The Visual SLAM component focuses on mapping and navigation of the construction environment. Visual SLAM uses images (video) for sensing the environment, and, on a high level, works on feature matching between adjacent frames. We use the Intel RealSense 435D as our visual sensor, which generates an RGB-D iamge used by the SLAM algorithm. Along with the RGB-D, we use the visual odometry and the Husky's IMU sensors (fused with an Extended Kalman Filter algorithm), for SLAM using by RTAB-Map, an open source Visual SLAM package. The ROS navigation stack uses the map generated by RTAB-Map and does path planning, and this information, along with the odometry, is used for moving the Husky in the environment using move_base. We make use of tf and URDF files to link the coordinate systems of the base (Husky) and the camera (RealSense). This setup makes it easy for move_base to perform navigation.

• Clearpath Husky
• Intel RealSense D435
• Laptop with Clearpath Ubuntu 18.04

• ROS Melodic
• Rviz
• Clearpath Ubuntu 18.04
• realsense_ros https://github.com/IntelRealSense/realsense-ros
• rtabmap_ros http://wiki.ros.org/rtabmap_ros
• move_base http://wiki.ros.org/move_base

roslaunch realsense2_camera liu_nav.launch
cd integration
python monitor_and_move.py
python control_k.py