Realtime message scheduling in wireless sensor networks

1
Real-time message scheduling in wireless sensor
networks
Kavitha Balasubramanian Teaching Assistant, CprE
458/558 Dept. of Electrical and Computer
Engineering Iowa State University, Ames, IA 50011
2
Agenda

• Introduction
• Wireless Sensor Networks
• Scheduling Algorithms in Wireless Sensor Networks
• Implementation
• Conclusion

3
Agenda

• Introduction
• Wireless Sensor Networks
• Scheduling Algorithms in Wireless Sensor Networks
• Implementation
• Conclusion

4
Agenda

• Introduction
• Wireless Sensor Networks
• Scheduling Algorithms in Wireless Sensor Networks
• Implementation
• Conclusion

5
Agenda

• Introduction
• Wireless Sensor Networks
• Scheduling Algorithms in Wireless Sensor Networks
• Implementation
• Conclusion

6
Agenda

• Introduction
• Wireless Sensor Networks
• Scheduling Algorithms in Wireless Sensor Networks
• Implementation
• Conclusion

7
Introduction

• Real-time systems
• Hard real-time
• Guarantee deadlines
• Soft real-time
• Improve hit ratio
• Wireless sensor network applications require
  real-time support
• Surveillance and tracking
• Border patrol
• Fire fighting.

8
Wireless Sensor Networks

• Operate in Adhoc networks
• Communication patterns
• Local coordination
• Sensors coordinate with one another in order to
  aggregate data
• Sensor-base communication
• Sends data from the local group to the base
  station
• Real time sensor applications need timeliness
  guarantees
• Schedule messages based on deadlines
• Exploit spatial reuse of wireless channel
• Simultaneous transmissions
• Explicitly avoid collisions
• Prevent false blocking

9
Scheduling Algorithms in Wireless Sensor Networks

• RAP A Real-Time Communication Architecture for
  Large-Scale Wireless Sensor Networks
• Chenyang Lu Brian M. Blum Tarek F. Abdelzaher
  John A. Stankovic Tian He
• Scheduling messages with deadline in real time
  multi-hop wireless sensor networks
• Huan Li, Prashant Shenoy, Krithi Ramamritham

10
RAP(A Real-Time Communication Architecture for
Large-Scale Wireless Sensor Networks)

• Velocity monotonic scheduling
• Deadline and Distance aware
• Static monotonic velocity (SMV)
• Dynamic monotonic velocity (DMV)
• Higher velocity denotes higher priority

11
RAP(A Real-Time Communication Architecture for
Large-Scale Wireless Sensor Networks)

• Static monotonic velocity

• Dynamic monotonic velocity (DMV)

12
Scheduling messages with deadline in real time
multi-hop wireless sensor networks

• Application Background
• Definitions
• False Blocking
• Algorithm details
• Conclusions

13
Application Background

• Examples
• Searching for trapped people in a building on
  fire, Building map of an unknown environment
• Each robot carries sub-set of sensors and has
  wireless connectivity
• Communication over an adhoc network

14
Assumptions

• Global topology information available
• Messages to be transmitted available
• Routing information available

15
Definitions

• Arrival Time (AT)
• Time at which message arrives at a node
• Transmission Duration
• Difference between instants when first bit is
  sent out and last bit is received by receiver
• Data Validity
• Time interval for which data value produced by
  sensor is valid
• Also determined by start time of consuming task
• Effective Deadline (ed)
• Minimum of data validity deadline and start time
  of consuming task
• Latest Start time
• latest time by which a hop must start
  transmitting for it to reach its destination by
  effective deadline

16
Definitions

• Message needs to travel h hops from source to
  destination
• mi denotes transmission of message at the ith
  hop
• Pd(mi) transmission time incurred on remaining
  hops to destination.
• LST(mi) ed(m) pd(mi)

17
False Blocking

• One hop transmission
• 1 -gt 0 RTS
• 0 -gt 1 CTS
• 2 Receives RTS of 1 and is blocked during the
  transmission of m1
• 3 -gt 2 RTS
• 2 Blocked and does not respond
• 4 Receives RTS sent by 3 and is blocked
• 5 -gt 4 RTS
• 4 Blocked and does not respond
• Even though m1 and m3 can be transmitted
  simultaneously and do not interfere

18
False Blocking

• Deadline miss
• m1 0 to 2
• m2 2 to 7
• m3 7 to 9 (misses deadline)
• Parallel transmissions can reduce deadline misses

19
False Blocking

• Deadlock
• False blocking propagates
• Propagation is along circular
• path
• In cycle A,B,C,D,E,F,A
• every second node A,C,E is sending RTS to the
  next node B,D,F that is already blocked and
  because of this RTS, the previous node gets
  blocked

20
Goal

• Avoid collisions
• False blocking needs to be eliminated for meeting
  deadlines
• Do parallel transmissions (carefully) to meet
  deadline constraints
• Consider potential impact of scheduling messages
  on future message transmissions

21
Algorithm

• Channel reuse based on smallest LST first
• Partition message transmissions into disjoint
  sets
• Messages in one set can be transmitted together
• Transmission in one set should complete before
  next one begins
• Once message is scheduled at hop i, it is
  considered for scheduling at hop i1

22
Initial state
23
Divide into sets
m1
m3
m2
m4
Set1
Set2
Set3
24
Transmit Messages
m1
m3
m2
m4
Set1
Set2
Set3
m5
m6
25
Transmit Messages
m1
m3
m2
m4
Set1
Set2
Set3
m5
m6
m7
26
Transmit Messages
m1
m3
m2
m4
Set1
Set2
Set3
27
Calculate LST
m1
m3
m2
m4
Set1
Set2
Set3
28
Sort the Message queue
m1
m3
m2
m4
Set1
Set2
Set3
m7
m6
m5
m8
29
Repeat the process
m1
m3
m2
m4
Set1
Set2
Set3
m7
m6
m5
m8
30
Set Construction

• Condition to join a set
• Exploit parallelism
• Arrival time of the message lt Finish time
• Schedule within deadline
• Finish time of message is no greater than the
  effective deadline.
• No interference
• Current message transmission does not interfere
  with existing message transmissions in the set
• Does not violate scheduled transmissions
• Inserting transmission into the set does not
  cause deadline violations for already scheduled
  transmissions in other sets

31
Example
32
Example

• Message m1
• m1 selected first since it has smallest LST _at_time
  0
• S1 T(m1)
• s(S1) AT 0
• f(S1) 2 lt 6

33
Example

• Message m2
• m1 and m2 interfere.
• m2 cannot be added to S1.
• Create new set S2 T(m2)
• s(S2) f(S1) 2
• f(S2) s(S2) 5 2 5 7 lt 8
• S1 T(m1), S2 T(m2)

34
Example

• Message m3
• Condition 1 Exploit parallelism
• f(S1) 2 gt a(m3)
• Condition 2 Schedule within deadline
• f(m3) max(s(S1),a(m3)) tt(m3) 1 2 3 lt
  d(m3) 8
• Condition 3 No interference
• m3 and m1 do not interfere
• Condition 4 Does not violate scheduled
  transmissions
• Since f(m3) 3 gt f(S1) fnew(S1) f(m3) 3
  snew(S2) fnew(S1) 3 fnew(m2) 3 5 8

35
Example

• Therefore S1 is feasible for T(m3)
• Final schedule
• S1 T(m1),T(m3)
• S2 T(m2)
• m1 and m3 are transmitted in parallel followed by
  m2

36
Example

• Therefore S1 is feasible for T(m3)
• Final schedule
• S1 T(m1),T(m3)
• S2 T(m2)
• m1 and m3 are transmitted in parallel followed by
  m2

37
Conclusion

• Advantages
• Explicitly avoids collisions
• Doesnt inject infeasible packets into the system
• Exploits spatial channel reuse

38
Implementation

• Implemented in C
• Input
• Network topology
• Collision Matrix
• Distance Matrix
• Routing table
• Message Size
• Events

39
Implementation

• Results
• Comparison of performance of the RAP and the
  CR-SLF
• Deadline miss
• Impact of parameter settings
• Message size
• Network topology
• Distance
• Number of events

40
Summary

• RAP Distributed
• CR-SLF Centralized
• SLF without channel reuse is same as Dynamic
  Velocity Monotonic Scheduler
• CR-SLF performs better because of channel reuse
  and collision avoidance

41
Questions
42
References

• Scheduling Messages with Deadlines in Multi-hop
  Real-time Sensor Networks
• 11th IEEE Real-Time and Embedded Technology and
  Applications Symposium, San Francisco,
  California, March 2005.
• RAP A Real Time Communication Architecture for
  Large Scale Wireless Sensor Networks
• IEEE Real-Time Technology and Applications
  Symposium, San Jose, California, September 2002.