Recognizing driver speed intention such as acceleration and deceleration is of great significance for intelligent assisted driving systems, drive energy management, and gear decision of automatic transmissions, among other applications. However, existing studies have mainly focused on recognizing only a few typical speed intentions. They have not adequately considered the effects of various factors of the driving environment, including road slopes, curves, as well as other critical factors like lane changes and vehicle gears, on intention recognition. To address this gap, this study comprehensively categorizes speed intentions and establishes a speed intention recognition model that considers the influence of these factors. First, naturalistic driving data is collected to ensure the robustness and practicality of the model. To integrate the effects of the driving environment into speed intention recognition, the road slope and turning/lane-changing operations of the driver are extracted from driving data. Furthermore, the speed intention is comprehensively categorized. The effects of road slope, vehicle gear, and turning/lane changing on the intention recognition are analyzed separately, and the Toeplitz inverse covariance-based clustering algorithm is used to label the driving data while considering these effects. Finally, a supervised feature selection algorithm is used to select intention recognition features, and a deep-learning-based hierarchical recognition model is established for speed intentions. Validation results indicate that the constructed intention recognition model exhibits excellent recognition performance and satisfies the requirements for real-time recognition.
Deep Learning Approach for Driver Speed Intention Recognition Based on Naturalistic Driving Data
IEEE Transactions on Intelligent Transportation Systems ; 25 , 10 ; 14546-14559
2024-10-01
4459821 byte
Article (Journal)
Electronic Resource
English
Individual driver risk assessment using naturalistic driving data
Online Contents | 2013
|An Unsupervised Approach for Inferring Driver Behavior From Naturalistic Driving Data
Online Contents | 2015
|