MBZIRC

As part of my involvement in the Computational Multiphysics Systems Laboratory at Virginia Tech, I worked one year for the MBZIRC 2020 robotics challenge.

I led the graduate students as team coordinator. I organized weekly meetings and progress updates on software and hardware development. I also played an important role in the project proposal and progress reports which were submitted to the competition committee. Our team was ranked 1st for its progress report among US teams and successfully passed all phases and competed in MBZIRC.

Our progress reports are available:

My technical work included assembling, interfacing and programming of the Clearpath Husky robot, an unmanned ground vehicle. I designed the systems interface, namely power system, network connections, ROS compatibility, sensors configuration and data collection. I used the 3D Quanergy Lidar and the 2D Sick LMS lidar. Other hardware used included the ZED stereo-camera and the robot manipulator Sawyer.

I also built software to perform Extended Kalman Filter Simultaneous Localization and Mapping, written in C++. I used distinct landmarks in the environment which were recognized by a Deep Learning based object detector, You Only Look Once (YOLOv3), while receiving depth information from the ZED camera and a semi-supervised method called stereo-DNN.

Moreover, I customized the robot_localization package for odometry estimation and for the prediction step of the EKF-SLAM. At the same time, I worked with the Google Cartographer package, as a second approach to the SLAM problem and used the Lidar for that purpose. I also programmed the Sawyer robot to perform autonomous grabbing of tagged building blocks using a customized magnetic gripper controlled with a microcontroller.