Software Team Lead

  I had not originally planned to be software lead for this Rover year, as I was already working two part time jobs and going to school full-time, but after my best friend Nick decided to become team and electrical lead, and my other best friend Dylan would create the rover's arm for his senior design project, I just couldn't say no. This also came hot off the heels of me being the emergency software lead for Rover the previous year, writing enough code in the nine days before competition to at least give them a fighting chance, which they had, so I was even more prepped and ready. I only had one hard requirement, joining as the software lead, and it was that we would have to create a serious ground station setup. When I first joined the robotics club in 2011, I did so because I was enraptured by the rover, but especially its ground station, which was a full-screen display full of stats, indicators, and buttons that got the nerd in me excited. If I was going to go all-in, I was going to do the same, and ended up making the ground station software my university senior design project.

  The upcoming year of work turned out to be some of the most intense I ever had at OSU, mostly self-inflicted due to not wanting to let this opportunity get wasted. It's worth noting, that not only was I writing the ground station software, but also the software running on the rover's onboard Jetson TX2 computer, all the firmware for each of the many distributed systems, on top of also hand-assembling/debugging most of the rover's PCBs, and creating many of its electrical wiring harnesses. In total, I put in over 2000 hours of work on this project, while working two part-time jobs and taking classes. It was a lot, but, it also turned out to be my proudest achievement while at OSU, so I'd argue it was worth it!

  After these countless hours, what resulted was some incredibly robust hardware and software, which didn't crash, and which allowed us to absolutely crush the competition at the Canadian International Rover Challenge in Drumheller, Alberta, Canada. We won first place, with a final score of 287 points. The second best team got 159, so it wasn't even close! On top of doing that well, the software was also feature complete by halfway through the competition, which was no small feat considering the amount of scope creep a project like this tends to encounter. I was particularly proud of the ground station by this point as well. Not only was it a treasure trove of information about the Rover's state, but included live GPS mapping, multiple switchable camera views with pan/tilt capabilities and quality-switching, artificial speed limiting, and automations to automatically complete some of the tasks from the competition with just a few button presses. Considering the robotics team hadn't placed over 3rd since before I'd started in 2011, finally getting a first place win was a breath of fresh air, and made it much more cathartic when graduating the following year.

  The ground station itself contained an Intel NUC with a Core i5, 16GB ram, nvme ssd, and was connected to two 23″ 1080p monitors. I chose two so we would be able to see all primary Rover systems without having to swap between tabs or pages, which I’d had to do on some other software I’d written and knew wouldn’t be efficient when we were in the middle of a timed event. The desktop ran Ubuntu 16.04 with the ROS2 stack. For the UI, PyQt was used, allowing for easy tweaking using QtCreator, and which provided access to QtSignals, a feature that made simple cross-thread communication much easier in Python. The Python code itself was Python 2, which was a limitation of the ROS2 stack. Control of the rover was done via USB xbox controllers, plus a keyboard and mouse.

  The Rover itself ran on an NVidia Jetson TX2 single-board computer powered from a 500Wh li-on battery. The electronics box also contained a fuse box, power cutoff, network switch, battery backup, composite to network video converter, usb hub, FrSky controller receiver, and custom usb->RS485 adapter board. The TX2 ran Ubuntu 16.04 with the ROS2 stack, just like the ground station. All remote control boards on the Rover were communicated to using the modbus protocol over RS485. Interfacing with the arm was achieved via RS485 as well, but using the custom simplemotion control library provided by the motor controller company. Using linux UDEV rules, usb->serial and cameras were symlinked to custom, repeatable names at /dev/rover/name so that devices could be accessed identically after reboots. Custom ROS nodes were written for each system to be controlled, and the whole Rover ROS package set to start automatically at boot. The Rover could be controlled with an FrSky X9D controller whether the ground station was running or not. The controller could also override all other control inputs on-the-fly by flipping a switch on the controller. This was a safety feature to ensure runaway ground station or autonomy code wouldn’t allow the Rover to take off on its own. Conversely, the Rover would also work without the controller powered on, but would immediately allow for this override if turned on at any time. Communication with the ground station was done using two Ubiquiti Rocket M2 long-range wifi radios, along with their 10dB omnidirectional antennas. This simple setup allowed for a transparent ethernet link to be made between the two systems, allowing for useful features such as remote debugging and code upload. The radios were tested to work at roughly a kilometer away, when mostly line-of-sight.