traffic lane change detection using mobile phone sensors and geolocation

Driver behavior is a critical factor in traffic safety. detecting abnormal driver behaviors through appropriate indicators and enforcing driving regulations will reduce high-risk driving behaviors and increase traffic safety. detecting dangerous driver behavior is beneficial for developing warning systems and preventing accidents. some high-risk driving behaviors, such as sudden lane changes, are dependent on determining the movement direction of the vehicle which has not received enough attention. the objective of this study is to determine the direction of movement of the vehicle and lane changes, using the sensors in the smartphone mounted on a vehicle. to achieve this goal, first, by using the samsung galaxy s6 smartphone and an accurate global positioning system (gps), longitudinal and angular accelerometer data and gps data are sampled as a dataset and combined by different types of neural networks. then, combined data is fed into a suggested neural network, and lane changes are detected. finally, the gps data is used as the ground truth for the training of the neural network. if the gps is not accessible, this neural network, just by receiving smartphone accelerometer data, can estimate the vehicle’s direction of movement with an accuracy of 0.5 to 4.8 meters compared to the ground truth up to 8 seconds after the gps is shut down. using the vehicle travel path, an algorithm is proposed that can correctly detect the change of driving lane in the sample data set with 94% accuracy, 93.62% precision, 88.00% recall, and f1-score of 90.72% which are acceptable values.