Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION Model
Ensuring the driver’s readiness to take over before a takeover request is issued by an autonomous driving system is crucial for a safe takeover. However, current takeover prediction models suffer from poor prediction accuracy and do not consider the time dependence of input features. In this regard,...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Systems |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-8954/13/1/46 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832587460057497600 |
|---|---|
| author | Lijie Chen Daofei Li Tao Wang Jun Chen Quan Yuan |
| author_facet | Lijie Chen Daofei Li Tao Wang Jun Chen Quan Yuan |
| author_sort | Lijie Chen |
| collection | DOAJ |
| description | Ensuring the driver’s readiness to take over before a takeover request is issued by an autonomous driving system is crucial for a safe takeover. However, current takeover prediction models suffer from poor prediction accuracy and do not consider the time dependence of input features. In this regard, this study proposes a hybrid LSTM-BiLSTM-ATTENTION algorithm for driver takeover performance prediction. By building a takeover scenario and conducting experiments in the driving simulation experimental platform under the human–machine co-driving environment, the relevant state indicators in the 15 s per second before the takeover request is sent are extracted from three perspectives, namely, driver state, traffic environment, and personal attributes, as model inputs, and the level of takeover performance was labeled; the hybrid LSTM-BiLSTM-ATTENTION algorithm is used to construct a driver takeover performance prediction model and compare it with other five algorithms. The results show that the algorithm proposed in this study performs optimally, with an accuracy of 93.11%, a precision of 93.02%, a recall of 93.28%, and an F1 score of 93.12%. This study provides new ideas and methods for realizing the accurate prediction of driver takeover performance, and it can provide a decision basis for the safe design of self-driving vehicles. |
| format | Article |
| id | doaj-art-d658f6281389423389d393d3f88809ca |
| institution | Kabale University |
| issn | 2079-8954 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Systems |
| spelling | doaj-art-d658f6281389423389d393d3f88809ca2025-01-24T13:50:36ZengMDPI AGSystems2079-89542025-01-011314610.3390/systems13010046Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION ModelLijie Chen0Daofei Li1Tao Wang2Jun Chen3Quan Yuan4College of Automotive Engineering, Guangxi Technological College of Machinery and Electricity, Nanning 530007, ChinaInformation Management Center of Transportation, Nanning 530000, ChinaGuangxi Key Laboratory of Intelligent Transportation System, Guilin 541004, ChinaSchool of Transportation, Southeast University, Nanjing 211189, ChinaState Key Laboratory of Automotive Safety and Energy, School of Vehicle & Mobility, Tsinghua University, Beijing 100084, ChinaEnsuring the driver’s readiness to take over before a takeover request is issued by an autonomous driving system is crucial for a safe takeover. However, current takeover prediction models suffer from poor prediction accuracy and do not consider the time dependence of input features. In this regard, this study proposes a hybrid LSTM-BiLSTM-ATTENTION algorithm for driver takeover performance prediction. By building a takeover scenario and conducting experiments in the driving simulation experimental platform under the human–machine co-driving environment, the relevant state indicators in the 15 s per second before the takeover request is sent are extracted from three perspectives, namely, driver state, traffic environment, and personal attributes, as model inputs, and the level of takeover performance was labeled; the hybrid LSTM-BiLSTM-ATTENTION algorithm is used to construct a driver takeover performance prediction model and compare it with other five algorithms. The results show that the algorithm proposed in this study performs optimally, with an accuracy of 93.11%, a precision of 93.02%, a recall of 93.28%, and an F1 score of 93.12%. This study provides new ideas and methods for realizing the accurate prediction of driver takeover performance, and it can provide a decision basis for the safe design of self-driving vehicles.https://www.mdpi.com/2079-8954/13/1/46human–machine codrivingdriving simulationeye trackingperformance prediction for takeoverdeep learning |
| spellingShingle | Lijie Chen Daofei Li Tao Wang Jun Chen Quan Yuan Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION Model Systems human–machine codriving driving simulation eye tracking performance prediction for takeover deep learning |
| title | Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION Model |
| title_full | Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION Model |
| title_fullStr | Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION Model |
| title_full_unstemmed | Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION Model |
| title_short | Driver Takeover Performance Prediction Based on LSTM-BiLSTM-ATTENTION Model |
| title_sort | driver takeover performance prediction based on lstm bilstm attention model |
| topic | human–machine codriving driving simulation eye tracking performance prediction for takeover deep learning |
| url | https://www.mdpi.com/2079-8954/13/1/46 |
| work_keys_str_mv | AT lijiechen drivertakeoverperformancepredictionbasedonlstmbilstmattentionmodel AT daofeili drivertakeoverperformancepredictionbasedonlstmbilstmattentionmodel AT taowang drivertakeoverperformancepredictionbasedonlstmbilstmattentionmodel AT junchen drivertakeoverperformancepredictionbasedonlstmbilstmattentionmodel AT quanyuan drivertakeoverperformancepredictionbasedonlstmbilstmattentionmodel |