New Publication: SEESys: Online Pose Error Estimation System for Visual SLAM

Duke I3T Lab, in collaboration with Dr. Guohao Lan, has demonstrated the first method for providing online pose error estimation for visual SLAM, which enables downstream applications to quantify the quality of pose information provided by a SLAM engine. As part of this research, the lab conducted an IRB-approved 30-participant case study demonstrating the use of the developed system for an audio error advisory. The study demonstrated that our platform-enabled audio advisories helps guide human operators of SLAM-enabled devices, reducing pose tracking errors by 25%. 

• SEESys: Paper PDF
• GitHub repository
• Demo video 

Image

Thanks to the National Science Foundation and Cisco for supporting this research!