook.html

19. Assess fake-beliefs
awareness
Picture taken from
http://www.asperger-advice.com/sally-and-anne.html

20. PUTTING ALL TOGETHER
 From many definitions and tests for
definingevaluate awareness:
 Extract what is relevant for autonomic components
 Making clear distinctions when needed
 Context and perception are critical when trying to
classify the degree of awareness.
The link between the three module internal architecture and the five degrees of
Self-consciuosness by Neisser et. Al[4]
[4]U. Neisser The roots of self-knowledge: Perceiving self, it, and thou in
Annals of the NY AoS., vol. 818, pp. 1933, 1997.

21. AWARENESS CAPABILITIES

22. WHAT CAN WE DO WITH THIS
CLASSIFICATION?
 Find the concept of distances in patterns:
 Every ensemble organization may require different
features of the trees.
 A requirement may be internal or external, strong or
weak.

23. EXAMPLE
 A Master-Slave configuration requires:
 Direct communication
 Be aware of other team mates
 A proactive internal architecture
 Heterogeneous composition
 …
 A Swarm configuration requires:
 Stigmergy
 Be aware of the environment and events
 A reactive internal architecture
 Homogeneous composition
 …

24. CONCLUSION AND FUTURE WORK
 Provide a conceptual map for Multi robot systems
 Adaptation
 Awareness
 Software and Hardware features
 Study how to express the concept of distances in
patterns:
 Estimate the effort for dynamically change a
coordination pattern
 More formal definitions
 Implementing test scenarios
 Model checking

25. THANK YOU
 The work is partially supported by:
(EU FP7-FET, Contract No. 257414) www.ascens-ist.eu
(nicola.capodieci, giacomo.cabri)@unimore.it
http://agentgroup.unimo.it/