Development of an enhanced simulated driver autonomous vehicle******

开发增强型模拟驾驶员自动驾驶汽车******

• 批准号: 533744-2018
• 负责人: VanderLoos, Hendrik
• 金额: $ 1.82万
• 依托单位: University of British Columbia
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=667510
• 关键词: Development; enhanced; simulated; driver; autonomous

The long-term goal of this research is to develop an autonomous vehicle (cargo trailer) that, unlike existing products, can drive on the street without requiring a driver sitting behind the wheel. The proposed work entails exploring the state of the art and arriving at a small-scale trailer hardware and software prototype. **There are major challenges toward this objective, as follows,**1. The envisioned following vehicle, FV, or cargo trailer, is an autonomous vehicle. In this project we may use an existing autonomous vehicle with some modification or develop our own. The lead vehicle, LV, or towing vehicle does not pull the FV through a conventional load-bearing physical link.**2. To satisfy the regulatory condition for autonomous vehicles that require a fulltime attending operator behind the wheel, we need to develop an alternate way to simulate a driver. For this purpose, we use the LV driver's behaviour as a resource through a non-load-bearing physical link with 7 degrees of freedom. Inverse kinematics will be used to calculate the locations of the LV and FV.**3. Reinforcement learning (RL) algorithms, along with machine learning, computer vision and image processing, are considered as effective approaches to solve complex sequential decision/control problems.

这项研究的长期目标是开发一种自动驾驶汽车（货运拖车），与现有产品不同，它可以在街道上行驶，而不需要驾驶员坐在方向盘后面。拟议的工作需要探索最先进的技术并获得小型拖车硬件和软件原型。 **实现这一目标面临以下主要挑战，**1。 设想的后续车辆 FV 或货运拖车是自动驾驶车辆。在这个项目中，我们可能会使用现有的自动驾驶汽车进行一些修改或开发我们自己的自动驾驶汽车。领头车辆、LV 或牵引车不会通过传统的承载物理链路拉动 FV。**2。 为了满足自动驾驶汽车需要全职操作员驾驶的监管条件，我们需要开发一种替代方法来模拟驾驶员。为此，我们通过 7 个自由度的非承载物理链路，将 LV 驱动器的行为作为资源。逆运动学将用于计算 LV 和 FV 的位置。**3。 强化学习（RL）算法以及机器学习、计算机视觉和图像处理被认为是解决复杂的顺序决策/控制问题的有效方法。