EnviroTrack and JAM

1
EnviroTrack and JAM

• Presented by Chien-Liang Fok

2
EnviroTrack

• A entity-tracking middleware for sensor networks
• Provides an single context label abstraction that
  is dynamically created and logically associated
  with an entity
• Abstracts details of inter-node communication
• Performs of group maintenance
• Programmer interacts with context label rather
  than changing set of nodes

3
Context Label

• A form of attribute-based naming
• Serves as an address to the entity
• Is distinguished by type (e.g., fire or car)
• Contains aggregate state about the entity
• Hosts application-defined tracking objects
• Active elements that perform context-specific
  computation or action
• Abstracts a group of nodes that can sense the
  entity

4
Problem Definition

• Maintain context labels such that
• Each entity has one context label
• Context label follows the entity
• Group management, leader maintenance
• Approximate aggregate state is maintained
• Freshness constraint, Le
• Critical mass constraint, Ne
• Simplify application development
• Language support for defining context labels,
  aggregate state variables, and tracking objects
• Directory service

5
Programming Model
6
Context Label Specification

• To declare a context label of type e, the
  following must be provided
• sensee() specifies a pattern in the environment
  to watch for
• statee() the state data to be maintained within
  the context label
• EnviroTrack provides a library of statee()
  functions
• Accessible by all tracking objects
• Which tracking objects to associate with the label

7
Group Maintenance

• A group of sensors that detect the entity should
  produce a single context label for it
• The algorithm must be lightweight and dynamic
• All members of the group must satisfy sensee()
• There is at most one majority leader within the
  group

8
Group Maintenance Algorithm

• If a leader exists, the leader sends heartbeat
  containing e
• Propagates h hops beyond group boundary
• Tells member nodes leader is still present
• Notifies non-members of the existence of context
  label e

9
G. M. Algorithm Cont.

• When a member receives a heartbeat, it sets a
  timer. If timer expires before receiving another
  heartbeat, a leadership takeover protocol is
  triggered
• This is a backup to the standard leadership
  handoff

10
G. M. Algorithm Cont.

• When a non-member node receives a heartbeat, it
  sets a timer
• If the entity is sensed before expiration, the
  node joins the group
• If the entity is sensed after timer expires, the
  node creates a new group (and context label) and
  declares itself leader

11
Spurious Leaders

• The group member algorithm may result in more
  than one leader
• Solution
• Each context labels leader contains a weight,
  initially zero
• Increment weight each time data is received from
  the members
• Weight is inherited during leadership handoff
• Context label with greater weight wins
• Unlikely for spurious leaders attain Ne

12
Approximate Aggregate State

• The leader of the group maintains the approximate
  aggregate data state
• Leader sends each member Le
• The member periodically sends sensor data to
  leader with a period of Pe Le d
• d network delay processing time
• The leader performs aggregation every Pe
• A valid flag is set if Ne is not met

13
Directory Service

• Provides access to a particular type of context
  label
• Use a hash function
• context label type ? physical location (x,y)
• Nodes within one hop of (x,y) are called
  directory object
• Context labels periodically send their state and
  location to directory object

14
Language

• EnviroTrack provides a language for declaring
  context labels and aggregate data state
• Example code

15
Evaluation

• Theoretical example scenario
• A T-72 tank moves through a sensor field
• Sensors can detect tank 100m away
• Sensors arranged in grid 140m apart
• At max speed, tank moves one hop every 11.2
  seconds (45km/h)
• Actual test setup
• 10001 model of the above scenario
• 10s/hop (50km/h) and 15s/hop (33km/h)

16
Tracked tank trajectory

• Nodes located at integer coordinates
• Actual path y 0.5
• Nodes have no notion of proximity

17
Leadership Handover
18
Communication Performance

• System handles message loss
• Message loss not due to link usage
• Protocol uses very little bandwidth
• Scalable to greater tracking difficulty

19
Maximum Trackable Speed

• Affected most by heartbeat period
• Best results when
• Receive timer 2.1 HB
• Time to take over leadership
• Wait timer 4.1HB
• Time to create new group

20
Maximum Trackable Speed
21
Communication Radius Sensing Radius Ratio
22
Conclusions

• EnviroTrack is a middleware for tracking entities
• Relieves applications from details of
• Group management
• Directory service
• Data aggregation

23
JAM

• A protocol that allows applications to view a
  jammed region as an entity rather than a
  collection of broken links and congested nodes
• Allows data to be rerouted around a jammed region
  and other evasive actions

24
General Algorithm

• Nodes within the jammed area notify nodes outside
  of the jammed area that they are jammed
• Border nodes just outside jammed area coordinate
  to form groups of jammed nodes

jammed area
25
Architecture
26
Jam Detection

• A node considers itself jammed when the utility
  of its communication link falls below a certain
  threshold
• Based on local data (failed channel access,
  protocol violations, low SNR)
• When a node considers itself jammed, it
  periodically sends a JAMMED message
• Modify MAC to bypass carrier sense phase
• Hopefully an un-jammed neighbor receives it

27
Mapping protocol

• Border nodes receive a JAMMED messages from its
  jammed neighbors and calculates normal direction
  vector to the jammed nodes
• These nodes are called mapping nodes
• If it is not aware of a compatible group, it
  creates a new one with a random ID, waits a
  while, and broadcasts a BUILD message
• Two groups are compatible if the angular
  difference between their direction vectors is
  below some threshold

28
Group Coalescing

• If two groups are compatible, they are coalesced
  after a certain time
• The group with the larger group ID dominates
• Jammed nodes in the subordinate group is merged
  with the dominant group
• Dominant groups BUILD message is augmented with
  subordinate group ID

29
BUILD Message

• Contains
• ID of original sender
• Group ID
• Sequence number
• List of known jammed nodes
• List of subordinate group IDs
• When received
• If a regular node, store data locally
• If a mapping node, update the list of known
  jammed nodes and rebroadcast it

30
Eager Eavesdropping

• JAM attempts to quickly diffuse knowledge of
  jammed region
• Forward information diffusion As a BUILD message
  is relayed around mapping nodes, each node
  updates list of jammed nodes
• Back Flooding When a a BUILD message is
  rebroadcasts, the single hop upstream mapping
  node processes it
• Updates its list of jammed nodes
• Does not relay it

31
Edge Nodes

• A mapping node that does not have a neighbor left
  or right of the direction vector
• If a neighbor is not discovered within some time,
  it periodically sends a PROBE message to detect
  neighboring mapping node(s) or form bridge nodes

jammed area
edge node
32
Bridge Nodes

• Does not actually receive any JAMMED messages
• Connects two edge nodes
• Masks areas of low connectivity
• A node becomes a bridge node when it detects a
  PROBE message from a group it knows about
• A bridge nodes coalescing timer is longer to
  prevent spurious bridging nodes

33
Recovery

• When a jammed node becomes un-jammed, it
  periodically sends an UNJAMMED message
• When a mapping node receives it, it updates the
  list of jammed nodes and propagates an TEARDOWN
  message
• The opposite of a BUILD message
• When all of a mapping nodes jammed neighbors are
  un-jammed, the mapping node becomes a regular node

34
Example
1
2
3
4
35
Example (Cont.)
5
6
7
8
36
Simulation Methodology

• GloMoSim simulator
• 4000 x 4000 meter field
• 400 nodes placed at 200m interval grid
• Radio settings inbetween MICA2 and WaveLAN
• MACA IEEE802.11 MAC layer
• Jammed area ranged from 5-144 nodes
• Connectivity
• Low 4 neighbors
• Moderate 8 and 12 neighbors

37
Simulation Results

• Varying jammer range

38
Simulation Results Cont.

• Time to map region, 12 neighbor case
• 90 of mapping nodes know 1/3 of jammed region in
  0.5-3 seconds

39
Simulation Results Cont.

• Node failure rate

40
Conclusions

• JAM provides a protocol for mapping a jammed
  region
• Loose group semantics and eager eavesdropping
• Quick convergence and tolerance to failures when
  network is moderately connected