Naturalistic driving data were applied to study driver acceleration behaviour, and a probability model of the driver was proposed. First, the question of whether the database is large enough is resolved using kernel density estimation and Kullback-Liebler divergence. Next, the convergence database is utilised to achieve the bivariate acceleration distribution pattern. Subsequently, two probability models are proposed to explain the pattern. Finally, the statistical characteristics of the acceleration behaviours are studied to verify the probability models. The longitudinal and lateral acceleration behaviours always approximate a similar Pareto distribution. The braking, accelerating, and steering manoeuvres become more intense at first and then less intense as the velocity increases. These behaviours characteristics reveal the mechanism of the quadrangle bivariate acceleration distribution pattern. The bivariate acceleration behaviour of the driver will never reach a circle-shaped pattern. The bivariate Pareto distribution model can be applied to describe the bivariate acceleration behaviour of the driver.

Linear quadratic (LQ) game has been widely used in shared driving to resolve conflicts between drivers and driving automation systems (DAS). However, the system model in LQ game is assumed to be linear time-invariant, which does not consider the effect of changing longitudinal speed on the lateral control. A novel cooperative LPV/MPC approach is proposed, which characterizes the longitudinal and lateral coupled vehicle control in a dynamic game. The stability of the cooperative LPV/MPC system is analyzed in the presence of inconsistent targets of driver and DAS. Driver behavior is studied using 2,701 lane-changing cases extracted from naturalistic driving data (NDD). A human-like risk assessment method is achieved according to the driver behavior characteristics. Finally, a shared driving strategy in lane-changing scenarios is presented based on the cooperative LPV/MPC algorithm and the human-like risk assessment method. Simulation validations show that the shared driving strategy can ensure vehicle stability when longitudinal control, lateral control and dynamic game are performed simultaneously. In addition, the shared driving strategy can realize the driving weight gradual handover, which can help to better assist the driver in risk avoidance.

In this paper, driving data distribution of human drivers in urban driving condition are presented by using China-FOT data. Longitudinal acceleration, lateral acceleration and longitudinal velocity are the three vehicle dynamic parameters which are concerned in the study of human driver behaviors. The data distributions of these three parameters are studied. It is found that the lateral acceleration versus longitudinal acceleration diagram shows the dual triangle distribution, and the speed based diagrams present piecewise linear distribution. Next, whether the database is sufficiently large to get steady distribution and boundary is discussed. And the data distribution boundaries of these three parameters are achieved. In the last, acceleration distribution at different speed region is presented.

Reliable motion prediction of surrounding vehicles is the key to safe and efficient driving of autonomous vehicles, especially at urban intersections with complex traffic environments. This study models driving intentions and future driving speeds at urban intersections and improves model prediction performance by considering traffic environment characteristics. Key feature parameters including environmental characteristics are first extracted through driving behavior analysis and existing research experience. Then models with different input combinations are constructed to explore the effectiveness of different factors in predicting driving intention and future speed. In particular, in vehicle speed modeling, a target detection algorithm is used to identify traffic participants. Based on the identified traffic participant and vehicle position information, a new method for speed prediction that can reflect the dynamic interaction characteristics between the driver and the traffic environment is proposed. Models are trained and tested using natural driving data from China. Finally, the models with the simplest input and the best effect are determined. The driving intention recognition model can accurately predict the driving maneuvers of straight-ahead, stopping, turning left and right 4 seconds before reaching the intersection. The speed prediction model can significantly improve the speed prediction accuracy, and shows stronger robustness and adaptability than existing models. This research provides important technical support for developing intelligent driving systems suitable for complex urban traffic environments.

Accurate estimation of current position and attitude of a vehicle is one of the key technologies for autonomous driving. Due to the defect of LiDAR intrinsic parameter and the sparsity of LiDAR beam in the vertical direction, current LiDAR-based simultaneous localization and mapping (SLAM) system generally suffers from the problem of inaccurate height positioning. In this study, a LiDAR and inertial measurement unit (IMU) tightly coupled localization algorithm considering ground constraint is proposed, which is developed based on a pose graph optimization framework. At the front end, the ground segmentation algorithm Patchwork is improved to obtain a point cloud with higher verticality, which is added to the LiDAR inertial odometry. Moreover, constraints are constructed by using current frame ground points and world map ground points, which are added to factor map optimization to limit elevation errors. At the back end, SC++ descriptors are used to construct loop constraints to eliminate accumulated errors. Verifications based on KITTI dataset show that the height positioning accuracy will be improved through introducing ground constraint factor and loop detection factor. Real vehicle tests indicate that the proposed algorithm has better height positioning accuracy and better robustness compared with the LeGO-LOAM algorithm.

With the rapid and wide implementation of adaptive cruise control system (ACC), the testing and evaluation method becomes an important question. Based on the human driver behavior characteristics extracted from naturalistic driving studies (NDS), this paper proposed the testing and evaluation method for ACC systems, which considers safety and human-like at the same time. Firstly, usage scenarios of ACC systems are defined and test scenarios are extracted and categorized as safety test scenarios and human-like test scenarios according to the collision likelihood. Then, the characteristic of human driving behavior is analyzed in terms of time to collision and acceleration distribution extracted from NDS. According to the dynamic parameters distribution probability, the driving behavior is divided into safe, critical, and dangerous behavior regarding safety and aggressive and normal behavior regarding human-like according to different quantiles. Then, the baselines for evaluation are designed and the weights of different scenarios are determined according to exposure frequency, resulting in a comprehensive evaluation method. Finally, an ACC system is tested in the selected test scenarios and evaluated with the proposed method. The tested vehicle finally got a safety score of 0.9496 (full score: 1) and a human-like score as fail. The results revealed the tested vehicle has a remarkably different driving pattern to human drivers, which may lead to uncomfortable ride experience and user-distrust of the system.