Get the Insights You Need.
Our data processing algorithms enable a homogenous data stream, which is essential to providing unbiased, fair, and consistent behavior intelligence. The underlying challenge is due to monitoring devices (either phones or dedicated hardware) sourcing from a wide range of different sensor platforms. Each platform may sense the same movements differently, and thus provide different results in the ultimate behavior analysis. These sensing differences are due to calibration deficits, sensor drift, sensor platform quality, or the overall embedded software package. Our advanced algorithms combine signal processing techniques (e.g. calibration, orientation, noise reduction, signal separation, etc.) and artificial intelligence (e.g. self-learning and decision making) to overcome these challenges.
Our patented telematics data processing techniques turn non-homogenous data streams into homogenous data streams. As a result, data streams from different types of devices (e.g., smartphones or dedicated telematics hardware) utilizing arbitrary sensor platforms produce the same data - and consequently the same behavioral analysis - when exposed to the same movements.
Our mobility behavior analytics platform incorporates the detection of specific mobility events that provide indications of aggressive, distracted, and/or inefficient behavior. These events include hard braking, sharp turning, phone handling, etc. Each event is annotated with time, duration, and location information that indicate when and where the event happened. The events are extracted based on thresholds that are deemed acceptable for different types of maneuvers as well as considering the environment the maneuver occurred in (e.g., on-ramp vs. off-ramp, up-hill vs. down-hill). Hence, these thresholds are set based on different driving conditions, environments, and population characteristics (e.g. location, vehicle type, weather trends, road type, etc.). For example, the braking or turning threshold for a large commercial truck is different than that for a sports car or motorcycle. Similarly, these thresholds are different for driving at high speeds on the highway than they are for driving through city streets. Further, drivers naturally adjust driving behavior based on location (e.g., driving in a large city, such as New York or Los Angeles, is vastly different than driving in a rural or suburban location). This information is all taken into account in our driving event detection.
Our behavior-based mobility intelligence platform enables fair and consistent mobility scoring results. Mobility behavior is influenced by many factors. Some of these are internal factors that reflect the active decisions the driver makes. Additionally, there are external factors (e.g. local weather, traffic, infrastructure, culture, etc.) that are taken into account to quantify a behavior. The biases these external factors can introduce are considered and removed from the analysis due to the artificial intelligence layer. However, the driver’s reaction to external factors is an important component of behavior-based analysis and is incorporated in our behavior intelligence.
Behavioral biometrics is crucial to providing accurate mobility scoring. For example, a trip with two hard braking events might indicate good, average, or poor driving performance depending on the other factors involved. To illustrate, consider two extremes: 1) a driver on a straight road with few vehicles in a region where keeping a large distance from the car ahead is typical, and 2) a driver on a busy city street with poor road layout surrounded by an aggressive driving culture. In the first case, two hard brakes is much harder to justify than in the second case. It is more likely in the second case that the driver had to brake suddenly to avoid the car that just cut in front or the pedestrian that ran across the street.
Our patented technology takes these factors into account to provide a fair and consistent evaluation of a driver’s tendencies. The statistical performance over the entire drive is monitored in addition to looking for specific driving events. For example, a driver that continually brakes hard at every stop sign and speeds off quickly (though does not cross the thresholds set for driving events) indicates more aggressive and less efficient driving behavior than one who steadily reduces speed but may have one or two braking events in a city. Additionally, our behavior-based analytics factor in regional differences, infrastructure (e.g. mountains v. flat land), vehicle type, weather trends, and additional information to compare driving performance across different regions, environments, etc. The driver is evaluated over time to derive a deep understanding of a driver’s behavior. However, individual trips may be evaluated in the same manner to provide insight into a driver’s current trends or to distinguish between driver and passenger.
One of the core facts we understood from the very beginning is that a smartphone and most motion tracking devices are not built to capture highly accurate data. While those devices are equipped with sensors, (e.g., location, accelerometer, magnetometer, gyroscope or other sensors), the capability of producing a high quality, homogenous data stream is very limited. Modern smartphones, for example, are primarily utilizing motion sensors to either rotate pictures as the device is rotated or to detect repetitive motion sequences such as walking or running.