Lidar-pointcloud-based motion estimation to support 3D modelling and autonomous route driving

点击翻译按钮获取中文摘要

Mobile robotics technology has advanced at a rapid pace in the last decade, making applications such as automated mining, agriculture, safety and security, and self-driving cars a near-term reality. One of the core tools needed to enable this vision is the ability to estimate motion from a moving platform. This motion-estimation capability is a key prerequisite for accurate three-dimensional modelling and autonomous route driving. Over the last several years, the University of Toronto has been developing a lidar-based motion estimation pipeline based on tracking a sparse set of points in intensity images. While this technique works well, it is limited to using data from sensors that provide these intensity images, which are not particularly common and typically have a small field of view. The main goal of this proposal is to modify this core technique to allow for motion to be robustly estimated (in realtime) from the common spinning-lidar sensor configuration, found on many of today's robot (e.g., the Google self-driving car). This sensor setup produces pointcloud data rather than intensity images. The project will be carried out in partnership with Applanix Corporation, a well-established world player in vehicle positioning located in Toronto, Canada. Applanix Corporation is already a top global provider of positioning systems for ships, aircraft, and self-driving cars. Incorporating lidar into Applanix's products, will open up new applications for the company, particularly in autonomous mining and automated surveying.