In order to meet the requirements of autonomous vehicle testing on the realness of the test scenes， a coupled feature modeling method is proposed， aiming at the interaction relationship between traffic vehicles. The dataset of traffic vehicle behaviors is established by combining the virtual data built based on mechanism model and the real scene data. Deep learning is adopted to set up the behavior decision model， following model and lane change model of traffic vehicles with local information response. The single model with relatively simple structure can enhance the expandability of scene simulation. In view of the high requirements of autonomous vehicle testing on model robustness， a framework of distributed execution and centralized adversarial training is constructed to conduct traffic vehicle model optimization for enhancing its robustness to input disturbance. The simulation environment for traffic vehicle interaction is created with a simulation on the model performed， and the effectiveness of the model is verified by comparing simulation data distribution with real data one and quantifying evaluation indicators.
Enlightened by that the vibration can be suppressed by periodic structure， a periodic type control structure for vehicle platoon is proposed in this paper， with its interference suppression characteristics emphatically studied. First of all， the vehicle platoon model with periodic type control structure is established， and the external interference is introduced into the acceleration control input to construct the spacing error equation matrix of the whole platoon. Then the transfer function equations from external interference and leading vehicle disturbance to spacing following error are derived. Finally， the interference suppression characteristics of the vehicle platoon with periodic type control structure are analyzed by the maximum singular value of the spacing error transfer function matrix. The results show that the rational design of periodic type control structure can effectively reduce the peak of the maximum singular value of the transfer function matrix from interference to spacing error， and enhance the interference suppression capability of vehicle platoon， even for a heterogeneous one. In addition， the vehicle platoon with periodic type control structure also exhibits a good scale expandability.
Intelligent vehicle path tracking control is faced with disturbances such as system model simplification， parameter uncertainty， the delay of actuator and sensor signals， and road curvature changes， which will generate system disturbance errors， resulting in reduced tracking accuracy. A model predictive control （MPC） method that considers the complex disturbance of the tracking system is proposed in this paper. Firstly， a model prediction tracking system is established based on the single-track vehicle dynamics model， and a dynamic adjustment method of the preview distance based on real-time path planning and speed information is designed to obtain the best preview point to improve the delay disturbance of the actuators and sensor signals of the intelligent vehicle chassis. Then， an extended state observer （ESO） is introduced to estimate the unknown disturbance to the system due to the simplified vehicle model in real time and use it for feed-forward compensation. At the same time， considering the steady-state disturbance error caused by the change of the road reference curvature to the system， a feed-forward control （FFC） method with curvature constraints is designed to eliminate this disturbance； and finally the steering angle control law of the superposition of the feedback input of the MPC controller， the ESO anti-interference compensation input and the FFC input is formed. Finally， real vehicle test and comparison analysis are carried out in a low-speed park scene with a certain brand of intelligent vehicle platform， which verifies the feasibility and superiority of the improved MPC method of integrating disturbance compensation.
In traffic intensive scenarios， the existing DSRC and C-V2X technologies can not ensure efficient data transmission. Firstly， this paper introduces the network architecture of the Internet of vehicles. Secondly， the design of an IEEE 802.11p mechanism based on adaptive primary and secondary window partition is proposed， in order to reduce the collision rate of vehicle nodes and then increase the data transmission rate. Then， by establishing a two-dimensional Markov model， analytical expressions of transmission delay and other performance indexes are obtained. Finally， the improved mechanism is numerically verified. The simulation results show that compared with the classical mechanism， the improved mechanism can obtain better network performance in both sparse and dense traffic scenes. At the same time， a multi-objective optimization model with minimum delay and highest reliability is established to optimize the proposed mechanism so as to obtain balanced system performance.
Aiming at the decision-making difficulties of the trade-off between the vehicle following risk and the driving efficiency caused by heterogeneous vehicle types in the complex traffic environment， a human-like vehicle following model for intelligent vehicles based on dynamic balance of perception risk is proposed on the basis of analyzing natural driving data. Firstly， an empirical model of the vehicle following distance is established for vehicle following modes with four different truck-car combination， and the “two invariances” law of time headway （THW） and inverse time to collision （i-TTC） existing in the driver’s steady-state vehicle following behavior is discovered， with the balance lines obtained by drawing method. Then， the mechanism of vehicle following decision-making is revealed from the perspective of the dynamic balance between perception risk and acceleration response during driving， and the commonly-used vehicle following models are unified within the framework of dynamic balance of perception risk. Finally， a simple nonlinear function is proposed as a mathematical expression of dynamic balance of perception risk， and the accuracy of the model is verified by using the tested vehicle following data.
An all-wheel steering path tracking strategy based on model predictive control with robust invariant set is proposed for a five-axle heavy-duty vehicle in this paper. Firstly， an all-wheel steering path tracking strategy based on first-axle and fifth-axle steering angle control is put forward to make the multi-axle vehicle more flexible in control， with the lateral force response synchronized and fully utilized. Then with considerations of the uncertainty of tire parameters and the bounded disturbance caused by side-wind in the control model， the Tube MPC is used to solve the path tracking problem. Meanwhile， a simplified minimum robust positively invariant set （mRPI） based on support function calculation is adopted to replace the general Minkowski sum-based mRPI operation， effectively saving the offline computation time of mRPI， reducing the number of vertices in the invariant set， and ensuring the online implementation of Tube MPC. Finally， a hardware-in-the-loop simulation is carried out with a result verifying that the Tube MPC-based all-wheel steering strategy proposed has higher path tracking accuracy and vehicle stability and stronger robustness when facing unknown disturbance， compared with ordinary all-wheel steering strategy.
When the sum of the driving rights of the driver and autonomous driving system is fixed to 1， the vehicle steering demand may not match the output of the shared steering control （SSC） system in complex scenarios， which will affect the driving safety of the vehicles. To solve the problem， an authority dynamic allocation approach is proposed， and a robust controller is designed considering the uncertainty of the authority levels and model parameters. Firstly， based on the parallel SSC system framework， an SSC system expansion framework including the driver model， trajectory tracking controller， authority dynamic allocation model， and vehicle model is proposed. The driving state evaluation model， vehicle trajectory tracking state evaluation model， and authority level allocating and decision-making model are proposed. The results of driver’s driving state evaluation and vehicle trajectory tracking state evaluation results are used as the basis for decision-making， combined with the calculation result of the authority allocation model， the authority level of the driver and controller can be determined dynamically. Then， the robust feedback controller is proposed considering the uncertainty of the authority level and vehicle model parameters. Finally， the effectiveness of the proposed SSC system is verified by simulation. The simulation results show that the proposed SSC system can effectively reduce the influence of interference and driver’s misoperation on vehicle safety， and can reduce the driver’s driving load， psychology load and controller workload during steering.
The vehicle dynamics modeling process based on mechanism analysis is usually simplified with assumptions，which can't accurately calculate the dynamic changes of actual vehicles under different road conditions， thus causing problems such as low trajectory tracking control accuracy and instability of intelligent automotive. To tackle the above-mentioned problems， this paper proposes a non-linear modeling and control method based on hybrid modeling technology. By constructing mechanism analysis - data-driven vehicle dynamics series hybrid model， the vehicle state and control data are calculated and processed by the mechanism model， and then used as the input of the data-driven module after a level combination. Besides， long-short-term memory network used as the backbone realizes the nonlinear correlation feature extraction of time-series data and the final model output calculation. The test results show that the model can supplement some unmodeled dynamics in the mechanism model， improve the model calculation accuracy and has the ability to implicitly understand different road adhesion conditions. In addition， the Euler integration is used to complete the discretization of the prediction model and design the model predictive control track tracking algorithm. The feedforward feedback control algorithm is designed to provide external input required by the prediction model in the horizontal control while realizing the longitudinal control of the vehicle， finally achieving more accurate trajectory tracking control effect that is more in line with the actual driving environment. The co-simulation results of Carsim / Simulink show that the method achieves accurate output of different road attachment coefficients， synchronously enhances the intelligent automotive trajectory tracking control accuracy and stability， and has good horizontal and longitudinal coordination control.
In order to ensure the safety of pedestrians and the stability of vehicle in the emergency pedestrian avoidance of human-machine co-driven vehicle， a human （driver）-vehicle cooperative game collision avoidance strategy based on pedestrian asymmetric potential field is proposed. Firstly， with full consideration of the street-crossing characteristics of pedestrian and his relative motion with vehicle， an asymmetric double elliptical pedestrian potential field is established for better characterize pedestrian risk， based on which the path planning for collision avoidance is performed. Then， for enhancing vehicle stability during collision avoidance and ensuring trajectory tracking performance， a non-cooperative game-based driver-AFS-ARS three-way synergy controller is constructed with a simulation on the condition of pedestrian avoidance is conducted for verification. The results show that with the ARS control added， not only the trajectory tracking performance in collision avoidance is ensured， the stability of vehicle is also apparently enhanced， with its average absolute value of error in lateral speed being 46.43% less than driver-AFS cooperative control.
In view of the lowering of the trajectory tracking accuracy and the solution efficiency caused by the increase of nonlinear degree and dynamic constraints of unmanned vehicles under complex working conditions， an efficient algorithm based on nonlinear model predictive control （NMPC） is proposed in this paper. Firstly， in consideration of the nonlinear factors of the vehicle model， the dynamic model and the magic formula tire model are established. A terminal state is integrated to the performance index. The multi-constraint conditions within the stability range are added， and barrier function method is used to solve nonlinear inequality constraints to ensure the smoothness of the solution process. Then in order to reduce the computational burden caused by solving nonlinear optimization problems， an improved continuous/generalized minimum residual （improved-C/GMRES） algorithm is proposed. Compared with the traditional C/GMRES algorithm， the continuously increasing penalty factor is introduced to speed up the numerical calculation efficiency and reduce the computational burden of the algorithm. Finally， based on the joint simulation platform of Simulink and Carsim， the trajectory tracking accuracy and solution efficiency are verified in double-shift line motion and serpentine motion. Simulation results show that compared with the traditional C/GMRES algorithm， the proposed algorithm can significantly improve the tracking accuracy and driving stability of trajectory tracking， and greatly accelerates the solution efficiency.
L3 level automatic driving system allows the driver to leave from driving tasks temporarily within its Operational Design Domain （ODD）. When the system reaches its boundary， it issues a Take-over Request （TOR）， alerting the driver to take over the vehicle in time. In order to explore the take-over performance of drivers with different driving experience in L3 automatic driving， emergency obstacle avoidance driving scenarios with non-driving tasks and visual assistance are designed for 24 participants of different ages with significantly different driving experience to carry out the take-over experiment on the driving simulator platform. The results show that different driving experience has no significant effect on reaction time， but there is significant difference in take-over performance. Experienced drivers have a significant improvement in the metrics like maximum steering wheel Angle and standard deviation of steering wheel Angle； i.e.， experienced drivers can resume the control of the vehicle more steadily and safely. In addition， the difference between experienced and inexperienced is more significant without visual assistance.
The limitations on performance， insufficient specifications or reasonably foreseeable misuse lead to the issues related to the safety of the intended functionality （SOTIF） emerging one after another， seriously hindering the rapid development of intelligent vehicles. This paper focuses on the key technologies to ensure the SOTIF， and systematically summarizes the related state of art of research from system development， functional improvement， and operation three phases. Finally， the research prospect is foreseen from the perspectives of basic theory， risk prevention and update mechanism. The review can provide an important reference for the research on the safety of the intended functionality for intelligent vehicles.
The safety risks the autonomous vehicles face not only come from the deficiencies of functional safety and information safety， but also stem from the insufficiencies of the safety of the intended functionality （SOTIF）. As an important part of autonomous driving system， the automatic emergency braking （AEB） system has SOTIF insufficiencies in control strategies. In this paper， the system theoretical process analysis method is adopted to analyze the safety of AEB system， find out the trigger events that may cause harm and put forward the corresponding safety objectives. Aiming at the safety objectives， a control strategy for AEB system based on detailed scenes are proposed. The control strategy proposed for AEB system is then verified in CarSim-MATLAB/Simulink environment. The results show that after functional corrections in both the event acceptance criterion and total risk acceptance criterion， the risk level of the system becomes much lower and the safety level of the system is apparently enhanced.
Based on a low-cost monocular visual perception system， a multi-target trajectory prediction method considering the influence of vehicle motion is proposed in this paper. Firstly， the ego vehicle motion estimation model composed of the depth estimation network and the position and orientation estimation network is established to achieve effective calculation of ego vehicle visual odometer in image sequence. Then， a prediction model is built by using the historical position and orientation sequences of ego vehicle， and a normalization processing on the historical positions of surrounding targets is fulfilled under the current perspective of ego vehicle camera. Finally， the prediction network is constructed based on the historical trajectory information and regional image information to realize the effective prediction of surrounding multi-target trajectories around intelligent vehicles. The innovation points of this paper are combining visual SLAM method with trajectory predictive model and putting forward the new motion estimation model and ConvLSTM-based multi-target trajectory predictive network. The model proposed overcomes the adverse influence of existing research on the trajectory prediction accuracy of the surrounding target caused by ignoring the ego vehicle motion state， and achieves a better prediction results under the condition of using monocular vision perception only. The results of test on public data sets show that with a prediction time step of 1.5s， the model proposed has a MSEcenter of 321， i.e. a 52% lower than that of the existing baseline model， with an excellent performance also in the long-time-step trajectory prediction in the future.
In order to enhance the driving efficiency of connected and automated vehicles （CAVs） at traffic bottlenecks， a platoon cooperative control strategy at an unsignalized intersection is proposed. Firstly， a control framework for the allocation of the right of way for platoons is put forward based on the traffic flow model and occupied time of platoons at the intersection. Then， a Q-learning model is designed to conditionally select platoon sizes， with instantaneous efficiency and travel delays as compound indicators. Finally， an online trajectory planning simulation is carried out for the grouped vehicles based on vehicle following model. The results show that the Q-learning model can flexibly allocate the platooning commands according to different working conditions and ensure the overall safety of platoons during driving process. Compared with the nonplatoon scheme， the traffic capacity of the intersection is increased by around 36.1%.
The modeling and cooperative control method of mixed vehicle platoon consisting of both human-driven vehicle （HDV） and connected and automated vehicle （CAV） under multi-delay condition is studies in this paper. Firstly， a generalized model for mixed vehicle platoon system is constructed to characterize the number and spatial distribution of CAVs， with the time delay of driver’s response introduced into the model. Then， with consideration of V2V communication time delay and sensor measurement time delay， a state feedback controller is designed， the characteristic equation of the closed-loop system is derived， and the sufficient conditions for the closed-loop stability of the system are determined according to the Routh Hurwitz criterion to guide the parameter design of the controller. Finally， numerical simulation is conducted to verify the effectiveness of the controller design， quantitatively exhibiting the role of CAV in attenuating traffic fluctuation， with the effects of various time delays on traffic flow fluctuation analyzed.
Aiming at the issue of safety of intended functionality （SOTIF） of adaptive cruise control （ACC） system caused by the large unintended measurement error of target movement of millimeter wave radar in heavy rainfall scenes， a SOTIF safety control strategy for ACC system in heavy rainfall scenes is proposed in this paper. Firstly， double-state Chi-square is used to check the target information output by millimeter-wave radar to determine whether there exist SOTIF risks. Then the target information with safety risk is corrected by Kalman filter， with the target information corrected input into ACC controller to control vehicle motion state for achieving safety control. Finally， the Prescan/Simulink co-simulation platform is built to conduct verification simulation on the safety control strategy proposed. The results show that double-state Chi-square test can timely detect risk information with a detection time deviation within 1.31 s and a distance correction error of Kalman filter within 3.66 m， effectively ensuring the safe and stable operation of ACC system in heavy rainfall scenes.
The research and application of unmanned ground vehicles （UGVs） become more popular in both civilian and military fields. UGVs completely eliminate the human driving operating mechanisms， and adopt full X-by-wire control framework and distributed all-wheel independent drive/braking/steering techniques instead， with each wheel independently controlled to realize multi-steering modes and obtain high maneuver trajectory tracking ability which traditional vehicle cannot achieved. This paper aims to conduct a research on the all-wheel diagonal steering （vividly called crab walk steering） control of full X-by-wire controlled UGVs. Firstly， a stability robust control algorithm with a reconfigurable model is proposed， the robust control target and parameter perturbation dynamics model for diagonal maneuver are established， and a robust H2/H∞ stability controller is constructed to achieve a point-to-point yaw-less direct movement of UGVs. Then a decoupled control between body motion posture and movement trajectory is realized through comprehensive cooperated dynamics control， so significantly enhancing the trajectory tracking capability of UGVs in complex space. Finally， a test on a six-wheel UGV prototype is carried out to verify the control effects of the robust controller designed.
In view of that the following speed of network-connected hybrid electric truck in a fleet not only affects the driving safety， energy demand and distribution and battery aging rate， but also influence the aerodynamic drag and even the energy consumption through vehicle spacing， the speed planning and energy management strategy under vehicle following scenes are proposed in this paper with concurrent considerations of multi-objectives covering driving safety， energy consumption， aerodynamic drag and battery aging. Firstly， the vehicle following safety is quantified based on aerodynamics. Then， a real time control strategy based on model predictive control is constructed with minimizing the total equivalent cost consisting of safety cost， energy consumption cost and battery aging cost as objective function， in which the speed of front vehicle is predicted by using long- and short-term memory-based neural network， and the optimization problem in rolling time domain is solved with dynamic programming. The results show that the cooperative control strategy proposed can lower the battery aging cost via restraining the charging and discharging current， and reduce the aerodynamic drag and hence the energy cost by flexibly adjust following vehicle spacing. The results of comparison with human-driven-model-based following strategies verify the feasibility of cooperative control strategy proposed.
Reservation-based approach is a classical method of multi-vehicle cooperative controlling at traffic-lightless intersections. The traditional reservation-based approach usually assumes that only intelligent and connected vehicles （ICVs） exist at the intersection， which lacks discussion on the coexistence of human-driven vehicles （HDVs） and ICVs in this scenario. In this paper， based on the reservation-based approach， the cooperative control method for traffic-lightless intersections with mixed ICVs and HDVs is studied， and the impact of ICV penetration on traffic efficiency at traffice-lightless intersections is explored. Firstly， the functional zones of the traffic-lightless intersection under mixed traffic flow are divided， and the architecture of the reservation-based multi-vehicle cooperative control system for mixed traffic flow is proposed. Furthermore， the reservation-based multi-vehicle cooperative control strategy for mixed traffic flow is formulated， including the ICV cell reservation strategy and ICV speed control strategy considering the characteristics of HDVs. Finally， the SUMO/Python joint simulation platform is built to simulate the strategy under different traffic flows with the ICV penetration of 30%， 60%， 90% and 100% respectively， with the intersection passing rate and the average speed of the road section as the evaluation indicators. The results show that the proposed control strategy can ensure that all vehicles can pass the intersection safely， and the efficiency of traffic-lightless intersections increases with the increase of ICV penetration.
