This week's update is also relatively short. We finished designing our presentation for this Friday early in the week and successfully presented our updates today.
With regard to our goals for this week, we successfully finished the majority of what we wanted to accomplish. Our algorithm has been, from what we can tell, successfully implemented in the latest version of ROS and is able to handle all four test situations (4, 8, 16, and 32 UAVs) without major issues. Roughly in the middle of this week, we ran into complications involving UAVs ignoring our algorithm six minutes into the simulation; however, we discovered later that this was caused by our visualization using Google Earth taking up too much time (writing to a file and then reading it for each UAV). Once we optimized our visualization, this issue went away.
Aside from debugging issues with ROS integration and malfunctions due to inefficiency, we also worked on determining a good ratio for repulsive and attractive forces. Furthermore, we brought back the expanding field method, which had previously been discarded. Right-hand turning was fine-tuned once again as new situations were discovered that resulted in UAVs colliding. While the system is not entirely perfect, it has seen vast improvement from earlier weeks and is able to handle 32 UAVs very well in a 1 km by 1 km range.
Overall, we feel that our algorithm has reached a point where it is fairly stable provided that a realistic amount of space is given based on the number of UAVs to be flown. Small tweaks are still being made to see if any last-minute improvements can be made, but the overall system is working successfully within the ROS framework.
Working Progress >