AUTO-CALIBRATION FOR ROAD TRAFFIC PREDICTION

• US 20150269838A1
• Filed: 10/20/2014
• Published: 09/24/2015
• Est. Priority Date: 03/19/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Joseph, Jojo, Nair, Raghunath E.
• Granted Patent: US 9,412,267 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G08G 1/0112 from the vehicle, e.g. floating car data [FCD]
• CPC Class Codes: G08G 1/0116 from roadside infrastructure, e.g. beacons
• CPC Class Codes: G08G 1/0129 for creating historical data or processing based on historical data
• CPC Class Codes: G08G 1/0133 for classifying traffic situation

Abstract

A method for auto-calibrating parameters in traffic prediction. The method includes determining a first subnet of traffic links that is associated with a plurality of traffic links in a traffic network. The method includes determining a second subnet of traffic links that is associated with the first subnet of traffic links and has a first traffic predicting accuracy value. The method includes generating a set of optimized traffic predicting parameters associated with the second subnet of traffic links, and applying the set of optimized traffic parameters onto a third subnet of traffic links. The method includes determining the set of optimized traffic predicting parameters used to calculate prediction results having a second traffic predicting accuracy value, and applying said set of optimized traffic predicting parameters to subnets associated with the traffic network. Further, the first traffic predicting accuracy value is lower than the second traffic predicting accuracy value.

Claims

1. A computer-implemented method for auto-calibrating parameters in traffic prediction, the method comprising:

determining, by one or more computer processors, a first subnet of traffic links having an association with a plurality of traffic links in a traffic network;

determining, by one or more computer processors, a second subnet of traffic links that is associated with the first subnet of traffic links, wherein the second subnet of traffic links has a first traffic predicting accuracy value;

generating, by one or more computer processors, a set of optimized traffic predicting parameters associated with the second subnet of traffic links;

applying, by one or more computer processors, the set of optimized traffic predicting parameters onto a third subnet of traffic links;

determining, by one or more computer processors, the set of optimized traffic predicting parameters used to calculate prediction results having a second traffic predicting accuracy value;

applying, by one ore more computer processors, the set of optimized traffic predicting parameters used to calculated prediction results associated with the second traffic predicting accuracy to subnets associated with the traffic network; and

wherein the first traffic predicting accuracy value is lower than the second traffic predicting accuracy value.

2. The method of claim 1, wherein the first traffic predicting accuracy value and/or second traffic predicting accuracy value is proportional to the difference between:

a predicted value obtained using parameter data from all links in the traffic network; and

an observed value, obtained using a real-time vehicular traffic information feed associated with all links in the traffic network.

3. The method of claim 1, wherein the set of traffic predicting parameters includes:

an alpha parameter that reflects a weight applied to a recent past versus a more distant past;

a beta parameter that reflects a number of steps, or hops, between traffic links;

a gamma parameter that reflects a number of weeks of historical data used for a mean calculation;

a zeta parameter that reflects a number of weeks of historical data used for an estimate calculation;

a delta parameter that reflects a number of data points of past data; and

a theta parameter that reflects a quality of data input from a real-time vehicular traffic information feed.

4. The method of claim 3, wherein the mean calculation comprises a calculation of a historical mean value using the gamma parameter.

5. The method of claim 3, wherein the estimate calculation comprises a calculation of a traffic volume of each link in the second subnet, using the zeta parameter.

6. The method of claim 3, wherein the step of generating, by the one or more computer processors, the set of optimized traffic predicting parameters comprises:

selecting the second subnet of traffic links;

increasing the beta parameter one hop;

increasing the alpha parameter, the gamma parameter, the zeta parameter, and the delta parameter by one step; and

executing at least one of the mean calculation, the estimate calculation, and a traffic predicting accuracy calculation, using one or more of the increased alpha, beta, gamma, zeta, and delta parameters.