3. 3
Intelligent TransportationIntelligent Transportation
M2m Cloud
Bus Tracking System
Location ( GPS), Speed, Accelerometer, Passenger
Ids
Valid passenger lists, Route Info
• Can we map the Road Condition?
• Can we predict Vehicle Condition?
• Can we detect Bad Driving Behavior?
Pilot at TCS, Siruseri Campus
Business Problems
4. 4
Vehicle Model Driven Sensor Data AnalysisVehicle Model Driven Sensor Data Analysis
CONSTANTS WE CAN MEASURE
Vehicle Type & Driving Behavior Road Condition Monitoring
Road Condition & Driving Behavior Car Prognosis
Road Condition & Vehicle Type Driving Behavior Analysis
Acceleration a(t) = f (H(t), v(t), R(t), D(t))
H(t)
System
Identification Tools
5. 5
MotivationMotivation
Need for a synthetic data generation
Experimentation is time consuming and can't provide exhaustive
analysis
Measurement influenced by multiple parameters (e.g. mobile
position, car type)
Our Work
Model based approach to synthetic accelerometer data
generation.
Preliminary results on using synthetically generated data to
understand road-vehicle interaction
6. 6
ModelingModeling
Aim: Model the interaction of road anomaly – vehicle interaction as
captured by a mobile phone in the form of acceleration profile.
Features considered
Five Road types (A to E as per ISO classification)
Four car types (saloons: large, medium, small; small mass)
Selectable: mass, natural frequency, tire radius.
Road anomaly: step pothole.
Impact area types: smooth, loose pebbles, muddy soil
7. 7
Modeling - InteractionModeling - Interaction
Modeling approach:
First principles modeling of tire
motion
Tire launches as horizontally
launched projectile into pothole
Tire bounces out of the pothole.
Only one tire goes through the
pothole and hence the other axle is
firmly in place.
8. 8
Modeling – Road gradeModeling – Road grade
Road grade modeling
5 road types – PSD based ISO classes
Data generation based on approaches in literature
Generated road profile passed through quarter car to get
acceleration data
Road
Generator
Road
Type
Car
Model
Car type
Pothole
Generator
+
Pothole
parameters
Synthetic
Data
9. 9
Modeling - ResultsModeling - Results
Experiment:
Tata Nano (small mass) at 30
km/hr
5 cm deep pothole on a bad
(Class D) road.
Device: Samsung Galaxy smart
phone accelerometer at 18
Hz.
Limitations
Pothole exit modeling
10. 10
FeaturesFeatures
Features: Useful for pothole detection.
Literature: Time (amplitude) and freq. (PSD) domain.
In this work, we use Jerk → Rate of change of Z-direction acceleration:
Feature 1 – maximum of abs value of jerk - max(abs(diff(Az))
Feature 2 - energy of the jerk in a time window.
11. 11
Feature 1 – Different Road typesFeature 1 – Different Road types
12. 12
Feature 1 – Different Car typesFeature 1 – Different Car types
13. 13
Feature 2 – Different Road typesFeature 2 – Different Road types
14. 14
Feature 2 – Different Car typesFeature 2 – Different Car types
15. 15
Feature 1 Validation: Max. of Jerk in Window (5cm pothole)Feature 1 Validation: Max. of Jerk in Window (5cm pothole)
17. 17
Concluding RemarksConcluding Remarks
CPS: integration of process and computation → enables quick
reporting of events
We illustrated one such work for road grade monitoring.
We have illustrated the use of synthetically generation
accelerometer data for analysis.
Proposed features are useful to indicate sharp irregularities in a road
profile.