Modelling to Improve Positioning Accuracy in Urban Environments for Autonomous Vehicles

自动驾驶汽车（AVs）将引起交通运输的重大变革，并将提高车辆设计、能源消耗、车辆所有权和道路利用的效率（以及巨大的商机）。开发完全自动驾驶交通解决方案的一个关键挑战是，在目前基于gps定位系统的“盲点”位置进行导航，比如停车场和地下道路。我们的合作伙伴（VitalAlert）拥有可以解决这个问题的技术。他们目前的产品可以使用甚低频（VLF）磁感应与地下矿工进行通信，这种磁感应可以穿透100米的岩石和土壤。VitalAlert正在与大型汽车公司合作，寻求将这项技术应用于自动驾驶汽车。一个关键的挑战是VLF信号在穿过混凝土时会衰减，但这个问题可以通过精确的建模和校准来解决，以获得准确的位置估计。为了进行校准，VitalAlert开发了时域有限差分（FDTD）软件来估计VLF信号的传播。虽然这种方法可以提供可接受的结果，但它的执行时间很长（对于一个50米× 50米× 20米的停车场来说，需要数小时到数天）。改进这个计算时间很重要，因为它限制了产品的商业潜力。在这个项目中，我们将开发一种有限元频域（FEMFD）建模技术，以实现更快（即分钟）的精确建模。这反过来又将提高VLF定位系统的商业潜力。我们的方法是基于这项技术10年的经验。

Autonomous Vehicles (AVs) are poised to cause major transformations in transportation, and will enable efficiencies in vehicledesign, energy consumption, vehicle ownership, and road utilization (as well as large business opportunities).A key challenge for developing fully autonomous transport solutions is navigation through locations that are currently "blind spots"for GPS-based positioning systems, such parking garages and underground roads.Our corporate partner (VitalAlert) has technology which could address this issue. Their current product allows communication withunderground mine workers using Very Low Frequency (VLF) magnetic induction, which can penetrate 100s of meters of rock andsoil. VitalAlert is pursuing using this technology for AVs with large automotive companies.One key challenge is that the VLF signals are attenuated when passing through concrete, but this problem can be solved byaccurate modelling and calibration to obtain an accurate position estimates. To do this calibration, VitalAlert has developed aFinite-Difference Time-Domain (FDTD) software to estimate the propagation of the VLF signal. While this method can provideacceptable results, it has long execution times (hours to days for a 50m x 50m x 20m parking garage). It is important to improvethis computation time, as it limits the commercial potential of the product.In this project, we will develop an Finite-Element Frequency Domain (FEMFD) modelling technique to allow faster (i.e. minutes)accurate modelling. This, in turn, will improve the commercial potential of the VLF positioning system. Our approach is based on 10 years of experience in this technology.