Vision and Robotics Lab, AUB

Since I joined the Vision and Robotics Lab, my research with Prof. Daniel Asmar and Ms. Rema Daher was focused on simultaneous localization and mapping (SLAM) – mainly on improving the current multiple mapping algorithms. During my undergraduate research experience, I have worked on a new evaluation metric to assess the performance of map alignment techniques. The proposed metric automatically generates error measurements between the alignment displacement field and its corresponding ground truth. In addition, a ground truthing dataset upon which any map alignment method can be assessed was proposed. This paper will be submitted to IROS 2021. Having experienced how rewarding it is to solve problems in the most novel fields and contributing to them, I continued working closely with Ms. Rema Daher in developing her Master’s Thesis. We were able to formulate both a parametric model and a generative adversarial network (GAN) to create SLAM maps from as-built floor plans. These models act as high-fidelity replacement for real-time simulation. The machine learning model was also trained to provide as-built floor plans given a SLAM map which can be utilized to improve localization during real-time navigation. The culmination of this work will be published in the AUB Library and submitted to IROS 2021.