Object detection is a significant downstream task in computer vision. For the on-board edge computing platforms, a giant model is difficult to achieve the real-time detection requirement. And, a lightweight model built from a large number of the depth-wise separable convolution layers cannot achieve the sufficient accuracy. We introduce a new lightweight convolution technique, GSConv, to lighten the model but maintain the accuracy. The GSConv accomplishes an excellent trade-off between the model's accuracy and speed. And, we provide a design paradigm, slim-neck, to achieve a higher computational cost-effectiveness of the detectors. The effectiveness of our approach was robustly demonstrated in over twenty sets comparative experiments. In particular, the detectors of ameliorated by our approach obtains state-of-the-art results (e.g. 70.9% mAP0.5 for the SODA10M at a speed of ~ 100FPS on a Tesla T4 GPU) compared with the originals. Code is available at https://github.com/alanli1997/slim-neck-by-gsconv

Monocular camera and Lidar are the two most commonly used sensors in unmanned vehicles. Combining the advantages of the two is the current research focus of SLAM and semantic analysis. In this paper, we propose an improved SLAM and semantic reconstruction method based on the fusion of Lidar and monocular vision. We fuse the semantic image with the low-resolution 3D Lidar point clouds and generate dense semantic depth maps. Through visual odometry, ORB feature points with depth information are selected to improve positioning accuracy. Our method uses parallel threads to aggregate 3D semantic point clouds while positioning the unmanned vehicle. Experiments are conducted on the public CityScapes and KITTI Visual Odometry datasets, and the results show that compared with the ORB-SLAM2 and DynaSLAM, our positioning error is approximately reduced by 87%; compared with the DEMO and DVL-SLAM, our positioning accuracy improves in most sequences. Our 3D reconstruction quality is better than DynSLAM and contains semantic information. The proposed method has engineering application value in the unmanned vehicles field.

This paper focuses on the negative effect of in-wheel switched reluctance motor (SRM) with inclined airgap eccentricity for the handling stability of electric vehicle (EV). Firstly, the radial electromagnetic force of the SRM under inclined airgap eccentricity is deduced based on the Maxwell stress tensor and airgap permeability correction coefficient, and the calculation results are verified by the built measurement device. Then, a dynamics model of in-wheel motor driving EV is constructed according to the response relationship of driving wheels. Finally, the influence of unbalanced radial torque (URT) on the negative dynamic effect of EV is analyzed with angular step input and sine input respectively. The simulation results show that the EV will deviate from the expected trajectory due to the impact of URT, which will threaten driving safety in vehicle steering or overtaking. In addition, the vehicle yaw velocity will deteriorate with the increase of the driving speed and turning angle, which will affect the dynamics response of vehicle handling stability. This research, starting from the coupling relationship between in-wheel motor response characteristics and vehicle driving conditions, can provide theoretical support for the popularization and application for in-wheel motor driving EVs.

For investigating driver characteristic as well as control authority allocation during the process of human–vehicle shared control (HVSC) for an autonomous vehicle (AV), a HVSC dynamic mode with a driver’s neuromuscular (NMS) state parameters was proposed in this paper. It takes into account the driver’s NMS characteristics such as stretch reflection and reflex stiffness. By designing a model predictive control (MPC) controller, the vehicle’s state feedback and driver’s state are incorporated to construct the HVSC dynamic model. For the validation of the model, a field experiment was conducted. The vehicle state signals are collected by V-BOX, and the driver’s state signals are obtained with the electromyography instrument. Subsequently, the hierarchical least square (HLS) parameter identification algorithm was implemented to identify the parameters of the model based on the experimental results. Moreover, the Unscented Kalman Filter (UKF) was utilized to estimate the important NMS parameters which cannot be measured directly. The experimental results showed that the model we proposed has excellent accuracy in characterizing the vehicle’s dynamic state and estimating the driver’s NMS parameter. This paper will serve as a theoretical basis for the new control strategy allocation between human and vehicle for L3 class AVs.