This was a more established project at CERLab, which focuses on the development of an autonomous mobile welding platform for the assembly of ship hull parts. I joined this team toward the conclusion of this study, where I wrote a ROS (Robot Operating System) waypoint manager.

ROS Node: Waypoint Manager

 The manager sequentially feeds a series of optimized target locations to a local planner and tracks the position of the robot to ensure that each target is achieved. Created as a ROS Node in C++, it subscribes to a previously written global planner that generates a list of target locations, and the move_base/feedback topic, which tracks the state of the robot. Based on the location of the robot, the manager will post the next waypoint to the move_base/goal topic once the previous target is reached. This node is actively being developed. 

Video Explanation: The triangular shapes are the ship hull parts that will be welded. The global planner optimizes the locations and sequences of locations that allow the robot to execute welds along the supports seen in the center window. The right two windows represent the progress tracking reported by the waypoint manager. (Locations in the video were generated arbitrarily for testing)

I joined the UGV group when it first began at CERLab, where we explored the application of Semantic SLAM to UGVs, specifically a Clearpath Husky. As we established a workspace for the project, I performed a blind-spot analysis to select the optimal LIDAR sensor and location. 

Blind-Spot Analysis

First, I placed each LIDAR above the center of the UGV, ensuring that the heights of their "blindness cones"(areas outside the field of view of the LIDAR) were minimized and that the UGV remained in these regions. This maximizes the amount of usable data potentially generated by the sensor. Based on this analysis, the Ouster OS0-32 LIDAR was selected.