The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

Driving intention estimation of the vehicle is important for safe driving, and the existing research has focused on the estimation of single driving intention, such as whether a lane change will be made. While in a real-world environment, safe driving requires a more comprehensive estimation for driving intention, such as rapid deceleration, changing lane to left or right, turning left or right. This study presents a complete framework for driving intention estimation in the real world, which consists of three parts: the definition of the driving intention set, the analysis of factors that influence driving intention estimation, and the input data set for driving intention estimation. To verify the validity of the framework, exploit the Berkeley open-source data set-BDD100K as the data source, take into account the typical time series characteristics of driving intention estimation, we use LSTM (Long Short-Term Memory) neural network as an advanced algorithm to model this multi-classification problem. The results show that, in combination with the framework, LSTM is a promising model for estimating vehicle driving intentions in a real traffic environment.

CICTP