Power, Location,

1
Power, Location, Time in Wireless Sensor
Networks

• Selected Results from UCLAsDSN, PADS,
  SensorWare Research

Collaborators M. Potkonjak (UCLA), C. Chien
(Rockwell),J. Agre (JPL), B. Schott R. Parker
(ISI), M. Jones (Virginia Tech.)
2
Networked Embedded Devices
SmartSensorNode
T3
Event
Internet
SmartSensorNode
SmartSensorNode
Gateway
T2
T1
Event
SmartSensorNode

• Nodes coordinate local processing among
  neighbors to combine their results
• lower network traffic, higher-level sensory tasks
• Application large scale, dynamically changing,
  robust sensor colonies

3
Example Node Rockwells WINS

• Capabilities vibration, acoustic, accelerometer,
  magnetometer, temperature sensing
• Features modular, miniature form factor

4
Example Node UCLAs WAND

• Capabilities acoustic I/O, image I/O, GPS
• Features scalable codecs, adaptive link/MAC
  processor, embedded router

5
Design Trade-offs
Lifetime(power)
Rapidity(latency -1)
Quality(coverage, fidelity)
6
Outline

• Power-aware RTOS networking
• Network coverage algorithms
• Low-latency tracking
• Implementation and simulation

7
Power-aware Operation

• Intra-node
• hardware circuits
• software
• Inter-node
• network protocols

8
Power-aware Resource Management in RTOSs

• Latencies are critical
• unlike DSP where only rate (sample period)
  matters
• deadlines important in protocols, target tracking
• Tasks are dynamic
• cannot schedule the tasks statically
• Hardware provides many control knobs for
  power-performance trade-off
• CPUs with dynamic frequency/voltage
• radios with multiple modes and symbol rate choices

9
Fixed Priority Preemptive CPU Scheduling in RTOSs

• Consider task set (period, WCET, deadline)
• (10, 3, 10), (14, 7, 14)
• CPU utilization 3/10 7/14 80
• Obvious power management strategies
• Shutdown when idle
• saves 20 power
• Can we slow CPU by 20 ( reduce V) for more
  savings?
• NO, as deadlines will no longer be met
• However, can slow by x 14/13 and lower voltage to
  still meet deadlines, and shutdown during idle
  time
• saves 22.5 in power
• Problem uses WCET (worst case execution time)

10
Predictive Strategy for Exploiting Execution-time
Variation

• Significant variation in execution time of tasks
• WCETBCET often gtgt 1
• e.g. compressed speech playout task has different
  time for talkspurt vs. silence
• e.g. on test run, MPEG decoder time range
  0.003s, 0.15s with average 0.035s
• Obvious shut down or reduce voltage if task
  finishes earlier
• Even better predict execution time of task
  instance and dynamically scale voltage even more
  aggressively
• task-specific predictor
• but, some deadlines may be missed!
• leads to packet loss another form of noise!
• provides power-quality trade-off

11
Sample Simulation Result
12
Sample Simulation Result
13
Power-aware Networking

• Known results
• power-aware adaptive link protocols
• optimize computation and communication energy
  spent per good application level bit distributed
  (Joules/bit)
• multihop communication provides power capacity
  benefits
• traditional multihop routing is power unaware
• focus on topology changes, and metrics such as
  shortest hop, shortest delay, link quality etc.
• power wasted in quiescent state signaling
• using power-based routing metrics, and reactive
  protocols helps
• What more can be done?
• coordinated routing and MAC
• large number of collisions with CSMA MAC during
  broadcasts used by routing protocols for probing
• exploit path diversity, and node traffic
  redundancy

14
Power-aware NetworkingPath Diversity Node
Redundancy

• Idea combine data from different nodes, and
  distribute traffic over alternative paths
• increase network lifetime and coverage
• packet disperser and combiner entities
• works with variety of routing approaches
• Evaluation metrics
• time to breakdown
• of depleted nodes
• RMS energy distribution
• A problem do not know which nodes are important
  as it depends on future target traffic pattern
  and user movement
• traditional load balancing is based on only
  present activity
• goal is stochastic lifetime, but practical
  approaches need indirect measures

15
Data Combining

• Centralized approach Data combining entity for a
  cluster
• Distributed approach Data combining entities
  created at those nodes that receive packets from
  different sources for the same user
• soft state new DCE when node with DCE dies

16
Spreading Techniques

• Stochastic scheme
• e.g. stochastic GBR where traffic spread among
  paths with similar gradients
• Energy-based scheme
• e.g. node increases its height when its energy
  falls below a threshold to repel traffic
• Stream-based scheme
• take action even before a node begins to die
• divert new streams away from nodes that are
  handling other streams

17
Evaluating Energy Efficiency
Without load spreading
With load spreading

• An approximate way is a histogram
• area of histogram vs. shape of histogram
• but only approximate (cant average over all
  futures)
• Possible metric to capture the essential
  histogram info
• RMS of the histogram (measures total as well as
  spread)

18
A Path Diversity Scenario
B
C
A
F
D
E
user

• A and B generate 1 packet every 100 ms until 5s
• C generates 1 packet every 100 ms from 5s till 15s

19
of Nodes with gt 10 Battery
Packets received by t150 Normal
127 Stochastic 133 Energy Disperse
160 Stochastic ED 161 Divert streams 175
20
RMS Battery Energy Consumption
Lower Bound 2
Lower Bound 1
21
Load vs. Energy Oriented Path Selection
Nodes have energy to forward 20 packets
B
D
C
E
F
A
Time 1 to 10 A and E send data to B _at_ 1 packet /
time
22
Load vs. Energy Oriented Path Selection
Nodes have energy to forward 20 packets
B
Dies at time 20
D
C
E
F
A
Time 10 to 30 F sends data to B _at_ 1 packet /
time
23
However

• There are specific future scenarios where
  spreading can actually reduce network lifetime
• Need to discover bottleneck nodes and avoid
  them during spreading
• hard problem
• criticality of a node depends on how many
  redundant nodes are backing it up
• possible solution centralized algorithm at the
  user where network topology can be built up
• inform critical nodes of their level of
  criticality so that they repel spreaded traffic
• scalability issues

24
Coverage and Location Problems

• Queries not just about the target, but also about
  the network coverage
• Example
• where are the nodes located?
• what is the coverage of the sensor network?
• How do we
• where are the paths for useful metrics such as
• maximal breach
• minimal exposure
• maximal support
• where should additional nodes be located to
  improve coverage metrics?

25
Example Maximal Breach Path

• Problem find the path of maximal breach of
  surveillance between a starting region I and
  ending region F through a sensor network

Approximate coverage boundary
Sensor node site
Breach origin
I
Breach destination
F
26
Solution Step 1 Generate Voronoi Diagram
By construction, each line-segment maximizes
distance from the nearest point
(sensor). Consequence Path of Maximal Breach of
Surveillance in the sensor field lies on the
Voronoi diagram lines.
27
Solution Step 2 Weighted Graph

• Given Voronoi diagram D with vertex set V and
  line segment set L and sensors S
• Construct graph G(N,E)
• Each vertex vi?V corresponds to a node ni ?N
• Each line segment li ?L corresponds to an edge ei
  ?E
• Each edge ei?E, Weight(ei) Distance of li from
  closest sensor sk ?S
• Formulation Is there a path from I to F which
  uses no edge of weight less than K?

28
Solution Step 3 Search for Maximal Breach Path

• Breadth first search to establish the existence
  of path from I to F
• A combination of repeated binary search and BFS
  to find the path of the maximal Weights(ei),
  which will have the longest absolute distance
  from the sensor nodes
• This will be the maximal breach path optimized
  for absolute distances

29
Another Example Minimal Exposure Path
S1 (xs1,ys1)
Task Find the average distance of the sensor
from the path of the agent Find new weights for
the edges of the Voronoi diagram described
earlier and solve for the maximal breach path.
(xi,yi)
S3
n1
n2
dL
(x,y)
L
S1
(xj,yj)
S2
30
Maximal Coverage Path

• Given Delaunay Triangulation
• of the sensor nodes
• Construct graph G(N,E)
• The graph is dual to the Voronoi graph previously
  described
• Formulation what is the path from which the
  agent can best be observed while moving from I to
  F? (The path is embedded in the Delaunay graph of
  the sensors)
• Solution Similar to the max breach path
  algorithm, use BFS and binary search to find the
  shortest path on this Delaunay graph.

31
System Scenario
GATEWAY
COVERAGE SERVER
Network Management

• Coverage service
• Location discovery
• RSSI-based joint location and channel parameter
  estimation using iterative multilateration

32
Location Discovery in Sensor Networks

• Given a network of sensor nodes where a few nodes
  know their location how do we calculate the
  location of the nodes?
• GPS may not work at all nodes in all terrains
• Special long range beacons have problems
• low fault tolerance, low security, problems with
  obstacles, and high energy consumption
• New approach RSSI-based iterative multilateration

33
Distance Estimation and RSSI
Pr KPt/rn

• Problem K and n also unknown

34
Iterative Multilateration

• P of the sensors are beacons
• Sensor can only communicate in a limited range
  of distance (d)
• -The nodes which located themselves after the
  first pass (Direct trilateration DTL)
• -The nodes which located themselves after the
  second pass
• -The nodes which located themselves after the
  third pass
• -Nodes which never find their location
• - Nodes with defined location after
  indirect trilateration (ITL)

35
RSSI-based Iterative Multilateration

• Some nodes know their positions initially, and
  act as local beacons
• RSSI measurements used to obtain distance
  constraints between nodes
• carried as part of routing messages
• Statistical estimation of location of nodes with
  unknown positions, and radio propagation model
  parameters
• done iteratively
• alternative global estimation

36
Single-step vs. Iterative Multilateration
N1000 d0.01
N3000 d0.01
Ndensity of random nodes drange
37
Latency in Tracking Tasks

• Sensor nodes viewed by some as tiny database
  servers
• a user query is answered by aggregating results
• forms of queries
• Do you currently see X?
• Did you see X in the past timer T?
• Inform me when you see X.
• Tell me periodically what you see.
• Problem remote user in the loop
• leads to higher latency for tracking
• also, excessive traffic and wasted power

38
Low-latency Proactive Tracking

• Tracking request is a tiny control script sent
  to sensor node detecting the target
• Script spreads to select nodes in target
  vicinity, and organizes them into an active set
  for the target
• results filtered and combined locally
• Active set for a target is autonomously
  updated to minimize latency and traffic to user
• nodes woken up and added in advance using
  history-based target motion prediction

39
Tracking Scenario
Region A
40
Tracking Scenario
Region A
41
Our System Architecture
Transient External User (UAV)
Sensor Scripts
SCRIPT
Applications
APP
Power-ware RTOS
Sensor Middleware
Hardware Abstraction Layer
Sensor Node Hardware
Sensor 1
Sensor 2
42
SensorSim Hybrid Simulator

• Motivation study sensor network deployment,
  protocols, and applications at scale in a
  controlled setting
• Three key capabilities
• Hybrid simulation
• selected nodes in a simulation can be real
  nodes
• currently supports only higher layers in real
  nodes
• real applications can run on nodes in a
  simulation
• Power modeling
• Energy consumer models radio, CPU, sensors
• Energy source models batteries
• Sensor and target modeling
• Target, sensor channel, and sensor transducer
  characteristics
• Current implementation based on ns simulator

43
SensorSim Architecture
monitor and control hybrid network (local or
remote)
real sensor apps on virtual sensor nodes
app
GUI
app
socket comm
serial comm
ns
HS Interface
GUI Interface
RS232
Ethernet
gateway Dll (RSC)
V
V
R
V
Gateway Machine
V
R
modified event scheduler
Proxies for real sensor nodes
Simulation Machine
44
Conclusion
Lifetime(power)
Rapidity(latency -1)
Quality(coverage, fidelity)

• Task-specific multidimensional quality of service
• Hunt for the best protocol for sensor nets is
  futile
• e.g. different tasks work best with different
  routing
• boundary between networking other layers fuzzy
  in sensor net
• A distributed computing viewpoint is better
  suited
• application-specific protocols for different
  tasks
• approach a minimal scriptable substrate with
  power, location, and time aware services