Many people die each year in the world single vehicle roadway departure crashes caused by driver inattention, especially on freeway. Lane Departure Warning System (LDWS) is a useful system to avoid those accident, which, lane detection key issue. In this paper, after brief overview of existing methods, we present robust algorithm based geometrical model and Gabor filter. This two assumptions: road front approximately planar marked which are often correct highway freeway where most accidents...

The recognition and tracking of traffic lights for intelligent vehicles based on a vehicle-mounted camera are studied in this paper. candidate region the light is extracted using threshold segmentation method morphological operation. Then, algorithm machine learning employed. To avoid false negatives loss, target CAMSHIFT (Continuously Adaptive Mean Shift), which uses color histogram as model, adopted. In addition to signal pre-processing learning, initialization problem search window...

In this paper, we introduce a novel and efficient hybrid trajectory planning method for autonomous driving in highly constrained environments. The contributions of paper are fourfold. First, present framework that is able to handle geometry constraints, nonholonomic dynamics constraints cars humanlike layered fashion generate curvature-continuous, kinodynamically feasible, smooth, collision-free trajectories real time. Second, derivative-free global path modification algorithm extract...

Lane change maneuver of high-speed automated vehicles is complicated, since it involves highly nonlinear vehicle dynamics, which critical for the driving safety and handling stability. Addressing this challenge, we present dynamic modeling control lane maneuver. A single-track dynamics model a multisegment process are employed. Variable time steps utilized discretization to ensure long enough prediction horizon, while maintaining fidelity computational feasibility. Accordingly, addressed in...

In complex and dynamic urban traffic scenarios, the accurate prediction of trajectories surrounding participants (vehicles, pedestrians, etc) with interactive behaviours plays an important role in navigation motion planning ego vehicle. this paper, based on graph neural network (GNN), we propose a hierarchical GNN framework to model interactions heterogeneous pedestrians riders) combined LSTM predict their trajectories. The proposed consists two modules GNNs for events recognition (IER)...

Data visualization can communicate information clearly and effectively through graphical means. We developed an industry landscape map to help tobacco regulatory scientists policymakers understand a high-level overview of the US industry. This kind mapping market data deep companies their products benefits science public health policy in supporting potential knowledge gaps regulated

Road detection is an important problem with application to driver assistance systems and autonomous, self-guided vehicles. The focus of this paper on the feature extraction classification for front-view road detection. Specifically, we propose using Support Vector Machines (SVM) effective approach self-supervised online learning. proposed algorithm capable automatically updating training data which reduces possibility misclassifying non-road classes improves adaptability algorithm. presented...

Driver model adaptation (DMA) provides a way to the target driver when sufficient data are not available. Traditional DMA methods running at level restricted by specific structures and cannot make full use of historical data. In this paper, novel framework based on transfer learning (TL) is proposed deal with models in lane-changing scenarios level. Under framework, new TL approach named DTW-LPA that combines dynamic time warping (DTW) local Procrustes analysis (LPA) developed. Using DTW,...

Effectively predicting interactive behaviors of traffic participants in the urban road is key to successful decision-making and motion planning intelligent vehicles. In this article, based on data collected from vehicle on-board sensors, a graph-neural-network-based multitask learning framework (GNN-MTLF) proposed accurately predict trajectories with behaviors. The behavior considered research includes events that are modeled as spatial-temporal graphs using GNN. Under GNN-MTLF, prediction...

Due to the complex and dynamic character of intersection scenarios, autonomous driving strategy at intersections has been a difficult problem hot point in research intelligent transportation systems recent years. This paper gives brief summary state-of-the-art strategies intersections. Firstly, we enumerate analyze common types corresponding simulation platforms, as well related datasets. Secondly, by reviewing previous studies, have summarized characteristics existing classified them into...

Modelling, predicting and analysing driver behaviours are essential to advanced assistance systems (ADAS) the comprehensive understanding of complex driving scenarios. Recently, with development deep learning (DL), numerous behaviour (DBL) methods have been proposed applied in connected vehicles (CV) intelligent transportation (ITS). This study provides a review DBL, which mainly focuses on typical applications CV ITS. First, state-of-the-art DBL is presented. Next, Given constantly changing...

Simultaneous localization and mapping (SLAM), as an important tool for vehicle positioning mapping, plays role in the unmanned technology. This paper mainly presents a new solution to LIDAR-based SLAM vehicles off-road environment. Many methods have been proposed solve problems well. However, complex environment, especially it is difficult obtain stable results due rough road scene diversity. We propose algorithm based on grid which combining probability feature by Expectation-maximization...

Learning-based methods have gained increasing attention in the intelligent vehicle community for developing highly autonomous vehicles and advanced driving assistance systems (ADAS). However, traditional offline learning lack ability to adapt individual behavior. To overcome this limitation, a combined framework (CLF) based on Natural Actor Critic (NAC) general regression neural network (GRNN) is developed paper. GRNN can be trained historical data, while NAC carried out online. In way,...

In recent years, road safety has attracted significant attention from researchers and practitioners in the intelligent vehicle domain. As one of most common vulnerable groups users, pedestrians cause great concerns due to their unpredictable behaviour movement, as subtle misunderstandings vehicle-pedestrian interaction can easily lead risky situations or collisions. Existing methods are usually limited by poor generalization ability across scenarios high demand on human calibrations. This...

Considering the difficulty and high cost of collecting sufficient data in real world, driving simulators are used many studies as an alternative source, which can provide a much easier safer way to collect data. However, because inherent differences between virtual recognition model for behavior trained using simulation-based cannot fit scenes well. To fill gap simulation data, knowledge transfer framework is proposed this paper. Two learning (TL) methods namely semi-supervised manifold...

As the main component of an autonomous driving system, motion planner plays essential role for safe and efficient driving. However, traditional planners cannot make full use on-board sensing information lack ability to efficiently adapt different scenes behaviors drivers. To overcome this limitation, a personalized behavior learning system (PBLS) is proposed in paper improve performance planner. This based on neural reinforcement (NRL) technique, which can learn from human drivers online...

Due to advantages of handling problems with nonlinearity and uncertainty, Gaussian process regression (GPR) has been widely used in the area driver behaviour modelling. However, traditional GPR lacks ability transferring knowledge from one another, which limits generalisation GPR, especially when sufficient data for modelling are not available. To solve this limitation, paper, a novel model, Importance Weighted Process Regression (IWGPR) is proposed. The importance weight (IW) represents...

The path planning of unmanned differential steering vehicles (UDSVs) in the off-road environment not only needs to consider non-complete constraints but also faces challenges complex terrains and obstacles. In this paper, an integrated system is proposed handle influence kinematic vehicle model, obstacles systematically for UDSVs. To improve efficiency, a Pre-planning designed carried out using Voronoi diagram established 3D with By combining potential field functions (PFF) related passable...

Accurately recognizing braking intensity levels (BIL) of drivers is important for guaranteeing the safety and avoiding traffic accidents in intelligent transportation systems. In this article, an instance-level transfer learning framework proposed to recognize BIL a new driver with insufficient driving data by combining Gaussian mixture model (GMM) importance-weighted least-squares probabilistic classifier (IWLSPC). By considering statistic distribution, GMM applied cluster behaviors into...

Abstract As intelligent vehicles usually have complex overtaking process, a safe and efficient automated system (AOS) is vital to avoid accidents caused by wrong operation of drivers. Existing AOSs rarely consider longitudinal reactions the overtaken vehicle (OV) during overtaking. This paper proposed novel AOS based on hierarchical reinforcement learning, where reaction given data-driven social preference estimation. incorporates two modules that can function in different phases. The first...

As a core part of an autonomous driving system, motion planning plays important role in safe driving. However, traditional model- and rule-based methods lack the ability to learn interactively with environment, learning-based still have problems terms reliability. To overcome these problems, hybrid framework (HMPF) is proposed improve performance planning, which composed behavior optimization-based trajectory planning. The module adopts deep reinforcement learning (DRL) algorithm, can from...

Nowadays, vision oriented intelligent vehicle systems such as autonomous driving or transportation assistance can be optimized by enhancing the visual visibility of images acquired in bad weather conditions. The presence haze scenes is a critical threat. Image dehazing aims to restore spatial details from hazy images. There have emerged number image algorithms, designed increase those However, much less work has been focused on evaluating quality dehazed In this paper, we propose...

Gene regulatory networks model regulation in living organisms. Fuzzy logic can effectively gene and interaction to accurately reflect the underlying biology. A new multiscale fuzzy clustering method allows genes interact between pathways across different conditions at levels of detail. cluster centers be used quickly discover causal relationships groups coregulated genes. measures weight expert knowledge help quantify uncertainty about functions using annotations ontology database confirm...