The connected environment provides information on surrounding traffic and areas beyond the visual range traffic to improve driving behavior and avoid dangerous incidents. However, due to the novelty of the connected environment, there is a lack of studies on driver lane-changing intention (LCI). Therefore, this study builds a connected lane-changing scenario based on a driving simulator. The differences of the LCI feature parameters are analyzed, and a driver LCI recognition model is developed. The results show that there is a significant difference in vehicle motion parameters, driver operation parameters, and driver’s eye and head movement parameters during the lane-changing left, lane-keeping, and lane-changing right stages. In addition, the length of the intention time window with connected information (6.6 s) is longer than that of the non-connected information (4.1 s). A driver LCI recognition model is developed using Stacking ensemble learning and then compared with traditional algorithms. The results show that the accuracy of the LCI recognition model using Stacking ensemble learning is higher than that of the traditional algorithms when it is recognized 0.5 s and 3 s before the lane-changing maneuver, which are 98.24% and 92.23%, respectively. The driver LCI is sent to the macro-traffic system or surrounding vehicles through connected communication, which not only helps the macro-traffic system to coordinate and control the traffic flow but also helps surrounding vehicles plan the motion trajectory in advance. In addition, the accurate recognition of driver LCI is beneficial to the transfer and distribution of the human-machine co-driving system control rights.

Vehicle handing stability under the combined acceleration steering and braking steering conditions has a great impact on the vehicle safety. Therefore, the coupled steering and driving/braking stability control is essential. Thus, a novel controller based on differential flatness and vehicle stability region is proposed. First, the reference vehicle state variables are calculated and the vehicle stability region is established based on vehicle dynamics and bifurcation analysis. Then, the flatness-based controller is designed which can ensure the vehicle handing stability by controlling the vehicle states approaching their references. Based on the driver preview model, the vehicle can keep good path tracking effect by tracking the lateral deviation. The simulation results of the double lane change tests show that the proposed control method can ensure both the vehicle handing stability and the path tracking effect. Compared to vehicle equipped with the integration of active front steering and direct yaw moment control, vehicle controlled by flatness-based controller performs better in both handing stability and path tracking.

With the development of communication sensing technology and on-board technology, the future will enter a connected environment. Traffic information is transformed from static isolation to dynamic interconnection, and drivers’ attention distribution and environmental perception abilities are changed. At present, the study on driver lane-changing intention (LCI) has received little attention. Therefore, this paper presents a novel method for recognizing the driver LCI in the connected environment. Firstly, an innovative driving simulator experiment was designed to mimic the connected environment. Secondly, the statistical results demonstrate that the LCI time window is 6.6 s in the connected environment, and the feature parameters of LCI are determined. These parameters include yaw rate, vehicle lateral speed, vehicle lateral acceleration, steering wheel angle, steering wheel torque, gaze direction (x, y, z), head rotation (x, y, z). Finally, a novel driver LCI recognition model is established. The phase space reconstruction and recurrence plot techniques are used to convert the time-series feature parameters into images. The Swin Transformer algorithm with state-of-the-art performance is introduced to classify the images into three categories: lane-changing left, lane keeping, and lane-changing right. This novel method overcomes the problems of vanishing gradient and low recognition accuracy caused by the long time-series input. The accuracy of the proposed LCI model is 89.90% at 3 s prior to the lane-changing maneuver, which is higher than that of the classical machine learning algorithms. This study contributes to the accurate response of lane-changing assistance systems and the smooth transfer of control rights in human-machine co-driving systems.

Road obstacle detection is an important component of the advanced driver assistance system, and to improve the speed and accuracy of road obstacle detection method is a vital task. In this article, fast image region-matching method based on the maximally stable extremal regions method is proposed to improve the speed of image matching. The theoretical feasibility of detection method combining monocular camera with inertial measurement unit (IMU) is clarified. The fast road obstacle detection method based on maximally stable extremal regions combining fast image region-matching method based on maximally stable extremal regions and the vision-IMU-based obstacle detection method is proposed to bypass obstacle classification and to reduce time and space complexity for road environment perception. The AdaBoost cascade detector, the speeded-up robust features-based obstacle detection method, and the proposed method are used to detect obstacles in outdoor contrast tests. Test results show that the proposed method has higher accuracy, and the reason of high accuracy is analyzed. The processing time of AdaBoost cascade detector, speeded-up robust features-based obstacle detection method, and proposed method are compared, and the results show that the proposed method has faster processing speed, and the reason of faster processing speed is analyzed.

Accurate recognition of pedestrian crossing intentions is essential for the safe operation of autonomous vehicles on urban roads. However, the current pedestrian crossing intention recognition model has the problems of relatively low recognition accuracy and short recognition advance time. Based on the above problems, this paper carried out a study on the recognition model of pedestrian crossing intention. Firstly, the pedestrian and vehicle crossing data were collected through laser radar and a high-definition monitor, and 1980 groups of valid samples were selected. Secondly, the pedestrian crossing intention characterization parameter set was determined through statistical analysis. Finally, this paper proposes a pedestrian crossing intention recognition model based on stacking ensemble learning. The ensemble learning framework integrates random forest (RF), support vector machine (SVM), long short-term memory network (LSTM), an attention mechanism, and bidirectional LSTM (AT-Bi-LSTM). Compared with traditional machine learning methods, the proposed method shows greater advantages in recognition accuracy. The model recognition accuracy reaches 95.36% when the model is recognized at 0.5 s before crossing the zebra crossing, and the model recognition accuracy is 89.27% when the model is recognized at 1s before crossing the zebra crossing. The research in this paper is of great significance for building a more intelligent pedestrian-vehicle collaboration and promoting the industrial application of the autonomous vehicle.