Safe and Robust Autonomous Vehicle Technology

安全稳健的自动驾驶汽车技术

• 批准号: RGPIN-2017-06767
• 负责人: Najjaran, Homayoun
• 金额: $ 2.26万
• 依托单位: University of British Columbia
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2020
• 资助国家: 加拿大
• 起止时间: 2020-01-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=710562
• 关键词: Safe; Robust; Autonomous; Vehicle; Technology

The outcome of research in the global robotics and mechatronics community over the past two decades has revolutionized both the aerospace and automotive industries by providing the technologies used in unmanned aerial vehicles (UAVs) and self-driving cars. A number of automotive and aerospace manufacturers, as well as high-tech companies, have made substantial investments in their research and development programs to introduce the first flawless and fully autonomous vehicle technology and to gain a greater share of the market. However, the fate of these investments is linked to acceptance and adoption of autonomous technology by regulators and the public . The proposed research program focuses on improving the robustness of autonomous vehicle technology for safer and more reliable UAVs and self-driving cars. This goal will be achieved by investigating every component of the autonomous navigation process with specific emphasis on the following three aspects. First, innovative machine learning methods based on convolutional neural network deep learning will be developed and used to enable more efficient object detection in unknown and complex environments. In this way, a direct perception approach for robust autonomous navigation of UAVs and AVs is introduced. Second, approximate reasoning methods for intelligent high-level decision making will be developed to provide a systematic way of responding to vague states and contradictory evidence in unforeseen driving or flight scenarios. A versatile inference engine will be developed by incorporating network-based machine learning, fuzzy logic and evidential reasoning methods. Third, a stable hybrid control system will be developed that can provide fast deceleration and swerving of AVs, and high-pitch maneuvers. Using this hybrid control scheme, a proof-of-concept active fault tolerant control (AFTC) system will be developed for a quadrotor UAV. The contributions of the proposed research will be creating the potential for: i) a prototype autonomous car that is safe to operate on public roads, and ii) a commercial-grade UAV safe to fly beyond line of sight. The proposed research will contribute to the concurrent global research aiming to resolve the dilemma of AV technology by introducing innovative human-like approximate reasoning paradigms that accounts for moral, social and legal standards. In conclusion, independent, factual and transparent research originating from academia will influence the public acceptance and adoption of AV technology.

过去20年，全球机器人和机电一体化领域的研究成果为航空航天和汽车行业带来了革命性的变化，提供了用于无人机(UAV)和自动驾驶汽车的技术。许多汽车和航空航天制造商以及高科技公司已经在他们的研发计划中投入了大量资金，以引入第一个完美和完全自动驾驶的汽车技术，并获得更大的市场份额。然而，这些投资的命运与监管机构和公众对自主技术的接受和采用有关。拟议的研究计划侧重于提高自动驾驶汽车技术的稳健性，以实现更安全、更可靠的无人机和自动驾驶汽车。将通过调查自主导航进程的每一个组成部分来实现这一目标，并特别强调以下三个方面。 首先，将开发基于卷积神经网络深度学习的创新机器学习方法，并将其用于在未知和复杂环境中更有效地检测目标。通过这种方法，提出了一种用于无人机和无人机鲁棒自主导航的直接感知方法。 其次，将开发用于智能高层决策的近似推理方法，以提供一种系统地应对不可预见的驾驶或飞行场景中的模糊状态和矛盾证据的方法。通过融合基于网络的机器学习、模糊逻辑和证据推理方法，将开发出一个通用的推理机。 第三，将开发一种稳定的混合控制系统，可以提供无人机的快速减速和转向以及高俯仰机动。采用这种混合控制方案，将为四旋翼无人机开发一个概念验证的主动容错控制(AFTC)系统。 这项拟议研究的贡献将创造以下可能性：i)在公共道路上安全运行的自动驾驶汽车原型，以及ii)可以安全飞行在视线之外的商业级无人机。拟议的研究将有助于同时进行的全球研究，旨在通过引入创新的类人近似推理范式来解决反病毒技术的困境，这些范式解释了道德、社会和法律标准。总之，来自学术界的独立、事实和透明的研究将影响公众对反病毒技术的接受和采用。