Traffic speed estimation using temporal and spatial smoothing of GPS speed data

A method of filling in speed values missing from GPS data for a determination real-time traffic speed over a plurality of links in a transportation network, comprising:

• determining an initial link speed estimate from collected GPS data mapped to one or more links comprising a transportation network;
• calculating at least one set of neighboring links using a network distance limit to identify the at least one set of neighboring links at least within a specified degree of separation and a road distance limit to identify the at least one set of neighboring links within a fixed distance along a link;
• generating a model of re-scaled speed estimation from the collected GPS data mapped to the one or more links, by:
• estimating a free-flow speed using the initial link speed estimate, and re-scaling the initial link speed estimate using the free-flow speed;
• building a rescaled speed profile comprised of a link profile represented by an hourly median value over a specific period of time for a particular link, a global profile represented by a median value across all links in the one or more links;
• extracting a representative profile by performing a cluster analysis on re-scaled speed profiles, by calculating a cluster median profile and building a median profile for each road class for profile-eligible links;
• snapping known GPS data to the one or more links using the re-scaled speed value;
• and
• smoothing the collected GPS data by mapping the model of the re-scaled speed estimation to the one or more links comprising the transportation infrastructure network by applying a plurality of different values of re-scaled speed estimation for each observation of a link l, and comparing those different values with collected GPS data relative to at least one link l and at least one time period u.

A link speed estimation system, comprising:

• a plurality of input including collected GPS data mapped to one or more links comprising a transportation network;
• a plurality of data processing modules, executed by at least one processor within a computing environment, and configured to supply missing speed values among the collected GPS data by utilizing observed information from one link to estimate neighboring links without observed information, the plurality of data processing modules including a preparation module configured to model a re-scaled speed estimation from the collected GPS data mapped to the one or more links by locating a set of closest neighboring links to the one or more links to which collected GPS data is mapped, build a profile of rescaled speed for collected GPS data over a specified initial period at a link and a time period, and perform a cluster analysis to extrapolate observed speed data from the collected GPS data to the set of closest neighboring links to develop a rescaled speed value, and a smoothing module configured to process the collected GPS data by mapping the model of the re-scaled speed estimation to the one or more links comprising the transportation infrastructure network, apply a plurality of different values of re-scaled speed estimation for each observation of a link l, and compare those different values with collected GPS values relative to at least one link l and at least one time period u;
• and
• an estimation module configured to determine a link speed estimate for all links 1 in the transportation network at all time periods u,
• wherein output data generated by the estimation module enables dynamic, real-time routing information for traffic across the one or more links of the transportation network.