Frame counter: Achieving Accurate and Real-Time Link
Estimation in Low Power Wireless Sensor Networks
Daibo Liu, Zhichao Cao, Mengshu Hou and Yi
Zhang
IPSN 2016
Link Estimation in Low Power Wireless Sensor Networks
Accurately count arrived data frame for each active link
Monitor all routing/non-routing links in real-time during active state
Wireless sensor networks General network structure Low power transmission mode
Low Power Listening Protocol
X-MAC protocol Short frame Receive early ack
Sender
Receiver
Sender
Receiver
Receive wakes up
Target Address
Listen for additional Data
Listen for additional Data
Ideal Approach in Low Power WSNs
State-of-the Art
Passive Estimation & Active Estimation
Ignored receiving information
at non-routing links
Ignoring the lost
frame in routing
link
Ignoring the lost frame
Treating it as one frame loss
Capture of CorruptedFrames
Observation 1: The RSSI of corrupted frames can be captured in real-time.
Observation 2: It is feasible to distinguish ZigBee from other 2.4GHz technologies with RSSI features.
Non Real-Time
Root cause: Asynchronous transmission & nocentral coordinator
Results: Routing selection based on outdated estimation
Actions: Except for routing beacon, ignoring neighbors’data frames; Inaccurately counting beacon frame.
Defect of State-of-the-art and Observation
Empirical Study of Low Power WSNs
Inaccuracy Root cause: No synchronization mechanism
Results: Overestimation
Actions: Ignoring all lost frames if at least one frame is received; Treat no frame reception as one loss.
New Challenges
Duty cycle and asynchronous radio work mode
Accurate and Real-Time Link Estimation
Data transmission is organized as repeated data frames
Count all arrived frames (decoded or not decoded)
Several nodes may successively transmit
Monitor all neighboring links in real-time
As a neighbor, it is difficult to know which frame is corrupted
During active state, the transmitted frames by any neighbor
should be counted
Sender information of not decoded data frames is unknown
Overview the Basic Idea
Tx
Neighbor
Rx
Decoded data frame
Lost data frame
Sampled RSSI sequence
Repeat data frame transmission until be ACKed1
Frames arrive at receiver (decoded or lost)2
Sampled received signal strength in time domain4
Decoded frames math corresponding RSSIs5
#1 #3
#3
Neighbor also overhears arriving frames3
Necessary Information for Accuracy Estimation
Ongoing Sender
Which node is transmitting data frame?
Total arrived frames
How many frames have arrived during the node’s active state?
Decoded frames
How many frames have been successfully decoded by the node?
Solution 1: Extracting ID from decoded frame
Solution: Using RSSI sequence to count
Solution: Counting it according to frame receiving event
Solution 2: Using RSSI Feature to infer
Necessary Information for Accuracy Estimation
Ongoing Sender
Which node is transmitting data frame?
Total arrived frames
How many frames have arrived during the node’s active state?
Decoded frames
How many frames have been successfully decoded by the node
Solution 1: Extracting ID from decoded frame
Solution: Using RSSI sequence to count
Solution: Counting it according to frame receiving event
Solution 2: Using RSSI Feature to inferPurposes: To know the ongoing sender, number of arrived frames during
radio active state, number of decoded frames.
Utilization of RSSI Sequence
Frame Transmission by sender
Sampled RSSI Sequence at
Receiver/Neighbor
Noise floor
Signal strength
Translated to Step Pulse Signal
Low level
High level
Rising edge
Falling edge
Procedure
Counting the detected pulses to represent the
number of arrived frames
Impacts on Accurate Data Frame Counting
Coexistent Interference in 2.4GHz ISM Band
Distinguish ZigBee Data Frame and ACK Frame
Frame Type On-air time Interval
Data Frame [576, 4256] μs Larger than 512μs
ACK Packet 352μs Between 192μs and 512μs
ZigBee Frame Identification
Shorter on-air time
longer on-air time
Feature #1: On-air time
Valid range of on-air
time
Feature #2: Frame interval
Shorter packet intervalFixed frame interval
Longer packet interval
ZigBee Frame Identification
Feature #3: PAPR (Peak-to-Average Ratio) Feature #4: RSSI < Noise floor
Flat sequence
Large variation
TRUE
FALSE
ZigBee
Determination of Transmitter
Extracting transmitter ID from decoded frame
Exploiting low power RSSI features
Fixed inter-frame interval
Inter-frame interval (Tifi) is fixed;System congestion backoff > Tifi.
Check|Tifi(k) - Tifi|≤ δ
Valid RSSI bias, 1dBm
Determine whether two successive frames
are transmitted by the same sender.
Parameter forframe segment
Computing average frame RSSI Ravg;Comparing Ravg with each neighbor A’s RSSI (R(A)) by:Steadily
averaged RSSI |Ravg – R(A)|≤ Rδ
If passes the check, A is a possible candidate transmitter. If
more than one candidate, adopt deferred determination.
Deferred Transmitter Determination
…. ….
….
…. ….
Neighbor A
Neighbor B
Neighbor C
timeline……..
Time unit, 1wakeup interval
A{1, 0, 1, 0, 0, 0}
B{0, 1, 0, 0, 0, 0}
C{0, 0, 0, 0, 1, 1}
Neighbor Transmitting Time Bitmap
Neighbor A
Neighbor B
Neighbor C
Neighbors transmits data frames in different time
Different bit corresponds
to different time.
Using averaged RSSI features and compressed transmitting time to
accurately determine the transmitter.
Implementation and Evaluation
Implementation Evaluation setup
TinyOS-2.1.1
Combining with LPL
Beneath collection tree protocol (CTP)
Multiple-hop networks
Indoor & outdoor testbeds
Comparing with the state-of-the-art
With/without coexisting interference
Correct False negative False positive
Radio 97.3% 0.8% 1.9%
Accuracy
Number of arrived frames Determination of transmitter
For segment vectors, more than
92.5% with <6% FN, and 89.4% with
<8% FP.
Overall accuracy
About (>) 60% segments are determined
by decoded frames, for the remainder:
Averaged RSSI Deferred determination
Accuracy 95.8% 98.9%
DF
Averaged RSSIDeferred
determinationDecoded frame
≈60%
Timeliness
Link Quality Update Period
97.9% links can be updated
within 200 seconds.
Comparing with 4-bit link estimator
Estimation window size: 5 frames
Indoor & outdoor testbed with multiple-hop data collection networks
Trickle controls routing probe
Data packet interval: 2 minutes
More than 50% links can not
be updated one time for 8
minutes by 4-bit
Collection Tree Protocol Performance
Network reliability, Energy consumption, Delay, and Path length
More reliable Less energy
Lower delayLess transmission
hops
Conclusion
Existing passive and active link estimators can not
achieve accurate and real-time link estimation
Implement on configurable indoor/ourdoor testbed
Validate performance
Resolve accurate and real-time link estimation in Low
Power WSNs
Using decoded frames to directly determine the transmitter
RSSI features for ZigBee frame identification and frame counting
Averaged RSSI and deferred determination to accurately infer
transmitter
Q&A