Longitudinal Position Prediction of Vehicles at Intersections with a Weighted Average Model

The road user position prediction algorithm Weighted Average Model(WAM) is re-formulated to address one of its main limitations: that WAM can only be applied at geographic locations where data has been collected. This issue is resolved with a change of reference frame, but at the cost of restricting the model to prediction of vehicles' positions along paths. It is additionally shown on real world traffic data that including the curvature of the path a vehicle travels on as input increases WAM's performance. This increase is explained by using the structure of WAM.