Maneuver prediction for road vehicles based on a novel neuro-fuzzy dynamic architecture

Collision avoidance systems for road vehicles may benefit from timely predictions of vehicle maneuvers. This article presents a novel approach for the prediction of maneuvers that copes with noisy measurements and is based on a supervised version of a dynamic FasArt method (SdFasArt). Additionally, the use of size-dependent scatter matrices to compute the activation of the neurons makes the algorithm more adaptable to different data distributions. The results obtained in real tests confirm the goodness of the method.

Research highlights
► Vehicle maneuvers are predicted timely using a neuronal architecture.
► A dynamic FasArt based method copes well with noisy measurements.
► Size-dependent scatter matrices optimize the results and adapt better to different data distributions.