Robust multi-robot localization and mapping

SLAM （Simultaneous Localization And Mapping）在过去的几十年里一直是一个非常活跃和快速发展的领域。它试图解决真正自治系统中的基本问题：如何从数据中提取信息并融合信息以创建环境地图，从而促进后续决策过程。尽管进行了大量的工作并取得了巨大的进步，但由于当前SLAM系统的脆弱性，长期操作仍然是一个悬而未决的问题。特别是，感知混叠，即不同的地方看起来与传感器相似，阻止了“闭环检测”，即识别已经访问过的地方，以及在很长一段时间内实现可靠性能的地图估计。

SLAM (Simultaneous Localization And Mapping) has been an extremely active and fast-evolving field for the past decades. It attempts to solve the fundamental problem in truly autonomous systems: how to extract information from data and fuse the information to create a map of the environment, which facilitates the subsequent decision making process. Despite the extensive works and vast advances, long-term operations remain an open problem due to the brittleness of current SLAM systems. In particular, perceptual aliasing, where different places appear similar to the sensor, prevents ``loop closure detection'', i.e. recognizing a place that was already visited, as well as map estimation from achieving reliable performance over long periods of time. This project aims at studying novel ways to perform distributed SLAM and solve the aforementioned problems. At the end of the project, we expect to have new algorithms for sensor fusion, as well as a distributed SLAM solution that includes synchronization and map merging. This system will be integrated in our partner's set of products and used for accurate mapping of environments without GPS (indoor, caves, forests, etc). The project will contribute to a product for SLAM that will provide state-of-the-art solutions for Canada's drone and robotics industry in general, with a direct effect on the Canadian economy.