Monitoring system for in-vehicle riding comfort by using real-time GPS data and statistical process control

Abstract

Riding discomfort is one of the indicators which affects the service quality performance in transportation service industry. For the firm or agency that operates the buses, the driving style of the bus driver based on the acceleration and braking is the parameter which contributes to the riding quality. Therefore, this behavior should be monitored and controlled. This research studied the driving behavior in detail. Five onboard surveys on the Bus Line 515 from Salaya sub-district to Victory Monument, Bangkok were conducted using the GPS (Global Positioning System). There were three primary data: GPS data, trip record, and riding comfort assessment. In order to obtain the best results, the raw GPS data were filtered, cleaned, and smoothed. Then the computation of acceleration and jerk was estimated. Afterward, data were gathered and analyzed using the descriptive statistics. Sampling rate approach was used to derive the best interval, since it could represent the actual driving behavior event. The results indicated that the interval of 1-s tends to well represent the actual movement and acceleration behavior, where the interval of 2-s seems appropriate to represent jerk. Given the acceptable threshold of [-1.8, +1.8], only 3% on average of the drivers violated this range. Therefore, the drivers showed a good behavior during the trips. Yet, to determine and identify when the driving behaviors became poor, a control chart of Shewhart was used to analyze by determining the upper and lower control limits. By comparing the statistic results of GPS data, figures from literature, and the on-board survey report, 1.8 and -1.8 m/s2 represented the actual feeling and driving behavior events. Thus, by applying the 8 rules, unusual driving patterns can be detected. The rules were also coded using Python to develop the monitoring system. An auditory system and a displayed message were the features of this system. The findings showed that the proposed system could give the actual information of bad driving behavior approximately 23.3%, where 76.7% was the false alarm. Shewhart control chart proved as a very effective tool to detect the extreme outliers, the aggressive, and the uncomfortable driving style. In reverse, it was sensitive related to the dwell time, speeding, and accelerating.