Victoria Manfredi, Jim Kurose, Naceur Malouch,

1
Separation of Sensor Control and Data in
Closed-Loop Sensor Networks

• Victoria Manfredi, Jim Kurose, Naceur Malouch,
• Chun Zhang, Michael Zink
• SECON 2009

2
Outline

• Why separate sensor control and data?
• Closed-Loop Sensor Networks
• Meteorological Application
• Network, Sensing, Tracking Models
• Simulation Results
• Summary
• Future Work

• Why separate sensor control and data?
• Closed-Loop Sensor Networks
• Meteorological Application
• Network, Sensing, Tracking Models
• Simulation Results
• Summary
• Future Work

3
Why separate sensor control and data?

• Sensor network
• Closed-loop sensor network

Bursty, high-bandwidth data
Many-to-one routing to sink
Congestion
Wireless links
Data
Data spatially, temporally redundant
Prefer to delay, drop data
Sensor Controls
How does prioritizing sensor control traffic over
data traffic impact application-level performance?
4
Why separate sensor control and data?
Related Work

• Service differentiation for different classes of
  traffic
• e.g., Fredj et al, Sigcomm 2001
• ? Do not consider effects of prioritizing only
  sensor control in a sensor network
• Prioritizing network control
• e.g., SS7, ATM, Kyasanur et al, Broadnets 2005
  
• ? Our focus prioritizing sensor control
• Networked control systems
• e.g., Lemmon et al, SenSys 2003
• data/sensor control are measurements/feedback of
  classical control system
• ? We assume amount of data ?? sensor control

5
Outline

• Why separate sensor control and data?
• Closed-Loop Sensor Networks
• Meteorological Application
• Network, Sensing, Tracking Models
• Simulation Results
• Summary
• Future Work

6
Closed-loop Sensor Networks

• Prioritizing sensor control
• impact on packet delays?
• impact on data collected?
• Control loop delay

Priority control delay
Data delay
FIFO control delay
Data
Data from control k
Data from control k-1
Control
k
k-1
k1
? Update interval
Small ? data delay, large ? control delay ? more
data collected in time to compute next sensor
control
7
Better Quality Data

• More data samples
• Cramer-Rao bound
• SD?(W) 1 / ?n I
• accuracy ? sub-linearly with n
• Effect of data packet drops?
• accuracy ? sub-linearly with n

Radars, Sonars, Cameras,
Fisher information
Std Dev of W from ?
of iid samples
Compute unbiased estimator W (sample mean) of
parameter ? (population mean)
Sensing accuracy ? and ? slowly with of samples
8
Outline

• Why separate sensor control and data?
• Closed-Loop Sensor Networks
• Meteorological Application
• Network, Sensing, Tracking Models
• Simulation Results
• Summary
• Future Work

9
CASA

• Collaborative Adaptive Sensing of the Atmosphere
• dense (sensor) network of low-power
  meteorological radars
• observe severe weather in lower 3km of atmosphere
• Collaborative
• multiple radars coordinated
• Adaptive
• can focus beam on phenomena

CASA radar network is a closed-loop sensor
network
10
Storm Tracking Application 3 Coupled Models
ld
?
?
Network model control, data delays, depend on
scheduling (FIFO, priority) Sensing model
given scan, quantity and quality of data,
estimated storm location Tracking model
predict storm location based on current, past
estimates and observations using Kalman filters
ld
?
Timeliness of control, data affects amount of
sensed data gathered
ld
lc
Quality of estimated storm location affects
tracking
(xk,yk)
(xk-1,yk-1)
Quality of tracking affects scan angle, quality
of estimates
11
Network Model
ld
?
?
ld
Obtain sensor control and data packet delays

• Wireless network
• radar data sent to control center, sensor control
  back to radars
• much more data traffic than sensor control
  traffic
• Delays at bottleneck link dominate control-loop
  delay

?
ld
lc
?control
Deterministic arrivals
?
?data
?other
Bursty arrivals

Obtain delays for FIFO, priority queuing using
simulation
12
Sensing Model
Convert packet delays into sensing error

• Radar
• transmits pulses to estimate reflectivity at
  point in space
• Reflectivity
• of particles in volume of atmosphere
• standard deviation,

radar SNR
where N c (D - (ab))?/?q
scan angle width
sensing time

Smaller angle, longer time sensing ? lower
sensing error
13
Tracking Model
Convert sensing error into location error,
perform tracking
(xk,yk)

• Location of storm centroid
• equals location of peak reflectivity
• standard deviation,
• Kalman filters
• generate trajectory of storm centroid
• track storm centroid

(xk-1,yk-1)
distance from radar
mid-range reflectivity value
?z used in measurement covariance matrix
Goal track storm centroid with highest possible
accuracy
14
Outline

• Why separate sensor control and data?
• Closed-Loop Sensor Networks
• Meteorological Application
• Network, Sensing, Tracking Models
• Simulation Results
• Summary
• Future Work

15
Data Quantity vs Quality
360? scans, ? 5sec, very bursty traffic
CDF
FIFO achieves at least 80 as many samples as
priority 80 of time
Priority has at least 90 as much uncertainty as
FIFO 90 of the time

NFIFO / Npriority

?r,Priority / ?r,FIFO
During times of congestion, prioritizing sensor
control ? quantity, quality of data
16
Tracking Quality


RMSE
intervals
?
(truet-obst)2
v
t1
intervals


idx 1
idx 25

idx 55
Per-interval performance gains/losses may
accumulate across multiple update intervals
17
Outline

• Why separate sensor control and data?
• Closed-Loop Sensor Networks
• Meteorological Application
• Network, Sensing, Tracking Models
• Simulation Results
• Summary
• Future Work

18
Summary and Future Work
When network congestion, prioritizing sensor
control in closed-loop sensor network ? quantity,
quality of data, and gives better
application-level performance

• Results parallel Fredj et al, Sigcomm 2001 for
  diffserv
• Future work
• how do errors accumulate across control update
  intervals?
• other applications where gains can accumulate?
• challenge, importance of quantifying impact of
  system design decisions on application-level
  performance

that performance is generally satisfactory in a
classical best effort network as long as link
load is not too close to 100, and that there
appears little scope for service differentiation
beyond the two broad categories of good enough
and too bad.
19
Thank You!
Questions?
Contact Victoria Manfredi
vmanfred_at_cs.umass.edu More info
www-net.cs.umass.edu/vmanfred
20
Data Quantity vs Quality
? sensing accuracy 1/sqrt(N)
prioritize sensor control
1/2 control loop delay
? data samples (N)
360? scans, ? 5sec, very bursty traffic
CDF
FIFO achieves at least 80 as many samples as
priority 80 of time
Priority has at least 90 as much uncertainty as
FIFO 90 of the time

NFIFO / Npriority

?r,Priority / ?r,FIFO
During times of congestion, prioritizing sensor
control ? quantity, quality of data
21
More Data

• Control loop delay
• Prioritizing sensor control ? ? to zero, ?
  virtually unchanged
• FIFO ? - ? - ?
• Priority ? - ?

Priority control delay
?
?
Data delay
FIFO control delay
Data from control k
Data from control k-1
k
k1
? Update interval
gain in time collecting data is at most
? / (? - ? - ?)
More data, but gain depends on size of update
interval
22
Kalman filter

• xk estimated (location, velocity)
• yk measured (location, velocity)
• noisy, with std deviation sr(q,ab)

Measure radar data received, measured position
yk, with sr(q,ab)
Filter estimate xk based on yk, predicted x-k
Estimated state error covariance matrix,
depends on velocity noise model, sr(q,ab)
Predict next x-(k1) 99 confidence region,
gives qk1 to scan next time step
23
Simulation Set-up

• Network parameters
• Kalman filter parameters
• initialize based on storm data
• 10 NS-2 simulation runs, 100,000 sec each

Vary burstiness of other traffic,
r1 1s
?control 1/? pkts/s
on
off
?1 p?o
?2 (1-p)?o
?data 2000/30 pkts/s
?
r2 1s
?other 2000/30 pkts/s
Index of dispersion
?control ?data?other ? 133.37 pkts/s
? 148.5 pkts/s
avg load ? 0.90
24
Data Quantity

Number of times more voxels scanned under
priority than under FIFO
idx 55
idx 25
idx 1
? (seconds)
As ? ? and burstiness ?, gains from prioritizing
increase
25
Data Quality
Assuming ?? 360?


Reflectivity Standard Deviation
Number of Pulses
F(x)
? 30sec
? 30sec
idx?1
F(x)
idx?1
idx?55
? 5sec
idx?55
? 5sec
idx?55
idx?55
idx?1
idx?1
x ?r,Priority / ?r,FIFO
x NFIFO / NPriority
Small decision epoch, bursty traffic FIFO
achieves 80 as many pulses as priority 80 of
time
Small decision epoch, bursty traffic priority
has at least 90 as much uncertainty as FIFO 90
of the time
26
Number of Pulses
FIFO and Priority each achieve about 6x as many
pulses per voxel for ? 30 sec vs ? 5 sec,
and total of pulses is independent of ?
27
Effect of Packet Loss
FIFO sensor control packets dropped
Capacity when ?gt1000, data dropped
?r (with loss) / ?r (no loss)
Priority no sensor control packets dropped
? pkts / second
As system goes into overload sensing accuracy
degrades (more) gracefully when sensor control is
prioritized
28
Related Work
Prioritize Network Control

• SS7 telephone signaling system
• ATM networks, IP networks
• 1998 Kalampoukas, Varma, Ramakrishan, 2002
  Balakrishnan et al,
• priority handling of TCP acks
• 2005 Kyasanur, Padhye, Bahl
• separate control channel for controlling access
  to shared medium in wireless

Service Differentiation for Different Classes of
Traffic

• 2001 Bhatnager, Deb, Nath
• assign priorities to packets, forwarding
  higher-priority packets more frequently over more
  paths to achieve higher delivery prob
• 2005 Karenos, Kalogeraki, Krishnamurthy
• allocate rates to flows based on class of traffic
  and estimated network load
• 2006 Tan, Yue, Lau
• bandwidth reservation for high-priority flows in
  wireless sensor networks
• 2008 Kumar, Crepadir, Rowaihy, Cao, Harris,
  Zorzi, La Porta
• differential service for high priority data
  traffic versus low-priority data traffic in
  congested areas of sensor network

Our focus prioritize sensor control
Networked Control Systems

• data, sensor control sent over network
• constrained to be feedback and measurements of
  classical control system
• ratio of data to control much smaller than that
  of closed-loop sensor network
• 2001 Walsh, Ye
• put error from network delays in control eqns
• 2003 Lemmon, Ling, Sun
• drop selected data during overload by analyzing
  effect on control equations

Do not consider effects of prioritizing only
sensor control in a sensor network
Sub-class of closed-loop sensor networks
considered here