面向自动协同驾驶的多车编队任务分配策略

自动驾驶的实现需要大量车载传感器的支持,然而,在有限车载计算资源条件下,由传感器所产生的庞大数据量使得自动驾驶任务的实时性难以满足,成为阻碍自动驾驶技术进一步发展的重要阻力。通过将驾驶任务进行协作处理,因而充分利用多个协作车辆的计算资源,自动协同驾驶成为解决该问题的新途径。而如何形成多车编队并实现编队中驾驶任务分配则是实现自动协同驾驶的关键。该文首先采用排队理论G/G/1模型建立一种普适性车辆编队网络拓扑分析模型,充分考虑编队内车辆间的任务协作能力和单个车辆的任务负荷,得出任务的处理时延和车辆系统中的平均任务数;其次,采用支持向量机(SVM)方法,基于车辆的负荷程度及处理能力将车辆的“空闲”、“繁忙”两状态进行分类,进而建立针对车辆协作任务分配的候选车辆集。最后,基于上述分析,该文提出面向多车编队协同驾驶的任务均衡策略——基于分类的贪婪均衡策略(C-GBS),以充分平衡编队内所有车辆的任务负荷并利用不同车辆的任务处理能力。仿真结果表明,该策略能够减小重负荷网络中的任务处理时延,有效提升自动驾驶车辆的任务处理效率。     

Real-time vision-based tracking control of an unmanned vehicle

Recent advances in manufacturing automation have motivated vision-based control of autonomous vehicles operated in unattended factories, material handling processes, warehouse operations, and hazardous environment explorations. Existing vision-based tracking control systems of autonomous vehicles, however, have been limited in real-time applications due to slow and/or expensive visual feedback and complicated dynamics and control with nonholonomic constraints. This paper presents a practical real-time sion-based cking control system of an unmanned vehicle, ViTra. Unlike the conventional RS170 video-based machine vision systems, ViTra uses a DSP-based flexible integrated vision system (FNS) which is characterized by low cost, computational efficiency, control flexibility, and a friendly user interface.

In particular, this paper focuses on developing a framework for vision tracking systems, designing generic fiducial patterns, and applying real-time vision systems to tracking control of autonomous vehicles. A laboratory prototype vision-based tracking system developed at Georgia Institute of Technology permits the uniquely designed fiducial landmarks to be evaluated experimentally, the control strategy and the path planning algorithm derived in the paper to be validated in real-time, and the issues of simplifying nonlinear dynamics and dealing with nonholonomic constraints to be addressed in practice. Experimental results reveal interesting insights into the design, manufacture, modeling, and control of vision-based tracking control systems of autonomous vehicles.     

Generating routes for autonomous driving in vehicle-to-infrastructure communications

The study of vehicular networks has attracted considerable interest in academia and the industry. In the broad area, connected vehicles and autonomous driving are technologies based on wireless data communication between vehicles or between vehicles and infrastructures. A Vehicle-to-Infrastructure (V2I) system consists of communications and computing over vehicles and related infrastructures. In such a system, wireless sensors are installed in some selected points along roads or driving areas. In autonomous driving, it is crucial for a vehicle to figure out the ideal routes by the communications between its equipped sensors and infrastructures then the vehicle is automatically moving along the routes. In this paper, we propose a Bezier curve based recursive algorithm, which effectively creates routes for vehicles through the communication between the On-Board Unit (OBU) and the Road-Side Units (RSUs). In addition, this approach generates a very low overhead. We conduct simulations to test the proposed algorithm in various situations. The experiment results demonstrate that our algorithm creates almost ideal routes.     

基于C-V2X的车路协同自动驾驶关键技术与应用

随着车联网技术的演进,自动驾驶在单车智能的基础上,又有了新的发展形态——车路协同自动驾驶。通过“人-车-路-云”深度融合形成的一体化复杂信息物理系统（cyber physical system,CPS）,可以与自动驾驶车辆实现协同感知、协同决策规划甚至协同控制,提升自动驾驶安全性,帮助克服各类复杂交通环境。首先介绍了车路协同的复杂信息物理系统的概念内涵和总体架构,并提出了车路协同自动驾驶的一系列典型应用场景、技术原理、C-V2X(cellular vehicle-to-everything)性能要求、车路协同系统功能与性能要求,可以为下一阶段智能网联汽车与智能交通的深度融合发展提供参考和解决思路。     

Development of Steering Control System for Autonomous Vehicle Using Geometry‐Based Path Tracking Algorithm

In this paper, a steering control system for the path tracking of autonomous vehicles is described. The steering control system consists of a path tracker and primitive driver. The path tracker generates the desired steering angle by using the look‐ahead distance, vehicle heading, and a lateral offset. A method for applying an autonomous vehicle to path tracking is an advanced pure pursuit method that can reduce cutting corners, which is a weakness of the pure pursuit method. The steering controller controls the steering actuator to follow the desired steering angle. A servo motor is installed to control the steering handle, and it can transmit the steering force using a belt and pulley. We designed a steering controller that is applied to a proportional integral differential controller. However, because of a dead band, the path tracking performance and stability of autonomous vehicles are reduced. To overcome the dead band, a dead band compensator was developed. As a result of the compensator, the path tracking performance and stability are improved.     

Study on the Take-over Performance of Level 3 Autonomous Vehicles Based on Subjective Driving Tendency Questionnaires and Machine Learning Methods

Level 3 autonomous vehicles require conditional autonomous driving in which autonomous and manual driving are alternately performed; whether the driver can resume manual driving within a limited time should be examined. This study investigates whether the demographics and subjective driving tendencies of drivers affect the take-over performance. We measured and analyzed the reengagement and stabilization time after a take-over request from the autonomous driving system to manual driving using a vehicle simulator that supports the driver's take-over mechanism. We discovered that the driver's reengagement and stabilization time correlated with the speeding and wild driving tendency as well as driving workload questionnaires. To verify the efficiency of subjective questionnaire information, we tested whether the driver with slow or fast reengagement and stabilization time can be detected based on machine learning techniques and obtained results. We expect to apply these results to training programs for autonomous vehicles' users and personalized human–vehicle interfaces for future autonomous vehicles.     

Parking maneuvers of industrial-like electrical vehicles with and without trailer

Maneuvering autonomous vehicles in constrained environments is not a trivial task. This paper concentrates in the practical maneuvering of electrical vehicles. The autonomous vehicles considered in the paper, ROMEO-3R and ROMEO-4R, have been developed at the University of Seville, Seville, Spain, as the result of the adaptation of a tricycle and a car-like conventional electrical vehicles for transportation of people, respectively. Moreover, maneuvering of ROMEO-4R backing up a trailer has also been considered. A particular maneuver, namely, autonomous parallel parking, has been used to illustrate the application of the presented methods to different electrical vehicles.     

Autonomous vehicles

There are various kinds of autonomous vehicles (AV's) which can operate with varying levels of autonomy. This paper is concerned with underwater, ground and aerial vehicles operating in a fully autonomous (nonteleoperated) mode. Further this paper deals with AV's as a special kind of device, rather than full-scale manned vehicles operating unmanned. The distinction is one in which the AV is likely to be designed for autonomous operation rather than being adapted for it as would be the case for manned vehicles. We provide a survey of the technological progress that has been made in AV's, the current research issues and approaches that are continuing that progress, and the applications which motivate this work. It should be noted that issues of control are pervasive regardless of the kind of AV being considered, but that there are special considerations in the design and operation of AV's depending on whether the focus is on vehicles underwater on the ground, or in the air. We have separated the discussion into sections treating each of these categories     

DiLO: Direct light detection and ranging odometry based on spherical range images for autonomous driving

Over the last few years, autonomous vehicles have progressed very rapidly. The odometry technique that estimates displacement from consecutive sensor inputs is an essential technique for autonomous driving. In this article, we propose a fast, robust, and accurate odometry technique. The proposed technique is light detection and ranging (LiDAR)-based direct odometry, which uses a spherical range image (SRI) that projects a three-dimensional point cloud onto a two-dimensional spherical image plane. Direct odometry is developed in a vision-based method, and a fast execution speed can be expected. However, applying LiDAR data is difficult because of the sparsity. To solve this problem, we propose an SRI generation method and mathematical analysis, two key point sampling methods using SRI to increase precision and robustness, and a fast optimization method. The proposed technique was tested with the KITTI dataset and real environments. Evaluation results yielded a translation error of 0.69%, a rotation error of 0.0031°/m in the KITTI training dataset, and an execution time of 17 ms. The results demonstrated high precision comparable with state-of-the-art and remarkably higher speed than conventional techniques.     

A Global Optimal Path Planning and Controller Design Algorithm for Intelligent Vehicles

Autonomous vehicle guidance and trajectory planning is one of the key technologies in the autonomous control system for intelligent vehicles. Firstly, the target pursuit model for intelligent vehicles was established and described in this text. Then, the research work for global motion planning was carried out based on Stackelberg Differential Game Theory, and the global optimal solution was obtained by using the survival type differential game. Finally, to overcome errors, we use a polynomial method to achieve the smooth motion planning. So, based on Terminal Sliding Mode method, the Active Front Steering controller design was used to calculate the desired active wheel angle for intelligent vehicle path tracking. The simulation and experiment results demonstrate the feasibility and effectiveness of this method for intelligent vehicles’ path planning and tracking.     

面向自动驾驶的边缘计算技术研究综述

边缘计算在自动驾驶的环境感知和数据处理方面有着极其重要的应用。自动驾驶汽车可以通过从边缘节点获得环境信息来扩大自身的感知范围,也可以向边缘节点卸载计算任务以解决计算资源不足的问题。相比于云计算,边缘计算避免了长距离数据传输所导致的高时延,能给自动驾驶车辆提供更快速的响应,并且降低了主干网络的负载。基于此,首先介绍了基于边缘计算的自动驾驶汽车协同感知和任务卸载技术及相关挑战性问题,然后对协同感知和任务卸载技术的研究现状进行了分析总结,最后讨论了该领域有待进一步研究的问题。     

Inter-vehicle communication: technical issues on vehicle controlapplication

Mutual exchange of status data between vehicles in close proximity is the basis for safe vehicle operation. Application areas range from driver assistance/warning to fully autonomous driving. To enable this data exchange we need an intervehicle communication (IVC) system. Vehicles involved in mutual data exchange form a kind of local area network; however, the characteristics of this network are quite different from those of conventional networks. The configuration of this kind of network changes from time to time, in terms of organization, proximity group, group position, and relative position; therefore, it is very challenging to establish. The authors have been working on this problem from the point of view of communication media and protocol. The characteristics and requirements of IVC are discussed, and several preliminary communication protocols are proposed. The authors also present some experimental results from the implementation of a function integrating communication by infrared ray with a gap-measuring technique by the vision system     

Control architecture and operator interface for a free-flyingrobotic vehicle

Space and underwater vehicles with robotic arms can severely tax the capability of conventional control systems. Submersible vehicles used in neutral buoyancy simulation are subject to even greater demands since they must simulate the dynamics of spacecraft in orbit as well as function as a remotely-operated underwater vehicle. In this report, the onboard control architecture, human-machine interface, and vehicle/operator communications are described for one such vehicle in operation at the University of Maryland Neutral Buoyancy Research Facility (NBRF). The Ranger Neutral Buoyancy Vehicle (RNBV) exemplifies the high-dimensional, computationally intensive nature of the current fleet of autonomous underwater vehicles while its complement of four manipulators exceeds the capabilities of most remotely operated vehicles in service today. The sensor-based, embedded onboard control system is described, and its implementation using multiple control stations is discussed     

基于视觉、LiDAR与IMU的实时无人车里程计研究

视觉/LiDAR里程计可以根据传感器数据对无人车在多个自由度上运动的过程进行估计,是无人车定位建图系统的重要组成部分。文中提出了一种使用视觉、LiDAR和IMU进行信息融合的里程计,支持多种运行模式和初始化方式。前端部分采用了改进后的ICP CUDA算法进行激光点云配准,利用光流法对视觉特征进行跟踪,并利用激光点云数据对视觉特征的深度进行估计。后端部分采用了基于滑动窗口的图优化模型,并为视觉和LiDAR关键帧创建状态节点,以前端结果作为量测,将相邻状态节点通过预积分因子关联。文中方案实验结果表明:在城市场景系统平均相对位移精度为0.2%~0.5%,系统全量传感器运行模式（VLIO模式）整体要比关闭视觉的模式（LIO模式）和关闭LiDAR的模式（VIO模式）精度高。文中提出的方法对于提高无人车定位建图系统的精度有着积极意义。     

One-layer neural-network controller with preprocessed inputs for autonomous underwater vehicles

Navigating, guiding, and controlling autonomous underwater vehicles (AUVs) are challenging and difficult tasks compared to the autonomous surface-level operations. Controlling the motion of such vehicles require the estimation of unknown hydrodynamic forces and moments and disturbances acting on these vehicles in the underwater environment. In this paper, a one-layer neural-network (NN) controller with preprocessed input signals is designed to control the vehicle track along a desired trajectory, which is specified in terms of desired position and attitude. In the absence of unknown disturbances and modeling errors, it is shown that the tracking error system is asymptotically stable. In the presence of any bounded ocean currents or wave disturbances, the uniformly ultimately boundedness of the tracking error and NN weight estimates are given. The NN does not require an initial offline training phase and weight initialization is straightforward. Simulation results are shown by using a scaled version of the Naval Post-Graduate School's AUV. Results indicate the superior performance of the NN controller over conventional controllers. Providing offline NN training may improve the transient performance.     

Systems for Safety and Autonomous Behavior in Cars: The DARPA Grand Challenge Experience

In this paper, we review technologies for autonomous ground vehicles and their present capabilities in research and in the automotive market. We outline technology requirements for enhanced functions and for infrastructure development. Since the recent Grand Challenge competition is a major force to advance technology in this field, we specifically refer to our experiences in developing a participating vehicle. We present a multisensor platform that has been proven in an off-road environment. It combines different sensing modalities that inherently yield uncertain information. Finite-state machines are formulated to generate rule-based autonomous behavior that enables fully autonomous off-road driving. Overall, the intent of the paper is to evaluate approaches and technologies used in the two Grand Challenges as they contribute to the needs of autonomous cars on the road     

无人机感知与规避技术综述

无人机广泛应用于军事侦察、打击以及民用航拍、植保等各个领域,无人机对飞行安全的影响已成为关注热点。无人机感知与规避技术可为无人机提供障碍物探测和碰撞告警能力, 是保障飞行安全的关键。本文系统地介绍了无人自主感知与规避系统的组成和基本功能,并对其未来发展进行了展望。     

汽车毫米波FMCW雷达中频信号的采样与处理

汽车毫米波雷达是智能交通系统的重要研究项目之一，本文介绍了汽车毫米波雷达的工作原理及信号处理的相关方法和技术。     

Visual Quality Assessment of Video and Image Sequences—A Human-based Approach

Most of the current quality assessment techniques interpret an image quality as a measure of its fidelity with another reference image, assuming the availability of that “perfect” image. It has been the concern of many researchers around the world to algorithmically assess the quality of image sequences based on human visual perception. This paper presents a novel technique for quantitatively assessing the quality of image sequences without the need for a reference image and in a way that precisely correlates to human judgement on quality. This research is a part of a larger framework that incorporates multi-objective optimisation algorithms to optimise the quality metrics of compressed videos acquired by autonomous vehicles and transmitted over low-bandwidth communication channels. Our system was trained on a dataset that involved 700 videos of 5 different categories. We validate the performance of our model and show that it highly correlates to the human subjective quality assessment.