Lane changing behavior recognition based on Artificial Neural Network-based State Machine approach

Abstract

Developing driving behavior prediction and recognition models play a crucial role in Advanced Driving Assistance Systems (ADAS). Developing these models generally requires the use of Machine Learning approaches. Often, Machine Learning approaches are difficult to interpret. In this contribution, an Artificial Neural Network (ANN)-based state machine driving behavior recognition model is developed to estimate three lane changing behaviors. A state machine topology defining the relationship between the states (driving behaviors) is developed. The state transitions to another state or remains in the same state based on specific conditions defined by estimations of ANN. Two options are developed: using one common ANN or using three ANN (for three states). Design parameters (weights and biases) defined using optimization describe the ANN estimations when trained. Based on this, lane changing behaviors for the models are estimated. Data from three participants were collected. The results show that this approach performs better than the conventional ANN in terms of DR and FAR with improvements up to 46 % for DR and 34 % for FAR. Based on the results, it can be concluded that the approach introduced realizes high accuracy (ACC), high detection rates (DR), and low false alarm rates (FAR).