Open Gaoee opened 6 years ago
Hi @Gaoee , sorry for my late reply. Once robot A knows that, in frame X, it observes the same scene as robot B in frame Y (from decentralized visual place recognition), robot A sends all data necessary for relative pose estimation between X and Y to robot B. Then, robot B establishes the pose of X relative to the pose of Y (no common frame of reference needed) , and sends it back to A. This data comprises feature point locations and their descriptors, and relative pose is established using P3P and RANSAC.
Many thanks for your reply.
Hello, I have read the paper "Data-Efficient Decentralized Visual SLAM" of you. And there has a question I can not understand. How to estimate the relative pose between robots, did the robot know the initial pose of other robots? Or is there have a common reference frame knew by all robots? Look forward to your reply, thanks.