Diagnosing Road Weather Conditions with Vehicle Probe Data

Abstract

Over the past 2 years, the U.S. Department of Transportation RITA funded an IntelliDrive vehicle probe data collection test bed in the northwest Detroit, Michigan, area. The purpose of the test bed was to provide the infrastructure for both public and private organizations to collect, process, and generate a robust observation data set for multiple purposes (e.g., crash avoidance, automated toll services, weather diagnostics). During April 2009, a weather-specific field study was performed over an 11-day period. The resulting data set was processed by a vehicle data translator (VDT), which parsed, quality controlled, and combined these data (with ancillary weather data) in the generation of road weather-specific algorithms. This paper briefly describes the VDT concept and then examines the accuracy of the quality-controlled temperature and pressure data (for several different stratifications) collected from 11 specially equipped vehicles operated during the study time period. Results show that the vehicles accurately measure the temperature (compared with a nearby fixed weather station, KDTW), but are not as accurate at measuring the barometric pressure. In addition, stratification by speed, vehicle type, time of day, and occurrence of precipitation do not affect the accuracy of the temperature and barometric pressure measurements.

A Geostatistical Investigation into the Effective Spatiotemporal Coverage of Road Weather Information Systems in Alberta, Canada

“…Next, data that fell outside of the reasonable ranges was recognized as erroneous. The reasonable ranges were determined based on sensor specifications, locationspecific climatological ranges, and historical data ranges [16,17]. Then, in the model analysis test, data that fell beyond acceptable regions was reported as invalid.…”

“…In addition to the basic atmospheric observations, the data from onboard vehicle sensors, such as wiper state and antilock braking systems, can also be used in road weather condition estimation. With CV-enabled data in hand, Drobot et al [16,17] designed a quality and accuracy check process, which includes crosschecks with sensor specifications, climatological ranges, neighboring vehicle, and station measurements, for weather-related data from probe vehicles. Other data analysis shows that the quality of temperature and pressure data is affected by many factors, such as vehicle type, speed, and precipitation occurrence [16], and the temperature data from probe vehicles closely resembles data from stationary weather stations [17].…”

Realizing the Potential of Vehicle-Based Observations

“…Potential Solutions: A solid set of sensor studies in the mobile environment may provide vehicle manufacturers and others with standards regarding elements such as optimal sensor type and placement. For example, studies indicate that placement of a temperature sensor near the front grill is optimal (2), producing temperatures accurate to less than 1.00°C, with a bias less than 0.25°C (9,10). To allow manufacturers the flexibility to select the optimal location for a given model, it may be necessary to articulate standards in terms of principles and guidelines rather than specific locations and other requirements.…”

“…Based on recent research (2,5,9,10) and expert opinion of the committee members, a list of desired vehicle data elements have been identified and are listed in Table 3. These data elements…”