A probabilistic approach for automated lane identification based on sensor information

The level lane location problem of sensor equipped vehicles circulating within arbitrary highway infrastructures is addressed. A first approach of a flexible probabilistic decision-making policy is developed utilizing sensor signals. Unmanned vehicles independently of the automation degree are related to challenging executive schemes such as adaptive cruise control systems, real time routing models involving lane changing options and speed control, platoon formation operations etc. An adaptive, closed loop methodology is presented localizing suitable detections while involving uncertainty within data, sensor vagueness and trust. The whole scheme is associated with low computational complexity where no additional investment on external devices is required. The outlined framework pronounces a significantly progressed study regarding a previously presented elementary pattern. The new model focuses in the case of invalid sensor detections due to traffic context, various environmental disturbances and failures for which no response was previously available. The effectiveness of the suggested scheme is measured when applied to detailed simulation scenarios fed by ground truth data. Different complex spatiotemporal contexts elicit varying driving profiles and pragmatic behavior-change interventions unaccessible from direct recordings provided by professional drivers. The proposed methodology is compared with a non-probabilistic model. Analysis illustrates noteworthy accuracy, precision and frequency on the resulting responses.