Abstract:
|
This project consists of aid an existing SLAM system by developing and implementing a motion estimator based on the Inertial measurements of a VI-sensor in the SLAM pipeline. This entire project has been carried out at “La Sapienza-Università di Roma” under the supervision of the PhD student Ciro Potena and professor Daniele Nardi. The project has consisted first on the development of an Inertial estimator, the evaluation of its accuracy and later the software implementation integrating the Inertial motion estimation into the feature tracking of the system to enhance its performance. The first chapter of the project is an introduction of what SLAM is. The second chapter explains the theoretical approach used to carry out the project. The third chapter explains all the modified code in the existing SLAM algorithm needed to make the system work with the IMU estimation and the results of all the experiments. First testing the IMU estimator and later testing the SLAM algorithm fused with the IMU estimators. Finally, in the fifth chapter, we explained the conclusions we got from the results of the experiments. All the developed software needed to integrate the IMU integration system in the SLAM pipeline has been coded in C++. We worked with the Robotic Operating System to interface with sensors, and to record datasets. For the analysis of the obtained results, we used the open-source dataset EuRoc. We used the Version-Control system GIT to store the developed code and to track the changes in the source code. The IMU estimator source code can be found on: “ https://gitlab.com/Carlos_IbCu/imu_project.git ” And the ProSlam algorithm can be found on: “ https://gitlab.com/Carlos_IbCu/srrg_proslam.git |