slides - Network and Systems Laboratory

Report
Low Power, Low Delay:
Opportunistic Routing meets Duty
Cycling
Olaf Landsiedel 1, Euhanna Ghadimi2, Simon Duquennoy3, Mikael Johansson2
1Chalmers University of Technology, Sweden
2KTH Royal Institute of Technology, Sweden
3Swedish Institute of Computer Science (SICS), Sweden
IPSN 2012
Presenter: SY
This Paper
• Opportunistic Routing for wireless sensor
network
– Duty cycled nodes
• Benefits
– Improve energy efficiency
– Reduce end-to-end delay
– Increase resilience to link dynamics
Unicast Routing in Duty‐Cycled WSNs
• Routing protocol: selects next hop
• MAC: wait for next hop to wakeup
– Assume: no synchronization
Unicast Routing in Duty‐Cycled WSNs
• Routing protocol: selects next hop
• MAC: wait for next hop to wakeup
– Assume: no synchronization
Opportunistic Forwarding
• The node that
– Wakes up first
– Successflly receives the packet
– Provides routing progress
• Forward the packet
Outline
• System design
• Evaluation
• Conclusion
DODAG
• Topology: DODAG
– Destination oriented directed acyclic graph
• Requirements of routing metric
– Builds loop free DODAG
– Minimize energy: radio-on time
– Minimize delay
Expected Duty Cycled Wakeups (EDC)
Single hop EDC
Link quality
• Single hop EDC: 1/(sum of neighbors link quality)
– Left case: A has a single neighbor with a perfect link,
its single hop EDC is 1/1 = 1;
– Right case: A has two neighbors both having perfect
links, its single hop EDC is 1/(1 + 1) = 0:5;
– Middle case: A has two neighbors with link qualities 1
and 0.25. Its single hop EDC is 1/(1 + 0:25) = 0:8.
Overall EDC
Single hop EDC
EDC of neighbor
Weight
• Overall EDC
– Sum of single hop EDC and neighbors’ EDC
• Which neighbors to include?
– Forwarder set
Forwarders Set
• Sort neighbor nodes by EDC
• Add one by one (from lowest)
• Find minimum EDC
Forwarding Cost
• Forwarding cost w
– Constant value, transmission penalty
– Increase w decrease forwarders set
• Fewer hops to destination
• Increase delay and energy consumption
– Too low: increase the risk of routing loop
• To balance delay and energy with routing
progress and stability
Link Estimation
• Link quality = (Rate of packet
overheard)/(forwarding rate)
• Rate of packet overheard
– Wakeup, listen to the radio, record packet
overheard
• Forwarding rate
– Header field contain the average forwarding rate
• Bootstrap
– Probing during initialization
Unique Forwarder
• Make sure only one node forward the packet
1. Majority of cases  only one receiver
Unique Forwarder – Cont.
• Coordination algorithm
– Demand a single ACK
• If (sender) receives multiple ACK
– Resend the packet
• If (receiver) detect link-layer duplicate
– Send second ACK with 50% probability
– Data transmission overhearing
• If overhears same packet, cancels transmission
– Network-layer duplication detection
• Detect duplication at network layer
Outline
• System design
• Evaluation
• Conclusion
Setup
• Testbed
– Indriya(Singapore): 120 nodes
– Twist(Berlin): 96 nodes
• Compare
– CTP
• Metrics
– Delay, Duty cycle, # of TX nodes, Reliability
• Implementation
– TinyOS, default MAC
– Wakeup every 2s (optimal for CTP)
– Randomly generate a packet every 4 minutes
System Calibration
• Choose w=0.1
Indriya, 0 dBm Tx Power
Indriya, -10 dBm Tx Power
Twist, 0 dBm Tx Power
Impart of Churn
• Remove average 10 nodes every 15 minutes
• Reduce from 120 to 30 nodes
Convergence
Wakeup Interval
Outline
• System design
• Evaluation
• Conclusion
Discussion And Limitation
• Works best at high network density
• Optimal at lower wakeup rates
– Compare to CTP
• At high wakeup rate
– CTP and ORW are similar
• Not well suited for high throughput
applications
Conclusion
• New routing metric
– Taken energy into account
• Real implementation
– Previous works mostly analytical and simulation
• Paper writing
– A bit harder to get the big picture

similar documents