際際滷

際際滷Share a Scribd company logo
Nature Knows Best ?

How can biological systems be exploited in engineered systems ?
People
 Cathy Scott c.scott@napier.ac.uk
 Husna Osman ho12@hw.ac.uk
 Ioannis Polyzos i.polyzos@gmail.com
 Michael Matscheko mm@pervasive.jku.at
 Petros Papadopoulos petros.papadopoulos@acm.org
 Sarah Clayton s.clayton@napier.ac.uk
Outline
 Introduction
 Aim
 Scenarios
 Network Settings
 Results
 Fungal Colonies
 Conclusion
 Questions ?
Introduction
 Routing strategies are highly important for computer
 networks performance.
 The emergence of mobile/sensor networks raised several
 novel problems mainly due to resource constrained
 devices.
 Among others biological systems with the same
 characteristics have been exploited in order to solve
 several discrete optimization problems such as :
  Ant Colonies
  Fungal Colonies
Aim

 This case study focus in :
   Comparing Ant Colony Optimization(ACO) with other
   discrete optimization algorithms like Nearest Neighborhood
   Tree(NNT).
   Exploring a recently introduced systems inspired by Fungal
   Colonies

 3 scenarios have been used in order to measure :
 Efficiency , Scalability and Robustness
Scenarios
 Scenario 1 : Static Network
   Focus on the performance of NNT and ACO in a static
   network where there is no update frequency/proactive ants.
 Scenario 2 :
   Focus on the performance of NNT(no update) and ACO(with
   proactive ants) in a static network.
 Scenario 3:
   Focus on the performance of NNT(with update) and
   ACO(with proactive ants) in a dynamic environment (possible
   failures).
Network Settings
 Typical Network1 :
   Population : 50
   Distance : 0.09
 Futuristic Network 2 :
   Population: 400
   Distance : 0.04
Routing Overhead
10




                                            bug ?
8
                                            configuration ?
                                            more tests ?


6




4

                                                      ACO FA 
                                                      ACO FA 1
2
                                                      ACO FA 10
                                                      NNT UR --
                                                      NNT UR 5
0
     0   5   10   15   20   25   30   35         40
                                                      NNT UR 10
Delivery Ratio
 1




0.8




0.6
                                                       NNT UR --
                                                       NNT UR 5
                                                       NNT UR 10
0.4



                                                       ACO FA 1
                                                       ACO FA 10
0.2



                     corresponds to                    ACO FA 
                     routing overhead
 0
      0   5   10   15discrepancy 25
                            20          30   35   40
Delivery Ratio (Failure t=20)
 1




                                            low delivery rates:
0.8
                                             congestion
                                             reduce data load?

0.6

                                                     NNT UR 5
                                                     NNT UR 10
0.4




0.2
                                                     ACO FA 1
                                                     ACO FA 10
 0
      0   5   10   15   20   25   30   35       40
Delivery Ratio (400 nodes)
 1




                                             far too much data
0.8
                                              for large network
                                             rerun with
                                              decreased data load
                                              or increased radio
0.6
                                              data rate


0.4




0.2                                                   NNT UR 5
                                                      NNT UR 10
                                                      ACO FA 1
 0                                                    ACO FA 10
      0   5   10   15   20   25   30   35        40
Conclusions
  Ant algorithm struggles with (almost) static network
  scenarious
  Network congestion leads to low delivery ratio

Future Work:
  Investigate situation in more dynamic networks
    Volatile environment (high node failure rates)
    Mobile nodes
  Optimization of parameters
  Longer simulation timespan
Fungi Colonies
Fungal Colonies
Processes                     Transactions
Explore (sparse branching)
Exploit (dense branching)
Degenerate
Processes             Triggers

 Explore               Fitness function
 (sparse branching)
 Exploit               Data message
 (dense branching)
 Degenerate            Energy < pdegenerate
Mappings

SpeckSim             Fungal Colony

 Energy               Nutrients
 Data                 Mobile biomass
 Established route    Structural biomass
Message Type               Transmission type

 Data message               Broadcast
 Explore message            Broadcast
 Establish route message    Unicast
Network Flow
Thank you 




        Questions ?

More Related Content

Nature knows best

  • 1. Nature Knows Best ? How can biological systems be exploited in engineered systems ?
  • 2. People Cathy Scott c.scott@napier.ac.uk Husna Osman ho12@hw.ac.uk Ioannis Polyzos i.polyzos@gmail.com Michael Matscheko mm@pervasive.jku.at Petros Papadopoulos petros.papadopoulos@acm.org Sarah Clayton s.clayton@napier.ac.uk
  • 3. Outline Introduction Aim Scenarios Network Settings Results Fungal Colonies Conclusion Questions ?
  • 4. Introduction Routing strategies are highly important for computer networks performance. The emergence of mobile/sensor networks raised several novel problems mainly due to resource constrained devices. Among others biological systems with the same characteristics have been exploited in order to solve several discrete optimization problems such as : Ant Colonies Fungal Colonies
  • 5. Aim This case study focus in : Comparing Ant Colony Optimization(ACO) with other discrete optimization algorithms like Nearest Neighborhood Tree(NNT). Exploring a recently introduced systems inspired by Fungal Colonies 3 scenarios have been used in order to measure : Efficiency , Scalability and Robustness
  • 6. Scenarios Scenario 1 : Static Network Focus on the performance of NNT and ACO in a static network where there is no update frequency/proactive ants. Scenario 2 : Focus on the performance of NNT(no update) and ACO(with proactive ants) in a static network. Scenario 3: Focus on the performance of NNT(with update) and ACO(with proactive ants) in a dynamic environment (possible failures).
  • 7. Network Settings Typical Network1 : Population : 50 Distance : 0.09 Futuristic Network 2 : Population: 400 Distance : 0.04
  • 8. Routing Overhead 10 bug ? 8 configuration ? more tests ? 6 4 ACO FA ACO FA 1 2 ACO FA 10 NNT UR -- NNT UR 5 0 0 5 10 15 20 25 30 35 40 NNT UR 10
  • 9. Delivery Ratio 1 0.8 0.6 NNT UR -- NNT UR 5 NNT UR 10 0.4 ACO FA 1 ACO FA 10 0.2 corresponds to ACO FA routing overhead 0 0 5 10 15discrepancy 25 20 30 35 40
  • 10. Delivery Ratio (Failure t=20) 1 low delivery rates: 0.8 congestion reduce data load? 0.6 NNT UR 5 NNT UR 10 0.4 0.2 ACO FA 1 ACO FA 10 0 0 5 10 15 20 25 30 35 40
  • 11. Delivery Ratio (400 nodes) 1 far too much data 0.8 for large network rerun with decreased data load or increased radio 0.6 data rate 0.4 0.2 NNT UR 5 NNT UR 10 ACO FA 1 0 ACO FA 10 0 5 10 15 20 25 30 35 40
  • 12. Conclusions Ant algorithm struggles with (almost) static network scenarious Network congestion leads to low delivery ratio Future Work: Investigate situation in more dynamic networks Volatile environment (high node failure rates) Mobile nodes Optimization of parameters Longer simulation timespan
  • 14. Fungal Colonies Processes Transactions Explore (sparse branching) Exploit (dense branching) Degenerate
  • 15. Processes Triggers Explore Fitness function (sparse branching) Exploit Data message (dense branching) Degenerate Energy < pdegenerate
  • 16. Mappings SpeckSim Fungal Colony Energy Nutrients Data Mobile biomass Established route Structural biomass
  • 17. Message Type Transmission type Data message Broadcast Explore message Broadcast Establish route message Unicast
  • 19. Thank you Questions ?