Explore the vast range of titles available.

High-precision lane keeping is essential for the future autonomous driving. However, due to the imbalanced and inaccurate datasets collected by human drivers, current end-to-end driving models have poor lane keeping the effect. To improve the precision of lane keeping, this study presents a novel multi-state model-based end-to-end lane keeping method. First, three driving states will be defined: going straight, turning right and turning left. Second, the finite-state machine (FSM) table as well as three kinds of training datasets will be generated based on the three driving states. Instead of collecting the dataset by human drivers, the accurate dataset will be collected by the high-performance path following controller. Third, three sets of parameters based on 3DCNN-LSTM model will be trained for going straight, turning left and turning right, which will be combined with FSM table to form a multi-state model. This study evaluates the multi-state model by testing it on five tracks and recording the lane keeping error. The result shows the multi-state model-based end-to-end method performs the higher precision of lane keeping than the traditional single end-to-end model.

Lane-keeping is a basic function of an intelligent vehicle. But the existing lane-keeping methods may not provide the expected effect. A vehicle often deviates from the desired lane despite the working lane-keeping controller in practice. For addressing this issue, we propose a novel lane-keeping control method based on the homogeneous domination control theory to improve the lane-keeping system performance in this paper. Firstly, a two-degree-of-freedom lane-keeping dynamic model is built. Then, the state equations of the lane-keeping control system are obtained based on the dynamic model. A lane-keeping state feedback controller is designed via the homogeneous domination method. We prove that the designed controller can globally asymptotically stabilize the system via the Lyapunov method. The proposed homogeneous domination method does not require the nonlinear terms of the nonlinear system to meet the strict linear growth condition. Numerical simulation and hardware-in-the-loop test results show that the proposed homogeneous controller has strong robustness, fast response, and low energy output which are more suitable for the lane-keeping system and improves the lane-keeping system performance.

In this paper, we present a data-driven modeling method for lateral motion control of unknown vehicle models. Vehicle’s motion can be modeled linearly but this model has complex and nonlinear characteristic. Therefore, it is necessary to know the exact information of the car chassis and requires a knowledge and understanding of dynamics. To solve these drawbacks, we linearly represent full vehicle's lateral dynamics which include nonlinear behavior using dynamic mode decomposition (DMD), one of the data driven modeling methods. To determine the validity of the model obtained using the DMD method, we conducted a simulation of the comparison of the output states between the existing model and the model obtained through DMD modeling, using the scenario of a dynamic maneuver called a double line change during lateral motion of a vehicle. After determination of validation is completed, we designed a lane keeping system by applying a model predictive control to specifically evaluate the model of the proposed method. Performance was derived by comparing the error caused by the vehicle driving on the course with the controller of the simulation. The performance of the proposed approach has been evaluated through simulations and is useful when the model is inaccurate.

According to WHO, 1.35 million people, every year are cut short in road accidents, most of them caused due to human misconduct and ignorance. To improve safety over the roads, road perception and lane detection play a crucial part in avoiding accidents. Lane Detection is a constitution for various Advanced Driver Assisting System (ADAS) like Lane Keeping Assisting System (LKAS) and Lane Departure Warning System (LDWS). It also enables fully assistive and autonomous navigation in self-driving vehicles. Therefore, it has been an effective field of research for the past few decades, but various milestones are yet to be achieved. The problem has encountered various challenging scenarios due to the past limitations of resources and technologies. In this paper, we reviewed the different approaches based on image processing and computer vision that have revolutionized the lane detection problem. This paper also summarizes the different benchmark data sets for lane detection, evaluation criteria. We implemented Lane detection system using Unet and Segnet model and applied it on Tusimple dataset. The Unet performance is better as compared to Segnet model. We also compare the detection performance and running time of various methods, and conclude with some current challenges and future trends for deep learning-based lane marking detection algorithm. Finally, we compare various researcher’s approaches with their performances. This paper concluded with the challenges to predict accurate lanes under different scenarios.

The lateral motion of an Automated Vehicle (AV) is highly affected by the model’s uncertainties and unknown external disturbances during its navigation in adverse environmental conditions. Among the variety of controllers, the sliding mode controller (SMC), known for its robustness towards disturbances, is considered to generate a robust control signal under uncertainties. However, conventional SMC suffers from the issue of high frequency oscillations, called chattering. To address the issue of chattering and reduce the effect of unknown external disturbances in the absence of precise model information, a radial basis function neural network (RBFNN) is employed to estimate the equivalent control. Further, a higher order sliding mode (HOSM) based switching control is proposed in this paper to compensate for the effect of external disturbances. The effectiveness of the proposed controller in terms of lane-keeping and lateral stability is demonstrated through simulation in a high-fidelity Carsim-Matlab Simulink environment under a variety of road and environmental conditions.

In recent years, driving simulators have been widely used by automotive manufacturers and researchers in human-in-the-loop experiments, because they can reduce time and prototyping costs, and provide unlimited parametrization, more safety, and higher repeatability. Simulators play an important role in studies about driver behavior in operating conditions or with unstable vehicles. The aim of the research is to study the effects that the force feedback (f.f.b.), provided to steering wheel by a lane-keeping-assist (LKA) system, has on a driver’s response in simulators. The steering’s force feedback system is tested by reproducing the conditions of criticality of the LKA system in order to minimize the distance required to recover the driving stability as a function of set f.f.b. intensity and speed. The results, obtained in three specific criticality conditions, show that the behaviour of the LKA system, reproduced in the simulator, is not immediately understood by the driver and, sometimes, it is in opposition with the interventions performed by the driver to ensure driving safety. The results also compare the performance of the subjects, either overall and classified into subgroups, with reference to the perception of the LKA system, evaluated by means of a questionnaire. The proposed experimental methodology is to be regarded as a contribution for the integration of acceptance tests in the evaluation of automation systems.

Recently, research concerning autonomous self-driving vehicles has become very popular. In autonomous vehicles (AVs), different decision-making and learning architectures have been proposed to predict multiple tasks (MTs) or MTs from various datasets or to improve performance among different MTs. In this paper, a novel and unified multitask learning framework, called Multi-Modal DenseNet ( M 2 -DenseNet), is proposed to predict MTs in a single network in which three long short-term memory units act as the output (MTs). Accordingly, the proposed M 2 -DenseNet can predict three different motion decision-making tasks, i.e., the steering angle, speed, and throttle, to control AV driving. Moreover, M 2 -DenseNet can greatly reduce the time complexity (e.g., to less than 5 ms) because the different prediction tasks can be predicted simultaneously. We conduct comprehensive experiments with the lane-keeping task based on two control mechanisms using the proposed M 2 -DenseNet and other existing methods to evaluate the performance. The experiments demonstrate that M 2 -DenseNet significantly outperforms other state-of-the-art methods with the accuracies of the three control tasks being approximately 98%, 99%, and 98%, respectively. The mean squared error between the predicted value and the ground truth is reported in the experiments, with values for the steering angle, speed, and throttle of 0.0250, 0.0210, and 0.0242, respectively.

In recent years, Reinforcement Learning (RL) has excelled in the realm of autonomous vehicle control, which is distinguished by the absence of limitations, such as specific training data or the necessity for explicit mathematical model identification. Particularly in the context of lane keeping, a diverse set of rewarding strategies yields a spectrum of realizable policies. Nevertheless, the challenge lies in discerning the optimal behavior that maximizes performance. Traditional approaches entail exhaustive training through a trial-and-error strategy across conceivable reward functions, which is a process notorious for its time-consuming nature and substantial financial implications. Contrary to conventional methodologies, the Monte Carlo Tree Search (MCTS) enables the prediction of reward function quality through Monte Carlo simulations, thereby eliminating the need for exhaustive training on all available reward functions. The findings obtained from MCTS simulations can be effectively leveraged to selectively train only the most suitable RL models. This approach helps alleviate the resource-heavy nature of traditional RL processes through altering the training pipeline. This paper validates the theoretical framework concerning the unique property of the Monte Carlo Tree Search algorithm by emphasizing its generality through highlighting crossalgorithmic and crossenvironmental capabilities while also showcasing its potential to reduce training costs.

Regarding the lane keeping system, path tracking accuracy and lateral stability at high speeds need to be taken into account especially for commercial vehicles due to the characteristics of larger mass, longer wheelbase and higher mass center. To improve the performance mentioned above comprehensively, the control strategy based on improved artificial potential field (APF) algorithm is proposed. In the paper, time to lane crossing (TLC) is introduced into the potential field function to enhance the accuracy of path tracking, meanwhile the vehicle dynamics parameters including yaw rate and lateral acceleration are chosen as the repulsive force field source. The lane keeping controller based on improved APF algorithm is designed and the stability of the control system is proved based on Lyapunov theory. In addition, adaptive inertial weight particle swarm optimization algorithm (AIWPSO) is applied to optimize the gain of each potential field function. The co-simulation results indicate that the comprehensive evaluation index respecting lane tracking accuracy and lateral stability is reduced remarkably. Finally, the proposed control strategy is verified by the HiL test. It provides a beneficial reference for dynamics control of commercial vehicles and enriches the theoretical development and practical application of artificial potential field method in the field of intelligent driving.

Autonomous driving has become a popular area of research in recent years, with accurate perception and recognition of the environment being critical for successful implementation. Traditional methods for recognizing and controlling steering rely on the color and shape of traffic lights and road lanes, which can limit their ability to handle complex scenarios and variations in data. This paper presents an optimization of the You Only Look Once (YOLO) object detection algorithm for traffic light detection and end-to-end steering control for lane-keeping in the simulation environment. The study compares the performance of YOLOv5, YOLOv6, YOLOv7, and YOLOv8 models for traffic light signal detection, with YOLOv8 achieving the best results with a mean Average Precision (mAP) of 98.5%. Additionally, the study proposes an end-to-end convolutional neural network (CNN) based steering angle controller that combines data from a classical proportional integral derivative (PID) controller and the steering angle controller from human perception. This controller predicts the steering angle accurately, outperforming conventional open-source computer vision (OpenCV) methods. The proposed algorithms are validated on an autonomous vehicle model in a simulated Gazebo environment of Robot Operating System 2 (ROS2).