Transcript Self Organization and Energy Efficient TDMA MAC Protocol

Self Organization and Energy Efficient
TDMA MAC Protocol by Wake Up for
Wireless Sensor Networks
Zhihui Chen and Ashfag Khokhar
ECE/CS University of Illinois at Chicago
IEEE SECON 2004
Presented by Jeffrey
2005/8/2
NTU NSLAB
1
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
2
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
3
Wireless Sensor Networks (WSNs)
Are Unique
• Traffic rate is very low
– Typical communication frequency is at
minutes or hours level
• Sensor networks are battery powered and
recharging is usually unavailable
– Energy is an extremely expensive resource
2005/8/2
NTU NSLAB
4
Wireless Sensor Networks (WSNs)
Are Unique
• Sensor nodes are generally stationary
after their deployment
• Sensor nodes coordinate with each other
to implement a certain function
– Traffic is not randomly generated as those in
mobile ad hoc networks
2005/8/2
NTU NSLAB
5
Previous Energy-Efficient MAC
Protocols for WSNs
• “An Energy-Efficient MAC Protocol for Wireless
Sensor Networks”
– W. Ye, J. Heidemann and D. Estrin
– IEEE INFOCOM ’02
– S-MAC (10% S-MAC)
2005/8/2
NTU NSLAB
6
Previous Energy-Efficient MAC
Protocols for WSNs
• “An Adaptive Energy-Efficient MAC Protocol for
Wireless Sensor Networks”
– T. Dam and K. Langendoen
– ACM SENSYS ’03“
– T-MAC
2005/8/2
NTU NSLAB
7
Concentrate traffic to fixed periods?
• Increases contention probability
• Incurs unnecessary retransmissions
• S-MAC proposes to perform RTS/CTS
handshake procedure
• Duty rate or portion of listening period of
S-MAC should be carefully chosen
• T-MAC adapts duty cycle to the traffic rate
2005/8/2
NTU NSLAB
8
Previous Energy-Efficient MAC
Protocols for WSNs
• “Energy-Efficient, Collision-Free Medium Access Control
for Wireless Sensor Networks”
– V. Rajendran, K. Obraczka and J.J. Garcia-Luna-Aceves
– ACM SENSYS ’03
– TRAMA
2005/8/2
NTU NSLAB
9
Scheduling
• Data transmissions are scheduled in advance to
avoid contention
• TDMA-W
–
–
–
–
2005/8/2
TDMA-Wakeup
Each node is assigned two slots
Transmission/Send slot (s-slot)
Wakeup slot (w-slot)
NTU NSLAB
10
5
6
1
ZZZ
ZZZ
2
Z
Z
3
ZZZ
ZZZ
4
Z
Z
ZZ
ZZ
ZZ
ZZ
7
8
1
2
3
4
5
6
7
8
1
2
Listen
Wake
up
Nd3
Send
ZZZ
ZZZ
ZZZ
ZZZ
ZZZ
Listen
ZZZ
ZZZ
ZZZ
Wake
up
Nd3
ZZZ
Send
ZZZ
ZZZ
ZZZ
ZZZ
ZZZ
Listen
ZZZ
ZZZ
Wake
up
Nd4
Listen
Listen
Listen
Send
ZZZ
ZZZ
ZZZ
ZZZ
Listen
ZZZ
ZZZ
Listen
Wake
up
Nd6
ZZZ
ZZZ
Listen
Send
ZZZ
ZZZ
Listen
ZZZ
ZZZ
ZZZ
ZZZ
ZZZ
ZZZ
ZZZ
Send
ZZZ
Listen
ZZZ
ZZZ
ZZZ
ZZZ
ZZZ
Listen
Listen
ZZZ
ZZZ
Listen
ZZZ
Z
ZZ
ZZ
ZZZ
Listen
Wake
up
Nd6
ZZ
ZZZ
ZZZ
Listen
5 Z
2005/8/2
6 Z
NTU NSLAB
ZZZ
11
ZZZ
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
12
Channel and Traffic Assumption
• Ideal physical layer
– The only reason for packet loss is
transmission contention
– No packet loss due to noise
• Three types of traffic pattern
– One-to-all broadcast
– All-to-one reduction
– One-hop random traffic
2005/8/2
NTU NSLAB
13
Channel and Traffic Assumption
• A TDMA-W frame lasts for Tframe seconds
• Tframe is known to all nodes and is preset
before deployment
• A TDMA-W frame is divided into slots
• Each node is assigned one slot for
transmission and one slot for wakeup
• Networks are synchronized
2005/8/2
NTU NSLAB
14
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
15
Self-Organization
• Assign time slots to the sensors within each
TDMA-W frame
• Assume sensor networks has data rate of 1
Mbps
• Transmission of a 512 byte packet occupies the
channel for about 3.9 ms
• Assume a TDMA-W frame of 1 second divided
into 256 slots
– Each slot is of 3.9 ms
– Capable of communicating 512 bytes
2005/8/2
NTU NSLAB
16
Self-Organization Scheme
1. Each node randomly selects a slot with uniform
probability among all slots to be its s-slot
2. During its selected s-slot, each node
broadcasts
– Its node ID
– Its s-slot number
– Its one-hop neighbors’ IDs and their s-slot
assignments
– Slot number of any s-slot during which this node has
identified a collision
2005/8/2
NTU NSLAB
17
Self-Organization Scheme
3. When a node is not transmitting, it turns
on its receiver circuit and listens to the
traffic from neighbors
•
The node should record all the information
being broadcast by all its neighbors
•
•
2005/8/2
Their s-slot assignments and their node IDs
The slot number of any slot being broadcast as a
collision-prone slot
NTU NSLAB
18
Self-Organization Scheme
4. If a node determines that
– it is involved in a collision
– or finds out that one of its two-hop neighbors
has the same s-slot
– It then randomly selects an unused slot and
go to step 2
2005/8/2
NTU NSLAB
19
Self-Organization Scheme
5. If
– no new nodes are joining in
– or s-slot assignments are not changing
– or no collisions are detected for a certain
period
– It implies all neighbor nodes are found and
all the s-slots are final
2005/8/2
NTU NSLAB
20
Self-Organization Scheme
6. Each node broadcasts the s-slot
selections of their two-hop neighbors.
•
•
Each node identifies an unused slot or any
s-slot being used by the nodes beyond its
two-hop neighbors and declares it as its wslot
Note that w-slots need not be unique
7. Each node broadcasts its w-slot and the
self-organization is complete
2005/8/2
NTU NSLAB
21
Can Detect Any Two-hop Collisions
2005/8/2
NTU NSLAB
22
Undetectable One Hop Collision
• To solve this problem
– Let a node go to the listening mode in its
assigned s-slot with a probability
2005/8/2
NTU NSLAB
23
Deadlock
• To listen during s-slot with a probability
• To set a collision counter
2005/8/2
NTU NSLAB
24
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
25
TDMA-W Channel Access Protocol
1. Each node maintains a pair of counters for
every neighbors
–
–
–
Outgoing counters
Incoming counters
These counters are preset to an initial value
2. If no outgoing data is sent to a node in a
TDMA-W frame
–
–
2005/8/2
The node decrements the corresponding outgoing
counter by one
Otherwise it resets the counter to the initial value
NTU NSLAB
26
TDMA-W Channel Access Protocol
3. If no incoming data is received from a
neighboring node in a TDMA-W frame
– The node decrements the corresponding
incoming counter by one
– If the counter is less than or equal to zero,
the node stop listening to that slot starting
from next TDMA-W frame
2005/8/2
NTU NSLAB
27
TDMA-W Channel Access Protocol
4. If a outgoing data transmission request
arrives
– The node first checks the outgoing counter
– If the counter is greater than zero, then the
link is considered active and the packet can
be sent out during the s-slot
– If the counter is less than or equal to zero, a
wakeup packet is sent out during the w-slot
of the destination node prior to the data
transmission
2005/8/2
NTU NSLAB
28
TDMA-W Channel Access Protocol
5. If a node receives a wakeup packet in its
w-slot
– It turns itself on during the s-slot
corresponding to the source node ID
contained in the wakeup packet
– If a collision is detected in the w-slot
•
•
2005/8/2
More than one node intends to send data
The node then searches all its neighbors for
incoming traffic
NTU NSLAB
29
Packet Content
• Wakeup packet contains only the source
and the destination information
• Data packet may only contain the
destination information
– Omit source ID since the source ID is
determined by the s-slot
2005/8/2
NTU NSLAB
30
Broadcast
• If a data packet is to be broadcast to
multiple nodes
– The destination address contains a special
identifier to mark it as a broadcast message
– Before sending a broadcast data packet
• The node should wakeup all its neighbors that
intend to receive this packet
• In the case when multiple users share the same wslot
– The destination field of the wakeup message should also
be set to a broadcast address
2005/8/2
NTU NSLAB
31
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
32
Performance Analysis of TDMA-W
• Let us fix the position of the w-slot
2005/8/2
NTU NSLAB
33
Average Delay
2005/8/2
NTU NSLAB
34
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
35
Deployment of Sensor Nodes
• Nodes are deployed randomly in a 500x500 sq.
ft. area
• Communication range is 100 feet for all nodes
• Assume an IEEE 802.11 basic rate of 1 Mbps as
the physical layer transmission rate
• Slot length is set to be 4 ms
– Long enough for transmitting a 512-byte packet
• Tframe is set to one second
– A TDMA-W frame has 250 slots
2005/8/2
NTU NSLAB
36
Simulation Results of SelfOrganization Protocol
2005/8/2
NTU NSLAB
37
Power Consumption
• Power consumption
– Transmission : Receiving/Listening : Sleeping
= 1.83 : 1 : 0.001
• 10% S-MAC
– Use RTS/CTS frames to reserve channel for node-tonode traffic
– Use ACK packet to acknowledge the successful
transmission
– If data or ACK packet is corrupted by collision, the
data packet is retransmitted
2005/8/2
NTU NSLAB
38
Power Consumption
• The network is synchronized
– All the nodes become active at the same time
• All data packets are fixed to be 256 bytes
in length
• Control packets (RTS, CTS, ACK in SMAC and Wakeup packet in TDMA-W) are
about 20 bytes in length
• Assume energy consumption for a control
packet is 1/10 of a data packet
2005/8/2
NTU NSLAB
39
Power Consumption
• Initial value for counters is set to 3
• Transmission buffer length is set to 50
packets
• Both TDMA-W and S-MAC are run for 10
minutes
2005/8/2
NTU NSLAB
40
Power Consumption of One-Hop
Random Traffic
2005/8/2
NTU NSLAB
41
Delay of Random One-Hop Traffic
2005/8/2
NTU NSLAB
42
Delay of All to One Reduction
Operation Traffic
2005/8/2
NTU NSLAB
43
Outline
• Introduction
• Channel and Traffic Assumption
• TDMA-W: Details
– Self-Organization
– TDMA-W Channel Access Protocol
– Performance Analysis of TDMA-W
• Simulation Results
• Conclusion
2005/8/2
NTU NSLAB
44
Conclusion
• Efficient protocols TDMA-W for selforganization and channel access control in
wireless sensor networks are proposed
• Proposed protocols were verified using
extensive simulations
• Proposed protocols only consume 1.5% to
15% power of 10% S-MAC
– 6 to 67 times better than 10% S-MAC
2005/8/2
NTU NSLAB
45
Conclusion
• Proposed scheme responds to the event
with a delay comparable to S-MAC for
one-hop traffic
• Proposed protocol is collision free for data
traffic so reliable transmission is
guaranteed for all types of traffic
2005/8/2
NTU NSLAB
46
Comments
• Strength
– Great improvement in the power consumption
• Weakness
– Verify results by using simulation (MATLAB)
with not so practical assumptions
– Delay could be significant
– Scalability would be poor
• Large overhead in memory
2005/8/2
NTU NSLAB
47
Thank you very much for
your attention!
2005/8/2
NTU NSLAB
48