Swarm Intelligent Networking

1
Swarm Intelligent Networking

• Martin Roth
• Cornell University
• Wednesday, April 23, 2003

2
What is Swarm Intelligence?

• Swarm Intelligence (SI) is the local interaction
  of many simple agents to achieve a global goal
• Emergence
• Unique global behavior arising from the
  interaction of many agents
• Stigmergy
• Indirect communication
• Generally through the environment

3
Properties of Swarm Intelligence

• Properties of Swarm Intelligence are
• Agents are assumed to be simple
• Indirect agent communication
• Global behavior may be emergent
• Specific local programming not necessary
• Behaviors are robust
• Required in unpredictable environments
• Individuals are not important

4
Swarm Intelligence Example

• The food foraging behavior of ants exhibits swarm
  intelligence

5
Principles of Swarm Intelligence

• What makes a Swarm Intelligent system work?
• Positive Feedback
• Negative Feedback
• Randomness
• Multiple Interactions

6
SI Positive Feedback

• Positive Feedback reinforces good solutions
• Ants are able to attract more help when a food
  source is found
• More ants on a trail increases pheromone and
  attracts even more ants

7
SI Negative Feedback

• Negative Feedback removes bad or old solutions
  from the collective memory
• Pheromone Decay
• Distant food sources are exploited last
• Pheromone has less time to decay on closer
  solutions

8
SI Randomness

• Randomness allows new solutions to arise and
  directs current ones
• Ant decisions are random
• Exploration probability
• Food sources are found randomly

9
SI Multiple Interactions

• No individual can solve a given problem. Only
  through the interaction of many can a solution be
  found
• One ant cannot forage for food pheromone would
  decay too fast
• Many ants are needed to sustain the pheromone
  trail
• More food can be found faster

10
Swarm Intelligence Conclusion

• SI is well suited to finding solutions that do
  not require precise control over how a goal is
  achieved
• Requires a large number of agents
• Agents may be simple
• Behaviors are robust

11
SI applied to MANETs

• An ad hoc network consists of many simple
  (cooperative?) agents with a set of problems that
  need to be solved robustly and with as little
  direct communication as possible
• Routing is an extension of Ant Foraging!
• Ants looking for food
• Packets looking for destinations
• Can routing be solved with SI?
• Can routing be an emergent behavior from the
  interaction of packets?

12
SI Routing Overview

• Ant-Based Control
• AntNet
• Mobile Ants Based Routing
• Ant Colony Based Routing Algorithm
• Termite

13
SI Routing Overview

• Ant-Based Control
• AntNet
• Mobile Ants Based Routing
• Ant Colony Based Routing Algorithm
• Termite

14
Ant-Based Control Introduction

• Ant Based Control (ABC) is introduced to route
  calls on a circuit-switched telephone network
• ABC is the first SI routing algorithm for
  telecommunications networks
• 1996

15
ABC Overview

• Ant packets are control packets
• Ants discover and maintain routes
• Pheromone is used to identify routes to each node
• Pheromone determines path probabilities
• Calls are placed over routes managed by ants
• Each node has a pheromone table maintaining the
  amount of pheromone for each destination it has
  seen
• Pheromone Table is the Routing Table

16
ABC Route Maintenance

• Ants are launched regularly to random
  destinations in the network
• Ants travel to their destination according to the
  next-hop probabilities at each intermediate node
• With a small exploration probability an ant will
  uniformly randomly choose a next hop
• Ants are removed from the network when they reach
  their destination

17
ABC Routing Probability Update

• Ants traveling from source s to destination d lay
  ss pheromone
• Ants lay a pheromone trail back to their source
  as they move
• Pheromone is unidirectional
• When a packet arrives at node n from previous hop
  r, and having source s, the routing probability
  to r from n for destination s increases

18
ABC Routing Probability Update

• Dp determined by age of packet
• Probabilities remain normalized

19
ABC Route Selection (Call Placement)

• When a call is originated, a circuit must be
  established
• The highest probability next hop is followed to
  the destination from the source
• If no circuit can be established in this way, the
  call is blocked

20
ABC Initialization

• Pheromone Tables are randomly initialized
• Ants are released onto the network to establish
  routes
• When routes are sufficiently short, actual calls
  are placed onto the network

21
ABC Conclusion

• Only the highest probability next hop is used to
  find a route
• Probabilities are changed according to current
  values and age of packet

22
Reference

• R. Schoonderwoerd, O. Holland, J. Bruten, L.
  Rothkranz, Ant-based load balancing in
  telecommunications networks, 1996.

23
SI Routing Overview

• Ant-Based Control
• AntNet
• Mobile Ants Based Routing
• Ant Colony Based Routing Algorithm
• Termite

24
AntNet Introduction

• AntNet is introduced to route information in a
  packet switched network
• AntNet is related to the Ant Colony Optimization
  (ACO) algorithm for solving Traveling Salesman
  type problems

25
AntNet Overview

• Ant packets are control packets
• Packets are forwarded based on next-hop
  probabilities
• Ants discover and maintain routes
• Internode trip times are used to adjust next-hop
  probabilities
• Ants are sent between source-destination pairs to
  create a test and feedback signal system

26
AntNet Route Maintenance(F)

• Forward Ants, F, are launched regularly to random
  destinations in the network
• F maintains a list of visited nodes and the time
  elapsed to arrive there
• Forward Ant packet grows as it moves through the
  network
• Loops are removed from the path list
• F is routed according to next-hop probability
  maintained in each nodes routing table
• A uniformly selected next hop is chosen with a
  small exploration probability
• If a particular next hop has already been
  visited, a uniformly random next hop is chosen

27
AntNet Route Maintnence(B)

• When F arrives at its destination, a Backward
  Ant, B, is returned to the source
• B follows the reverse path of F to the source
• At each node, B updates the routing table
• Next-hop probability to the destination
• Trip time statistics to the destination
• Mean
• Variance

28
AntNet Routing

• Data packets are routed using the next-hop
  probabilities
• Forward ants are routed at the same priority as
  data packets
• Forward Ants experience the same congestion and
  delay as data
• Backward ants are routed with higher priority
  than other packets

29
AntNet Conclusion

• AntNet is a routing algorithm for datagram
  networks
• Explicit test and feedback signals are
  established with Forward and Backward Ants
• Routing probabilities are updated according to
  trip time statistics

30
AntNet Reference

• G. Di Caro, M. Dorigo, Mobile Agents for Adaptive
  Routing, Technical Report, IRIDIA/97-12,
  Universit Libre de Bruxelles, Beligium, 1997.

31
SI Routing Overview

• Ant-Based Control
• AntNet
• Mobile Ants Based Routing
• Ant Colony Based Routing Algorithm
• Termite

32
Mobile Ants-Based Routing Intro

• Mobile Ants-Based Routing (MABR) is a MANET
  routing algorithm based on AntNet
• Location information is assumed
• GPS

33
MABR Overview

• MABR consists of three protocols
• Topology Abstracting Protocol (TAP)
• Simplifies network topology
• Mobile Ants-Based Routing (MABR)
• Routes over simplified topology
• Straight Packet Forwarding (SPF)
• Forward packets over simplified topology

34
MABR Topology Abstracting Protocol

• TAP generates a simplified network topology of
  logical routers and logical links
• All individual nodes are part of a logical router
  depending on their location
• A single routing table may be distributed over
  all nodes that are part of a logical router

35
MABR TAP

• Zones are created, each containing more logical
  routers than the last
• Zones are designated by their location
• Logical links are defined to these zones

36
MABR Routing

• An AntNet-like protocol with Forward and Backward
  ants is applied on the logical topology supplied
  by TAP
• Forward ants are sent to random destinations
• Ants are sent to the zones containing these
  destinations
• Ants collect path information during their trip
• Backward ants distribute the path information on
  the way back their source
• Logical link probabilities are updated

37
MABR Routing
38
MABR Straight Packet Forwarding

• Straight Packet Forwarding is responsible for
  moving packets between logical routers
• Any location based routing protocol could be used
• MABR is responsible for determining routes around
  holes in the network
• SPF should not have to worry about such situations

39
MABR Conclusion

• The network topology is abstracted to logical
  routers and links
• TAP
• Routing takes place on the abstracted topology
• MABR
• Packets are routed between logical routers to
  their destinations
• SPF
• MABR is still under development
• Results are not yet available

40
SI Routing Overview

• Ant-Based Control
• AntNet
• Mobile Ants Based Routing
• Ant Colony Based Routing Algorithm
• Termite

41
Ant Colony Based Routing Overview

• Ant-Colony Based Routing (ARA) uses pheromone to
  determine next hop probability
• Employs a flooding scheme to find destinations

42
ARA Route Discovery

• To discover a route
• A Forward Ant, F, is flooded through the network
  to the destination
• A Backward Ant, B, is returned to the source for
  each forward ant received

43
ARA Route Discovery

• Reverse routes are automatically established as
  forward ants move through the network
• Backward ants reinforce routes from destination
  to source

44
ARA Routing

• Next Hop Probabilities are determined from the
  pheromone on each neighbor link

45
ARA Pheromone Update

• When a packet is received from r at n with source
  s and destination d
• r updates its pheromone table
• n updates its pheromone table

46
ARA Pheromone Decay

• Pheromone is periodically decayed according to a
  decay rate, t

47
ARA Loop Prevention

• Loops may occur because route decisions are
  probabilistic
• If a packet is received twice, an error message
  is returned to the previous hop
• Packets identified based on source address and
  sequence number
• The previous hop sets Pn,d 0
• No more packets to destination d will be sent
  through next hop n

48
ARA Route Recovery

• A route error is recognized by the lack of a
  next-hop acknowledgement
• The previous hop node sets Pn,d 0
• An alternative next hop is calculated
• If no alternative next hop exists, the packet is
  returned to previous hop
• A new route request is issued if the data packet
  is returned to the source

49
ARA Conclusion

• ARA is a MANET routing algorithm
• Flooding is used to discover routes
• Automatic retransmit used to recover from a route
  failure
• Packet backtracking used if automatic retransmit
  fails
• Next Hop probability proportional to pheromone on
  each link

50
ARA Reference

• M. Gunes, U. Sorges, I. Bouaziz, ARA The
  Ant-Colony Based Routing Algorithm for MANETs,
  2003.

51
SI Routing Overview

• Ant-Based Control
• AntNet
• Mobile Ants Based Routing
• Ant Colony Based Routing Algorithm
• Termite

52
Termite Overview

• Termite is a MANET routing algorithm
• Termite uses pheromone to produce next-hop
  probabilities
• Random routing
• Termite aims to reduce control traffic
• Termite should scale across network size and
  volatility

53
Termite Routing

• Each packet is forwarded probabilistically based
  on the amount of destination pheromone on each
  neighbor link
• F, K used to tune the routing probabilities
• No packet is routed out the link it arrived on

54
Termite Pheromone Update

• When a packet arrives at a node n from previous
  hop r originally from source s, n updates it
  Pheromone Table

55
Termite Pheromone Decay

• Pheromone is periodically decayed according to a
  decay rate, t

56
Termite Route Recovery

• If a transmission to a neighbor fails
• The neighbor is removed from the Pheromone Table
• An alternative next-hop is calculated and the
  packet is resent
• If no alternative exists, the packet is dropped

57
Termite Route Discovery(RREQ)

• If a node does not contain a needed destination
  in its pheromone table, a route request is issued
• A route request (RREQ) packet follows a random
  walk through the network until a node is
  encountered containing some destination pheromone
• A route reply (RREP) is returned to the source

58
Termite Route Discovery(RREP)

• A route reply (RREP) packet follows the pheromone
  trail normally back to the RREQ source
• The source of the RREP is the requested node,
  regardless of which node actually originates the
  packet
• The requested nodes pheromone is automatically
  spread through the network

59
Termite

• Termite minimizes control traffic by allowing all
  packets to explore the network
• Path discovery uses random walk
• Route Discovery packets are unicast

60
Open Issues

• Termite still has many open questions
• How to automatically determine routing parameters
  based on local information
• Decay rate, t
• Seed rate and distance
• Number of RREQs per Route Request
• How good is random walk route discovery
• How exactly are the various parameters related?
  Can some be determined from others? How do they
  affect performance?

61
Simulation Implementation
62
Simulation Environment

• 10 m transmission radius
• 1 Mbps channel
• 64B data packets
• CBR source
• 2 packets per second with acknowledgement

63
Network Performance vs. Mobility
64
Path Length vs. Mobility
65
Next Hop PDF vs. Mobility
66
Termite Reference

• M. Roth, S. Wicker, Termite Emergent Ad-Hoc
  Networking, 2003.

67
SI Advantages

• SI based algorithms generally enjoy
• Multipath routing
• Probabilistic routing will send packets all over
  the network
• Fast route recovery
• Packets can easily be sent to other neighbors by
  recomputing next-hop probabilities
• Low Complexity
• Little special purpose information must be
  maintained aside from pheromone/probability
  information

68
More SI Advantages

• Scalability
• As with any colonies numbering in the millions,
  SI algorithms can potentially scale across
  several orders of magnitude
• Distributed Algorithm
• SI based algorithms are inherently distributed

69
SI Disadvantages

• SI also suffers from
• Directional Links
• Bidirectional links are generally assumed by
  using reverse paths
• Novelty
• SI is a relatively new approach to routing. It
  has not been characterized very well, analytically

70
Swarm Intelligence Conclusion

• The fundamental idea behind using SI for routing
  in MANETs is to use the interaction of many
  packets to generate routing tables while
  minimizing the use of explicit routing packets
• The arrival of packets is observed, which
  influences next-hop routing probabilities
• Critical packets may include specialized ant
  packets or all packets