Filtering of traffic-irrelevant GPS data

We proposed an algorithm for filtering unusual GPS data from multiple GPS-equipped cars. The GPS data was periodically transmitted to a base station via a radio link every 3 minutes and 10 seconds by 3000-4000 taxis around Bangkok and suburb. We manually analyzed the data and mark the data that seems irregular. From our analysis of the data, there are four main problems affect the accuracy of traffic estimation system: (1) Taxi driver drives round one place for a long time, (2) Taxi waiting for passengers, (3) The direction of a car does not flow along the direction of the street, (4) The measurement error of the GPS receivers. If these invalid data is filtered out before using it to estimate traffic congestion, the system will be more consistent and accurate. We used the coefficient of variations (C.V.) as a performance indicator, because invalid data is the cause of data inconsistency. From the experiment, our proposed algorithm could reduce the C.V. from 0.9179 to 0.8184.