INRIX’s proprietary Traffic Fusion Engine uses sophisticated Bayesian modeling and proprietary error correction technology to process the hundreds of data sources aggregated by the INRIX Smart Dust Network and generate accurate real-time and predictive traffic data.

The subject of multiple patents pending, the INRIX Traffic Fusion Engine is designed specifically to maximize data quality, employing:

• Detection of Malfunctioning Traffic Sensors – particularly in the case of public DOT loop detector and toll-tag reader data, feeds and the physical sensors and communications infrastructure are prone to failures and long-lead repair times. Uniquely INRIX detects incorrectly functioning sensors in near real-time, flagging them to be ignored as a data source for the Fusion Engine.
• Geospatial Filtering – due to its unique relationships with data suppliers, INRIX obtains location and time information, speed and heading for each vehicle together with additional metadata that provides context for the vehicles’ current status, allowing high-accuracy geospatial filtering of probe-derived data.
• Collaborative Filtering and Outlier Detection – INRIX combines data from all sources allowing collaboration between data points that agree with one another to identify statistical outlier data points and hence compute a high confidence estimate of real-time conditions together with an error estimate.
• Optimizing Spatial Granularity – due to the high density of raw data that INRIX receives it is frequently able to determine real-time conditions at or below the resolution of a single TMC segment. In real-time, the Traffic Fusion Engine is able to adjust the spatial granularity of the data it computes to maximize the statistical confidence in the data.
• Never Low Confidence Data – If, due to insufficient data at a given location, INRIX is unable to meet its error threshold for real-time data, no information is reported on the road segment in question.

To make useful predictions, the Traffic Fusion Engine first determines the combination of factors that will influence future traffic patterns and develops an in-depth understanding of what traffic is like in each metropolitan area. These include obvious factors such as the day of the week, the weather, accidents and road construction, as well as other events such as school schedules, sports games and concerts, and even uniquely local variables such as the legislative calendar in Washington, D.C. It also applies proprietary error-detection algorithms to road sensor data, so a single unreliable sensor does not lead to an incorrect reading for a particular stretch of road.

Since these kinds of predictions cannot be made using simple probabilities, the Traffic Fusion Engine makes use of Bayesian analysis, which calculates the ongoing probability of something happening in the face of many different factors. INRIX leverages sophisticated Bayesian analysis to create a model for traffic in specific metropolitan areas that map the relationships between real-time traffic, historical traffic patterns, weather, time of day, specific events and many other factors. These graphs help the Traffic Fusion Engine generate useful predictions in the face of countless and constantly-changing variables.

The Traffic Fusion Engine includes exclusive technologies from Microsoft Research, building on their innovations in statistical inference of traffic patterns, predictive analysis and visualization of real-time systems.