Disrupting the UK inertial sensor supply chain to increase resilience and safety for self-driving vehicles

颠覆英国惯性传感器供应链，提高自动驾驶车辆的弹性和安全性

• 批准号: 10065170
• 金额: $ 173.64万
• 依托单位: ZERO POINT MOTION LTD
• 依托单位国家: 英国
• 项目类别: BEIS-Funded Programmes
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=10065170
• 关键词: Disrupting; UK; inertial; sensor; supply

Zero Point Motion (ZPM) develops ultra sensitive chipscale devices to measure acceleration and rate of rotation of moving objects. These inertial sensors are 100x more sensitive than existing ones inside cars today because we use light instead of a noisy electrical readout, enabling our sensors to measure smaller motions without becoming larger in size or more expensive. Our sensors can improve the performance, safety and resilience of driverless cars because inertial sensors work anywhere, anytime. Currently, autonomous navigation relies on external reference signals from satellites (e.g. GPS), LIDAR/RADAR or cameras. However, all these external references output data at 10-100x lower rates than inertial sensors, and their signals can be spoofed, jammed or denied, for example, GPS is not available underground or in tunnels and is a weak signal prone to interference, whereas LIDAR/RADAR and cameras require huge computational processing powers to stitch together images which can fail if false objects are tracked or if there are no prominent features.Autonomous vehicles cannot reach Level 5 full autonomy using a single sensor solution, but instead, a multitude of inputs that are fused together (sensor fusion) such that LIDAR/RADAR and cameras enable obstacle tracking and avoidance, satellite navigation provides an absolute position relative to the Earth, and inertial sensors measure the car dynamics. Improving the inertial sensors is therefore critical to guarantee traceable decision making, as they provide measurements and not 'black box' software or AI/machine learning interpretations of static LIDAR/RADAR/camera data.To prepare our sensors for automotive certification, we will use standard packaging practices known to satisfy regulatory testing and begin our quality assurance process, running tests defined by the AEC-Q100 certification. Our partner WAE will conduct testing on motorsport platforms, exposing our sensor to real-world extreme dynamics and conditions. To verify how the inertial sensor sits within the navigation stack and the safety or mission critical role it plays compared to external references, our partner UWE will conduct independent research on integrity and resilience using their expertise on autonomous vehicles. We will work with our partner the Royal Institute of Navigation to ensure our IMUs and UWE's research is visible to the UK CAM hardware supply chain by holding two events involving academic and industry leaders in autonomous navigation, and to encourage debate and discussion on topics around sensor fusion, safety, resilience and the UK's position. These events are opportunities to generate new insights and collaborations between interested stakeholders.

零点运动(ZPM)开发了超灵敏的芯片级设备来测量运动物体的加速度和旋转速度。这些惯性传感器的灵敏度是目前汽车内现有传感器的100倍，因为我们使用光而不是嘈杂的电子读数，使我们的传感器能够测量较小的运动，而不会变得更大或更昂贵。我们的传感器可以提高无人驾驶汽车的性能、安全性和弹性，因为惯性传感器可以随时随地工作。目前，自主导航依赖于来自卫星(例如GPS)、激光雷达/雷达或相机的外部参考信号。然而，所有这些外部参考输出数据的速率都比惯性传感器低10-100倍，它们的信号可以被欺骗、干扰或拒绝，例如，GPS在地下或隧道中不可用，并且是容易受到干扰的弱信号，而激光雷达/雷达和相机需要巨大的计算处理能力来拼接图像，如果跟踪到错误的对象或如果没有显著特征，图像可能会失败。自动驾驶车辆不能使用单一传感器解决方案达到5级完全自主，而是使用融合在一起的大量输入(传感器融合)，以便激光雷达/雷达和相机能够跟踪和避障。卫星导航提供相对于地球的绝对位置，惯性传感器测量汽车的动态。因此，改进惯性传感器对于确保可追溯性决策至关重要，因为它们提供的是测量结果，而不是静态LIDAR/雷达/相机数据的‘黑匣子’软件或人工智能/机器学习解释。为了使我们的传感器为汽车认证做好准备，我们将使用已知可满足监管测试的标准封装实践，并开始我们的质量保证流程，运行AEC-Q100认证定义的测试。我们的合作伙伴WAE将在赛车运动平台上进行测试，将我们的传感器暴露在真实世界的极端动态和条件下。为了验证惯性传感器在导航堆栈中的位置，以及与外部参考相比它所扮演的安全或任务关键角色，我们的合作伙伴UWE将利用他们在自动驾驶车辆方面的专业知识，对完整性和弹性进行独立研究。我们将与我们的合作伙伴皇家导航研究所合作，通过举办两场有学术和行业领袖参与的自主导航活动，确保我们的IMU和UWE的研究成果能被英国CAM硬件供应链看到，并鼓励围绕传感器融合、安全、弹性和英国立场的话题进行辩论和讨论。这些活动是在感兴趣的利益相关者之间产生新的见解和合作的机会。