Outline

1
(No Transcript)
2
Outline

• Background on Landslides
• Landslides Prediction
• System Architecture
• Solution
• Evaluation

3
Landslides

• A landslide is an catastrophic event where a
  block of earthen mass slides downhill.
• Cause significant loss of life and billions of
  dollars each year.

4
Landslides Prediction

• Although a basic understanding of the landslides
  is available, system that predict the occurrence
  of a landslide do NOT exist.
• Why?
• The lack of field measurements over large
  temporal and spatial scales.

5
Landslides Prediction

• The development of a landslide is a temporal
  process
• takes as long as a year to develop
• Movement speed, several cm per month
• Landslides are spatial in nature.
• Position
• Movement direction

This paper propose a WSN to detect the early
signals preceding a landslide.
6
System Architecture

• Sensor Column
• Geophone
• Strain Gauge
• Pore Pressure Transducer

7
System Architecture

• Deployment
• A network of sensor columns
• Placed in vertical holes drilled over the hill
  surface
• Using sensor columns to detect movements

8
Solution Outline

• Geophones estimate displacement dij(t)
• Based on distance matrix Ddij(t)
• Detection
• Determine whether slip surface has formed
• Classification
• Estimate subset of sensors that moved
• Localization
• Compute location of slip surface

9
Example

• Detection Classification Localization

10
Detection

• Use strain gauges on each sensor column
• Can measure changes in their length due to
  deformation

11
Classification

• Determine which sensors are above and below the
  slip surface
• Basic principles Distance between two nodes
• Below the slip surface should not change
• Across the slip surface is likely to change
• Above the slip surface would see a small change
• The nodes located closest to the known rigid part
  are unlikely to move. (anchor nodes)

12
Classification

• Voting Algorithm
• Note state 0, 1, U
• Initially anchors0, restU
• Repeat until all nodes classified
• Update node is state based on votes from
  neighbors j
• If j in U, no vote
• If ?dij0, vote is equal to js state
• If ?dij?0, vote is equal to complement of js
  state

13
Localization

• Localize moved nodes using trilateration
• Slip surface estimation

14
Evaluation

• Evaluation metrics
• Average and std. dev of localization error
• Max. distance between actual and estimated slip
  plane
• Abstract network model
• Communications are error-free
• Nodes do not fail

15
(No Transcript)
16
Summary

• Propose a WSN for the prediction of landslides
• Design a system to detect the early signals
  preceding a landslides.

WSN
Civil Engineering
17
Remarks

• Localization for a specific application
  (landslides prediction)
• NOT for general purpose.
• Localization for a group of sensors (sensors
  above slip surface)
• NOT for a single sensor node.
• Localization for moving sensor
• although there is only a little movement

18
Thanks!