System and method for incident detection with spatiotemporal thresholds estimated via nonparametric quantile regression

1. A method for incident detection, the method comprising:

designating a plurality of decision variables for incident detection;

detecting, at a predetermined location over a predetermined time period, data for determining the plurality of decision variables;

determining the plurality of decision variables over the predetermined time period;

calibrating a plurality of bands for the incident detection based on the determined decision variables, wherein:

the plurality of bands define thresholds that are time-varying for all measurement locations in a network; and

the thresholds are estimated using nonparametric quantile regression;

detecting data corresponding to a decision variable at a particular time t;

determining the decision variable at the time t;

querying whether the decision variable at the time t falls outside a band of the plurality of bands to determine whether an incident has occurred; and

issuing an alert in response to a determination that the incident has occurred;

wherein the querying comprises comparing the decision variable at the time t with a time-varying threshold defined by the band at the time t;

wherein the detecting is performed by at least one sensor;

wherein the sensor records occupancy at the predetermined location over the predetermined time period by recording a passage of each traveling element on the network over a fixed loop;

wherein the alert is issued at a command center capable of responding to the incident; and

wherein one or more steps of the method are performed by a computer system comprising a memory and at least one processor coupled to the memory.

2. The method according to claim 1 , wherein a determination of an occurrence of a downstream incident and an occurrence of an upstream incident in the network is performed using a same band of the plurality of bands.

3. The method according to claim 1 , wherein at least two thresholds respectively comprise upper and lower limits of a band of the plurality of bands and a width of the band is variable over time.

4. The method according to claim 1 , wherein the decision variables represent a shock in occupancy and an occupancy level at a time.

5. The method according to claim 4 , wherein calibration is performed using a plurality of parameters, wherein first and second parameters are quantiles for shocks in occupancies, and third and fourth parameters are quantiles for occupancy levels.

6. The method according to claim 5 , wherein fifth and sixth parameters control a degree of smoothness for respective conditional quantile functions for the shocks in occupancies and the occupancy levels.

7. The method according to claim 1 , wherein:

the comparing comprises determining whether the decision variable at the time t is less than or greater than a conditional quantile function for the decision variable to determine whether the incident has occurred.

8. The method according to claim 1 , further comprising utilizing variable state values representing both upstream and downstream incidents in the network.