Vehicle Infrastructure Integration Application: Weather Data Translator

1
Vehicle Infrastructure IntegrationApplication
Weather Data Translator
Clarus ICC5 Meeting Session 6

• Andy Stern
• Consulting Meteorologist (Noblis)
• FHWA Road Weather Management Team

2
Dealing with a fire hose of data

• Potentially there could be
• millions of vehicles providing data
• 10s to 100s of millions of probe messages
• available 24/7/365
• on both interstates and arterials
• How do we deal with all of the data?
• Weather Data Translator (WDT)
• Data segmentation

3
Potential Vehicle-based Elements

• Hours of operation
• Elevation
• Accelerometer data
• Vehicle speed
• Heading
• Steering wheel rate of change
• Exterior temperature
• Windshield wiper rate
• Rain sensor
• Sun sensor
• Adaptive cruise control radar

• Impact sensor
• Barometric pressure
• Fog lights
• Headlights
• Relative humidity
• Anti-lock braking system
• Traction control
• Stability control
• Pavement temperature
• Brake boost
• Wiper status

4
Elements Available in PoC

• Hours of operation
• Elevation
• Accelerometer data
• Vehicle speed
• Heading
• Steering wheel rate of change
• Exterior temperature
• Windshield wiper rate
• Rain sensor
• Sun sensor
• Adaptive cruise control radar

• Impact sensor
• Barometric pressure
• Fog lights
• Headlights
• Relative humidity
• Anti-lock braking system
• Traction control
• Stability control
• Pavement temperature
• Brake boost
• Wiper status

5
Data Analysis

• Hoping to collect a diverse dataset (e.g. day,
  night, rain, snow, etc)
• Statistical Analysis
• Compare vehicle data to fixed mobile data
• Determine if there are biases, quality errors,
  outliers, etc
• Estimate minimum number of samples required to
  result in quality data

6
Data Integration

• Integrate VII-based data with other weather data
  sets
• Determine correlation of weather radar
  reflectivity data and wiper status/rate
• Determine correlation of ABS/VTC/VSC activation
  with slippery road surfaces
• Correlate air temperature with ABS/VTC/VSC events

7
Weather Data Translator (conceptual)
8
Data Segmentation

• Rather than forwarding every probe data message,
  consider statistical processing based on road
  segmentation
• For the purpose of this discussion, roads can be
  segmented in the following ways
• static segmentation
• dynamic segmentation

9
Static Segmentation

• Equally divide the interstate network into 1 mile
  segments note metadata features
• Using a consistent frequency across the nation
  (e.g., every 10 min), bin all collected
  observations into 1 mile segments along the
  interstate
• Use the observations to obtain averages, remove
  outliers test as a comparison with nearby in
  situ observations (e.g., ASOS, ESS, etc)

10
Static Segmentation Example
Probe data are binned according to mile
posts regardless of geographic or infrastructure
features
5
4 vehicles take snap shots in this segment
ESS
1
2
RSU
4
4
3
1
5
mp 15
mp 16
4
3
1 mile
mp 12
mp 14
mp 13
Valley
Hill
Hill
11
Dynamic Segmentation

• Divide the interstate network into segment
  lengths that take into account differences in
  geography or infrastructure (e.g., bridges)
• Using a consistent frequency across the nation
  (e.g., every 10 min), bin all collected
  observations into these different size bins
• Use the observations to obtain averages, remove
  outliers test against in situ observations
  utilizing metadata from each segment

12
Dynamic Segmentation Example
Process probe data binned according to geographic
or infrastructure features
2 vehicles provide snap shots in this
segment which contains a bridge
5
ESS
3
2
RSU
4
2
3
1
3
seg 15
seg 16
4
5
seg 12
1 mile
seg 14
seg 13
Valley
Hill
Hill
13
Derived Observations

• In addition to metadata, each segment could have
• Air temperature
• Precipitation occurrence (yes, no)
• Precipitation intensity (none, light, moderate,
  heavy)
• Precipitation type (liquid, frozen)
• Barometric pressure
• Pavement condition (not slippery, slippery)

14
Our Vision

• Increasing surface observation data density from
  thousands to millions

Credit Kevin Petty, NCAR