Event-Camera, Simultaneous Localisation and Mapping for Autonomous Systems and Mobile Devices

用于自主系统和移动设备的事件相机、同步定位和测绘

• 批准号: 103644
• 金额: $ 27.17万
• 依托单位: SLAMCORE LIMITED
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2017
• 资助国家: 英国
• 起止时间: 2017 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=103644
• 关键词: Event; Camera; Simultaneous; Localisation; Mapping

Robotics and autonomous systems use sensors to understand where they are in the real world. The information from these sensors feed into a model which the system needs to create and locate itself within simultaneously. This challenge is known as Simultaneous Localisation and Mapping and is so fundamental that all truly autonomous products will require a dedicated SLAM system. To date, accurate real-time SLAM systems require expensive sensors combined with high power (both computational and battery) hardware. A novel sensor called the Event Camera was developed nearly a decade ago which has the potential to replace existing SLAM approaches but the underlying algorithms have proven very difficult to design due to the Event-Cameras fundamental architecture. It works like the human eye with each pixel only responding if it detects a change in intensity. This means the amount of information it generates is multiple orders of magnitude smaller but structured in a completely different way. The Slamcore team are pioneers of SLAM solutions and have recently demonstrated the first real-time, Event-Camera SLAM system. This project will take Slamcore’s technology and develop it further to demonstrate how Event-Camera SLAM will enable a revolution in autonomous products.

机器人和自主系统使用传感器来了解它们在真实的世界中的位置。来自这些传感器的信息输入到系统需要同时创建和定位自身的模型中。这一挑战被称为同步定位和映射，并且是如此重要，以至于所有真正的自主产品都需要专用的SLAM系统。到目前为止，精确的实时SLAM系统需要昂贵的传感器与高功率（计算和电池）硬件相结合。近十年前开发了一种称为事件相机的新型传感器，它有可能取代现有的SLAM方法，但由于事件相机的基本架构，底层算法已被证明非常难以设计。它的工作原理就像人眼一样，每个像素只有在检测到强度变化时才会做出反应。这意味着它生成的信息量要小几个数量级，但结构方式完全不同。Slamcore团队是SLAM解决方案的先驱，最近展示了第一个实时的事件相机SLAM系统。该项目将采用Slamcore的技术并进一步开发，以展示Event-Camera SLAM将如何实现自主产品的革命。