Kinematic model predictive control (MPC) is well known for its simplicity and computational efficiency for path tracking of autonomous vehicles, however, it merely works well at low speed. In addition, earlier studies have demonstrated that tracking accuracy is improved by the feedback of yaw rate, as it improves the system transients. With this in mind, it is expected that the performance of path tracking can be improved by a cascaded controller that utilizes kinematic MPC to determine desired yaw rate rather than steering angle, and uses proportional-integral-derivative (PID) control to follow the reference yaw rate. However, directly combining MPC with PID feedback control of yaw rate results in a controller with poor tracking accuracy. The simulation results show that the cascaded MPC-PID controller has relatively stable but larger error compared to classic kinematic and dynamic MPC. Based on the analysis of vehicle sideslip angle, a novel path tracking control method is proposed, which is designed using a kinematic MPC to handle the disturbances on road curvature, a PID feedback control of yaw rate to reject uncertainties and modeling errors, and a vehicle sideslip angle compensator to correct the kinematic model prediction. The proposed controller performances involving steady-state and transient response, robustness, and computing efficiency were evaluated on Carsim/Matlab joint simulation environment. Furthermore, field experiments were conducted to validate the robustness against sensor disturbances and time lag. The results demonstrate that the developed vehicle sideslip compensator is sufficient to capture steer dynamics, and the developed controller significantly improves the performance of path tracking and follows the desired path very well, ranging from low speed to high speed even at the limits of handling.

For autonomous vehicles, it is an important requirement to obtain integrate static road information in real-time in dynamic driving environment. A comprehensive perception of the surrounding road should cover the accurate detection of the entire road area despite occlusion, the 3D geometry and the types of road topology in order to facilitate the practical applications in autonomous driving. To this end, we propose a lightweight and efficient LiDAR-based multi-task road perception network (LMRoadNet) to conduct occlusion-free road segmentation, road ground height estimation, and road topology recognition simultaneously. To optimize the proposed network, a corresponding multi-task dataset, named MultiRoad, is built semi-automatically based on the public SemanticKITTI dataset. Specifically, our network architecture uses road segmentation as the main task, and the remaining two tasks are directly decoded on a concentrated 1 / 4 scale feature map derived from the main task's feature maps of different scales and phases, which significantly reduces the complexity of the overall network while achieves high performance. In addition, a loss function with learnable weight of each task is adopted to train the neural network, which effectively balances the loss of each task and improves performance of the individual tasks. Extensive experiments on the test set show that the proposed network achieves great performance of the three tasks in real-time, outperforms the conventional multi-task architecture and is comparable to the state-of-the-art efficient methods. Finally, a fusion strategy is proposed to combine results on different directions to expand the field of view for practical applications.

With the continuous development of digital twin (DT) technology at this stage, research and applications surrounding DTs have gradually become a hot spot. The application of DT technology in the field of autonomous driving tests is being studied. In this paper, we intend to build a highly open DT autonomous driving test platform, combining it with functional units such as simulation test tools, communication equipment, real test vehicles, etc., to form a rich test verification environment. This model supports various types of autonomous driving solutions and algorithm verification tests and has the ability to carry out real vehicle testing and verification in complex virtual scenes under the condition of limited resources. Finally, this article provides a brand-new test method for autonomous driving vehicles.

Preceding vehicles have a significant impact on the safety of the vehicle, whether or not it has the same driving direction as an ego-vehicle. Reliable trajectory prediction of preceding vehicles is crucial for making safer planning. In this paper, we propose a framework for trajectory prediction of preceding target vehicles in an urban scenario using multi-sensor fusion. First, the preceding target vehicles historical trajectory is acquired using LIDAR, camera, and combined inertial navigation system fusion in the dynamic scene. Next, the Savitzky–Golay filter is taken to smooth the vehicle trajectory. Then, two transformer-based networks are built to predict preceding target vehicles’ future trajectory, which are the traditional transformer and the cluster-based transformer. In a traditional transformer, preceding target vehicles trajectories are predicted using velocities in the X-axis and Y-axis. In the cluster-based transformer, the k-means algorithm and transformer are combined to predict trajectory in a high-dimensional space based on classification. Driving data from the real-world environment in Wuhan, China, are collected to train and validate the proposed preceding target vehicles trajectory prediction algorithm in the experiments. The result of the performance analysis confirms that the proposed two transformers methods can effectively predict the trajectory using multi-sensor fusion and cluster-based transformer method can achieve better performance than the traditional transformer.

An accurate object pose is essential to assess its state and predict its movements. In recent years, scholars have often predicted object poses by matching an image with a virtual 3D model or by regressing the six-degree-of-freedom pose of the target directly from the pixel data via deep learning methods. However, these approaches may ignore a fact that was proposed in the early days of computer vision research, i.e., that object parts are strongly represented in the object pose. In this study, we propose a novel and lightweight deep learning framework, YAEN (yaw angle estimation network), for accurate object yaw angle prediction from a monocular camera based on the arrangement of parts. YAEN uses an encoding–decoding structure for vehicle yaw angle prediction. The vehicle part arrangement information is extracted by the part-encoding network, and the yaw angle is extracted from vehicle part arrangement information by the yaw angle decoding network. Because vehicle part information is refined by the encoder, the decoding network structure is lightweight; the YAEN model has low hardware requirements and can reach a detection speed of $97FPS$ on a $2070\mathrm{s}$graphics cards. To improve the performance of our model, we used asymmetric convolution and SSE (sum of squared errors) loss functions of adding the sign. To verify the effectiveness of this model, we constructed an accurate yaw angle dataset under real-world conditions with two vehicles equipped with high-precision positioning devices. Experimental results prove that our method can achieve satisfactory prediction performance in scenarios in which vehicles do not obscure each other, with an average prediction error of less than 3.1° and an accuracy of 96.45% for prediction errors of less than 10° in real driving scenarios.

The deep convolutional neural network has led the trend of vision-based road detection, however, obtaining a full road area despite the occlusion from monocular vision remains challenging due to the dynamic scenes in autonomous driving. Inferring the occluded road area requires a comprehensive understanding of the geometry and the semantics of the visible scene. To this end, we create a small but effective dataset based on the KITTI dataset named KITTI-OFRS (KITTI-occlusion-free road segmentation) dataset and propose a lightweight and efficient, fully convolutional neural network called OFRSNet (occlusion-free road segmentation network) that learns to predict occluded portions of the road in the semantic domain by looking around foreground objects and visible road layout. In particular, the global context module is used to build up the down-sampling and joint context up-sampling block in our network, which promotes the performance of the network. Moreover, a spatially-weighted cross-entropy loss is designed to significantly increases the accuracy of this task. Extensive experiments on different datasets verify the effectiveness of the proposed approach, and comparisons with current excellent methods show that the proposed method outperforms the baseline models by obtaining a better trade-off between accuracy and runtime, which makes our approach is able to be applied to autonomous vehicles in real-time.