Advanced Robotics Group (ARG) is a robotics startup based in Houston, Texas, where I interned over the summer of 2022.
They have not yet launched their product, and due to a plethora of NDAs I can't give any specifics on the purpose of the company. However I can on a higher level and a baseline engineering level describe my contributions and their impacts on the company.
Drivetrain Health Monitoring
In an autonomous robotics environment where once your mission has been launched, you can not interact with the robot whatsoever before mission completion, diagnostic and health data for the robot is essential. The ability to determine when components are nearing failure, and returning to base before then, can save millions of dollars and a lot of frustration. One of my jobs at ARG was to create a diagnostic tool using motor and motion data to determine the health of the drivetrain.
The drivetrain consists of many gears and wheels, and when looking at the frequency spectrum of the motor current, you can see spikes at specific frequencies relative to the ratios in the drivetrain. In a healthy robotic system, you can establish a baseline frequency spectrum, and then analyze deviation from that spectrum as the system runs. This can quickly identify fatigue in wheels or gears, and given test data you can learn what telltale signs of impending critical failure are. I developed a system which analyzed these spectrums and determined where potential failure points would lie. This system, along with many others used to monitor the health of the robot, are essential for mission success.
Testbed Design and Optimization
At ARG, design engineers would often waste significant time waiting for available test-bed space, or waste time transferring systems from the test-bed to the robot. I noticed this problem, and redesigned a new testbed, which included every electronic component of the robot, in locations that were easy to visualize, monitor, and change. All of the same cables and components could be thoroughly tested, and then deployed in a robot.
I also implemented systems which engineers could use to track availability of the testbed. Now, at just a glance, it could be seen if the testbed was available for use or not, as opposed to needing to go around the office asking other engineers each time something needed to be tested. This testbed system was so successful that the company went on to order multiple more.
Manufacturing and Documentation Frameworks
When working at a relatively small scale with very expensive components and custom made connectors and cables, room for error exists. Before my time at ARG, this had caused more than a little time and money to be wasted. Having cables that can be plugged in backwards, or cables that are not verified for correct pinouts, can at best waste time debugging the issue, and at worst instantly fry thousands of dollars of equipment.
One of my responsibilities was to take every custom electrical wire or component that was needed for the robotics platform, and find a system in which to implement it that was relatively infallible. I created extensive manufacturing documentation, and implemented tracking and labeling systems for each component. After implementing these solutions, time or money waste from these areas went from a weekly occurrance, to essentially unheard of.