gradient-based self-organisation patterns of anticipative adaptation

19
Gradient-based Self-organisation Patterns of Anticipative Adaptation Sara Montagna, Danilo Pianini and Mirko Viroli [email protected] Alma Mater Studiorum—Universit` a di Bologna a Cesena Sixth IEEE International Conference on Self-Adaptive and Self-Organizing Systems Lyon, France; 11th September 2012 Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 1 / 18

Upload: danilo-pianini

Post on 20-Aug-2015

345 views

Category:

Technology


3 download

TRANSCRIPT

Page 1: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Gradient-based Self-organisation Patterns of AnticipativeAdaptation

Sara Montagna, Danilo Pianini and Mirko [email protected]

Alma Mater Studiorum—Universita di Bologna a Cesena

Sixth IEEE International Conference onSelf-Adaptive and Self-Organizing Systems

Lyon, France; 11th September 2012

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 1 / 18

Page 2: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Outline

Goal and result

Start from a catalogue of design patterns for spatial self-organisation

Aim at extending the gradient pattern with temporal aspects

Design the antipative gradient pattern (“proacting” to known future!)

Build it by combination of simpler patterns

Qualitative evaluation by simulation

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 2 / 18

Page 3: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Self-organisation Patterns: reusable design elements

⇒ Start from the layered catalogue in [FMDMSM+12]

The gradient pattern case

Information about a source node becomes global knowledge

Information to reach the source is propagated hop-by-hop

Self-heal to changes, but tackling only “present-awareness”

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 3 / 18

Page 4: Gradient-based Self-organisation Patterns of Anticipative Adaptation

When spatial gradient is not enough

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 4 / 18

Page 5: Gradient-based Self-organisation Patterns of Anticipative Adaptation

When spatial gradient is not enough

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 5 / 18

Page 6: Gradient-based Self-organisation Patterns of Anticipative Adaptation

When spatial gradient is not enough

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 6 / 18

Page 7: Gradient-based Self-organisation Patterns of Anticipative Adaptation

When spatial gradient is not enough

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 7 / 18

Page 8: Gradient-based Self-organisation Patterns of Anticipative Adaptation

When spatial gradient is not enough

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 8 / 18

Page 9: Gradient-based Self-organisation Patterns of Anticipative Adaptation

When spatial gradient is not enough

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 9 / 18

Page 10: Gradient-based Self-organisation Patterns of Anticipative Adaptation

From Gradient to Anticipative Gradient

From space to space-time: anticipative gradient

The Gradient should deviate now to anticipate known later events

Some design guidelines

Design anticipative gradient by combining more elementary patterns

Assume estimated node-to-node average travelling time is available

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 10 / 18

Page 11: Gradient-based Self-organisation Patterns of Anticipative Adaptation

From Gradient to Anticipative Gradient

From space to space-time: anticipative gradient

The Gradient should deviate now to anticipate known later events

Some design guidelines

Design anticipative gradient by combining more elementary patterns

Assume estimated node-to-node average travelling time is available

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 10 / 18

Page 12: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Spatial Structure: Horizon Wave

Horizon Wave

Advertises a future event

Creates a shrinking crown around the source of the future event points

It’s the set of nodes possibly reaching the source during the event

Figure : Horizon Wave pattern, with its shrinking dynamics in evidence.

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 11 / 18

Page 13: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Spatial Structure: Gradient Shadow

Allows to identify multiple different paths toward the POI

Tags the gradient paths passing by some future event area

Nodes store (all) the paths transiting/non-transiting across FEs

Figure : Shadow spatial structure with overlapping Future Events

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 12 / 18

Page 14: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Spatial Structure: Future Event Warning

Future Event Warning

Users in this area will reach the event

Intersection of Gradient Shadow and Horizon Wave.

Figure : Warning spatial structure as intersection of Wave and Shadow

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 13 / 18

Page 15: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Spatial Structure: Anticipative Gradient

Anticipative Gradient

Chooses the time-shortest path

Penalises those paths passing through the event

Users travelling those paths must wait for the event to finish

Figure : Anticipative Gradient pattern (and “waiting distance” T ′).

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 14 / 18

Page 16: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Early evaluation

“Implementation” as chemical-like reactions (SAPERE [ZCF+11])

Used Alchemist simulator (http://alchemist.apice.unibo.it)

0

0.1

0.2

0.3

0.4

0.5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Figure : An Anticipative Gradient case: (Left) estimated distance normalised bythe maximum value; (Right) time-to-destination improvement factor and steeringdirection.

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 15 / 18

Page 17: Gradient-based Self-organisation Patterns of Anticipative Adaptation

Conclusions

1 Added proactive adaptation to the Gradient pattern2 By identification and composition of simpler patterns

Wave, Shadow, Warning, Anticipative Gradient

Future works

1 Dealing with a wider set of future events

2 Deep analysis of performance achievements

3 Application of the approach to real scenarios of traffic/crowd routing

4 Prototype implementation in the SAPERE framework [ZCF+11];

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 16 / 18

Page 18: Gradient-based Self-organisation Patterns of Anticipative Adaptation

References

References I

Jose Luis Fernandez-Marquez, Giovanna Di Marzo Serugendo, Sara Montagna, MirkoViroli, and Josep Lluis Arcos.Description and composition of bio-inspired design patterns: a complete overview.Natural Computing, May 2012.Online First.

Franco Zambonelli, Gabriella Castelli, Laura Ferrari, Marco Mamei, Alberto Rosi, GiovannaDi Marzo Serugendo, Matteo Risoldi, Akla-Esso Tchao, Simon Dobson, GraemeStevenson, Juan Ye, Elena Nardini, Andrea Omicini, Sara Montagna, Mirko Viroli, AloisFerscha, Sascha Maschek, and Bernhard Wally.Self-aware pervasive service ecosystems.Procedia CS, 7:197–199, 2011.

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 17 / 18

Page 19: Gradient-based Self-organisation Patterns of Anticipative Adaptation

References

Gradient-based Self-organisation Patterns of AnticipativeAdaptation

Sara Montagna, Danilo Pianini and Mirko [email protected]

Alma Mater Studiorum—Universita di Bologna a Cesena

Sixth IEEE International Conference onSelf-Adaptive and Self-Organizing Systems

Lyon, France; 11th September 2012

Montagna, Pianini, Viroli (UNIBO) Anticipative Gradient SASO 2012 18 / 18