CONNECTED AND AUTONOMOUS VEHICLES IN THE DEEP LEARNING ERA: A CASE STUDY ON COMPUTER-GUIDED STEERING
CONNECTED AND AUTONOMOUS VEHICLES IN THE DEEP LEARNING ERA: A CASE STUDY ON COMPUTER-GUIDED STEERING
复制标题
深度学习时代的联网自动驾驶汽车：计算机引导转向案例研究
DOI：
10.1142/9789811211072_0019
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
S. Ozer
中科院分区：
文献类型：
--
作者：
Rodolfo Valiente;Mahdi Zaman;Y. P. Fallah;S. Ozer
联网自动驾驶汽车 (CAV) 通常配备多个先进的车载传感器，可生成大量数据。利用和处理此类数据来提高 CAV 的性能是当前的研究领域。机器学习技术是在许多应用中利用此类数据的有效方法，并有许多成功案例。在本章中，首先，我们概述了机器学习在 CAV 新兴领域（包括特定应用）中应用的最新进展，并重点介绍了该领域的几个悬而未决的问题。其次，作为案例研究和特定应用，我们提出了一种新颖的深度学习方法来控制能够集成本地和远程信息的协作自动驾驶汽车的转向角度。在该应用中，我们通过考虑图像帧之间的时间依赖性来解决利用两辆自动驾驶车辆之间共享的多组图像来提高控制转向角的准确性的问题。这个问题尚未在文献中得到广泛研究。我们提出并研究了一种新的深度架构来自动预测转向角。我们的深层架构是一个端到端网络，利用卷积神经网络（CNN）、长短期记忆（LSTM）和全连接（FC）层；它处理当前和未来的图像（由前方车辆通过车对车 (V2V) 通信共享）作为控制转向角度的输入。在我们的模拟中，我们证明了结合使用感知和通信系统可以提高 CAV 的稳健性和安全性。与文献中的其他现有方法相比，我们的模型显示出最低的误差。
Connected and Autonomous Vehicles(CAVs) are typically equipped with multiple advanced on-board sensors generating a massive amount of data. Utilizing and processing such data to improve the performance of CAVs is a current research area. Machine learning techniques are effective ways of exploiting such data in many applications with many demonstrated success stories. In this chapter, first, we provide an overview of recent advances in applying machine learning in the emerging area of CAVs including particular applications and highlight several open issues in the area. Second, as a case study and a particular application, we present a novel deep learning approach to control the steering angle for cooperative self-driving cars capable of integrating both local and remote information. In that application, we tackle the problem of utilizing multiple sets of images shared between two autonomous vehicles to improve the accuracy of controlling the steering angle by considering the temporal dependencies between the image frames. This problem has not been studied in the literature widely. We present and study a new deep architecture to predict the steering angle automatically. Our deep architecture is an end-to-end network that utilizes Convolutional-Neural- Networks (CNN), Long-Short-Term-Memory (LSTM) and fully connected (FC) layers; it processes both present and future images (shared by a vehicle ahead via Vehicle-to-Vehicle (V2V) communication) as input to control the steering angle. In our simulations, we demonstrate that using a combination of perception and communication systems can improve robustness and safety of CAVs. Our model demonstrates the lowest error when compared to the other existing approaches in the literature.