Improving Agent-Based Route Predictions for As-The-Crow-Flies Navigation

Mobile navigation has become a ubiquitous technology through mobile devices and smartphones. One of these devices’ most-used navigation methods is the turn-by-turn method, first used in car-based navigation systems. However, pedestrians and cyclists increasingly use other methods such as the as-the-crow-flies (ATCF) navigation method instead of the classical turn-by-turn navigation. Instead of giving fixed instruction, ATCF navigation only indicates the straight-line direction to the destination and users have to make their own decision when and where to turn. In this paper, we improve the route predictions for such alternative navigation methods as ATFC by incorporating a model on users’ angle estimation error. We show that the route predictions are closer to actual human navigation behavior than methods proposed in related work.