This technology is a vision-based collaborative simultaneous localization and mapping (SLAM) system. To identify rendezvous situations between multiple platforms and fuse individual SLAM maps, it uses an optimization algorithm at a ground station to compare and calculate the movement of non-static features extracted from each platform, as well as the similarity in pose and map information between platforms, to perform inter-platform matching.
Existing collaborative SLAM systems have significant limitations in sensor configuration and operation, as they often require the mandatory use of visual markers or precise distance information for platform identification.
This technology proposes a method that tracks and manages non-static features—which are typically discarded—from footage collected by a monocular camera without the need for separate markers. By formulating the detection of inter-platform rendezvous as an optimization problem, it enables the fusion of individual map and pose data at a ground station, allowing for collaborative mapping among multiple robots with a simple sensor setup. It can be applied to multi-drone exploration, collaborative mapping in disaster zones, and indoor autonomous navigation. Since it is implemented using only a monocular camera without markers or expensive sensors, it significantly reduces system construction costs.
US12624952B2
