Search

Positioning and mapping technology is a difficult and hot topic in autonomous driving environment sensing systems. In a complex traffic environment, the signal of the Global Navigation Satellite System (GNSS) will be blocked, leading to inaccurate vehicle positioning. To ensure the security of automatic electric campus vehicles, this study is based on the Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain (LEGO-LOAM) algorithm with a monocular vision system added. An algorithm framework based on Lidar-IMU-Camera (Lidar means light detection and ranging) fusion was proposed. A lightweight monocular vision odometer model was used, and the LEGO-LOAM system was employed to initialize monocular vision. The visual odometer information was taken as the initial value of the laser odometer. At the back-end opti9mization phase error state, the Kalman filtering fusion algorithm was employed to fuse the visual odometer and LEGO-LOAM system for positioning. The visual word bag model was applied to perform loopback detection. Taking the test results into account, the laser radar loopback detection was further optimized, reducing the accumulated positioning error. The real car experiment results showed that our algorithm could improve the mapping quality and positioning accuracy in the campus environment. The Lidar-IMU-Camera algorithm framework was verified on the Hong Kong city dataset UrbanNav. Compared with the LEGO-LOAM algorithm, the results show that the proposed algorithm can effectively reduce map drift, improve map resolution, and output more accurate driving trajectory information.

Nowadays, most trajectory prediction algorithms have difficulty simulating actual traffic behavior, and there is still a problem of large prediction errors. Therefore, this paper proposes a multi-object trajectory prediction algorithm based on lane information and foresight information. A Hybrid Dilated Convolution module based on the Channel Attention mechanism (CA-HDC) is developed to extract features, which improves the lane feature extraction in complicated environments and solves the problem of poor robustness of the traditional PINet. A lane information fusion module and a trajectory adjustment module based on the foresight information are developed. A socially acceptable trajectory with Generative Adversarial Networks (S-GAN) is developed to reduce the error of the trajectory prediction algorithm. The lane detection accuracy in special scenarios such as crowded, shadow, arrow, crossroad, and night are improved on the CULane dataset. The average F1-measure of the proposed lane detection has been increased by 4.1% compared to the original PINet. The trajectory prediction test based on D2-City indicates that the average displacement error of the proposed trajectory prediction algorithm is reduced by 4.27%, and the final displacement error is reduced by 7.53%. The proposed algorithm can achieve good results in lane detection and multi-object trajectory prediction tasks.

For nonlinear systems, both the cubature Kalman filter (CKF) and square-root cubature Kalman filter (SCKF) can get good estimation performance under Gaussian noise. However, the actual driving environment noise mostly has non-Gaussian properties, leading to a significant reduction in robustness and accuracy for distributed vehicle state estimation. To address such problems, this paper uses the square-root cubature Kalman filter with the maximum correlation entropy criterion (MCSRCKF), establishing a seven degrees of freedom (7-DOF) nonlinear distributed vehicle dynamics model for accurately estimating longitudinal vehicle speed, lateral vehicle speed, yaw rate, and wheel rotation angular velocity using low-cost sensor signals. The co-simulation verification is verified by the CarSim/Simulink platform under double-lane change and serpentine conditions. Experimental results show that the MCSRCKF has high accuracy and enhanced robustness for distributed drive vehicle state estimation problems in real non-Gaussian noise environments.

As the final stage of intelligent vehicle, traffic accidents can be effectively reduced by automatic driving. However, neither the technology nor the regulations are mature for autonomous driving. The lane-keeping assist system is one of the important components of the advanced driver-assistance system. When driver fatigue or inattention is detected, the system can effectively prevent the vehicle departure from the lane. Information such as vehicle status, driver status, and external environment can be used by the lane-keeping assist system based on human–machine dynamic cooperative control, thereby smoothly changing the driving rights between the driver and the automatic controller. The system can keep the vehicle in the lane while complying with the driver's intention, thereby ensuring vehicle safety and driver comfort. The research status and future development suggestions on lane-departure decision models, dynamic allocation of driving rights, and performance evaluation were analyzed in this paper. Regarding lane-departure decision models, different decision models considering the driver's state should be developed. The decision model can be established as an adaptive adjustment model and also should allow the manual adjustment of the preset parameters according to the driver’s preferences and the external driving environment. Concerning the allocation of driving rights, a more reasonable dynamic allocation of driving rights should be explored, and intelligent optimization algorithms or control models should be designed. Regarding performance evaluation indicators, evaluation indicators related to the reduction of human–machine conflict and the amount of control effort should be added. A scientific and complete subjective evaluation system should be developed. Future studies on lane-keeping assist system based on human–machine cooperative control should deeply integrate driver factors, issue real-time warnings and active intervention, and perform complete testing and evaluation of the system.

Vehicle detection is one of the most challenging research works on environment perception for intelligent vehicle. The commonly used object detection network is too large and can only be realized in real-time on a high-performance server. Based on YOLOv3-tiny, the feature extraction was realized using light-weighted networks such as DarkNet-19 and ResNet-18 to improve accuracy. The K -means algorithm was used to cluster nine anchor boxes to achieve multi-scale prediction, especially for small targets. For automotive applicable scenarios, the proposed vehicle detection network was executed in an embedded device. The KITTI data sets were trained and tested. Experimental results show that the average accuracy is improved by 14.09% compared with the traditional YOLOv3-tiny, reaching 93.66%, and can reach 13 fps on the embedded device.

With the development of deep learning, the performance of object detection has made great progress. However, there are still some challenging problems, such as the detection accuracy of small objects and the efficiency of the detector. This study proposes an accurate and fast single shot multibox detector, which includes context comprehensive enhancement (CCE) module and feature enhancement module (FEM). To integrate more efficient information when aggregating context information, the conv4_3 and fc_7 feature maps are merged to design the CCE module. To obtain more fine‐grained feature information, this study presents a FEM and special feature enhancement module (FEM‐s) module that can fuse different receptive field sizes to better adapt to the scale change of the object. Compared to existing methods based on deep learning, the proposed method helps to gradually produce more detailed feature maps with better performance. Under the premise of ensuring real‐time speed, the authors network can achieve 81.2 mean average precision on the PASCAL VOC 2007 test with an input size of 320 × 320 on a single Nvidia 2080Ti GPU.

With the rapid development of intelligent transportation system, the research on vehicle stability can be a theoretical basis for realizing autonomous driving technology. The previous stability control strategies have not taken into account the tire force saturation factor, the slip rate, and the robustness of the control system sufficiency. According to the characteristic that the torque of each wheel can be distributed independently, a torque distribution algorithm under emergency conditions is proposed. The proposed torque distribution algorithm is constructed using three hierarchical controllers. The upper controller attempts to judge whether the vehicle is in stable state using the phase plane method. Also, it judges whether the wheels are slipping. The middle controller aims to calculate the demands for the desired traction force and yaw moment, whereas the lower controller is designed to translate those virtual signals into actual actuator commands. When designing the middle controller, a sliding mode control method is utilized to guarantee system stability and robustness by taking into account various factors, including lateral wind, and sensor noise. For the lower controller, the control allocation optimization method is utilized to determine an appropriate control input for each in-wheel motor by considering the road conditions, adhesion utilization, and maximum output torque of the motor. The numerical simulation studies are conducted to evaluate the performance of the torque distribution algorithm. Comparison results indicate that the proposed algorithm presents better performance to distribute the appropriate torque for each wheel and ensure the stability of the vehicle under emergency conditions.

A novel pedestrian detection system based on vision in urban traffic situations is presented to help the driver perceive the pedestrian ahead of the vehicle. To enhance the accuracy and to decrease the time spent on pedestrian detection in such complicated situations, the pedestrian is detected by dividing their body into several parts according to their corresponding features in the image. The candidate pedestrian leg is segmented based on the gentle AdaBoost algorithm by training the optimized histogram of gradient features. The candidate pedestrian head is located by matching the pedestrian head and shoulder model above the region of the candidate leg. Then the candidate leg, head and shoulder are combined by parts constraint and threshold adjustment to verify the existence of the pedestrian. Finally, the experiments in real urban traffic circumstances were conducted. The results show that the proposed pedestrian detection method can achieve pedestrian detection rate of 92.1% with the average detection time of 0.2257 s.

In this paper, with the aim of meeting the requirements of car following, safety, comfort, and economy for adaptive cruise control (ACC) system, an ACC algorithm based on model predictive control (MPC) using constraints softening is proposed. A higher-order kinematics model is established based on the mutual longitudinal kinematics between the host vehicle and the preceding vehicle that considers the changing characteristics of the inter-distance, relative velocity, acceleration, and jerk of the host vehicle. Performance indexes are adopted to represent the multi-objective demands and constraints of the ACC system. To avoid the solution becoming unfeasible because of the overlarge feedback correction, the constraint softening method was introduced to improve robustness. Finally, the proposed ACC method is verified in typical car-following scenarios. Through comparisons and case studies, the proposed method can improve the robustness and control precision of the ACC system, while satisfying the demands of safety, comfort, and economy.

Pedestrian detection and tracking is the key to autonomous vehicle navigation systems avoiding potentially dangerous situations. Firstly, the probability distribution of colour information is established after a pedestrian is located in an image. Then the detected results are utilized to initialize a Kalman filter to predict the possible position of the pedestrian centroid in the future frame. A Camshift tracking algorithm is used to track the pedestrian in the specific search window of the next frame based on the prediction results. The actual position of the pedestrian centroid is output from the Camshift tracking algorithm to update the gain and error covariance matrix of the Kalman filter. Experimental results in real traffic situations show the proposed pedestrian tracking algorithm can achieve good performance even when they are partly occluded in inconsistent illumination circumstances.

In order to detect the obstacle from the large amount of 3D LIDAR data in hybrid cross-country environment for unmanned ground vehicle, a new graph approach based on Markov random field was presented. Firstly, the preprocessing method based on the maximum blurred line is applied to segment the projection of every laser scan line in x-y plane. Then, based on K-means clustering algorithm, the same properties of the line are combined. Secondly, line segment nodes are precisely positioned by using corner detection method, and the next step is to take advantage of line segment nodes to build an undirected graph for Markov random field. Lastly, the energy function is calculated by means of analyzing line segment features and solved by graph cut. Two types of line mark are finally classified into two categories: ground and obstacle. Experiments prove the feasibility of the approach and show that it has better performance and runs in real time.

Abstract Visual odometry provides planetary exploration rovers with accurate knowledge of their position and orientation, which needs effective feature tracking results, especially in barren sandy terrains. In this paper, a stereovision based odometry algorithm is proposed for a lunar rover, which is composed of corner extraction, feature tracking and motion estimation. First, a morphology based image enhancement method is studied to guarantee enough corners are extracted. Second, a Random Sample Consensus (RANSAC) algorithm is proposed to make a robust estimation of the fundamental matrix, which is the basic and critical part of feature matching and tracking. Then, the 6 degrees of freedom rover position and orientation is estimated by the RANSAC algorithm. Finally, experiments are performed in a simulated lunar surface environment using a prototype rover, which have confirmed the feasibility and effectiveness of the proposed method.

Lane departure warning system (LDWS) has been regarded as an efficient method to lessen the damages of road traffic accident resulting from driver fatigue or inattention. Lane detection is one of the key techniques for LDWS. To overcome the contradiction between complexity of algorithm and the real-time requirement for vehicle onboard system, this paper introduces a new lane detection method based on intelligent CCD parameters regulation. In order to improve the real-time capability of the system, a CCD parameters regulating method is proposed which enhances the contrast between lane line and road surfaces and reduces image noise, so it lays a good foundation for the following lane detection. Hough transform algorithm is improved by selection and classification of seed points. Finally the lane line is extracted through some restrictions. Experimental results verify the effectiveness of the proposed method, which improves not only real-time capability but also the accuracy of the system.