For my senior project in the CU aerospace program I was part of the autonomous aerial localization team. Our RFP asked for an aerial platform that could autonomously navigate a defined mission area to search for two robotic ground vehicles (RGVs). Once an RGV was spotted, the system used an onboard camera enabling computer vision to calculate a precise inertial GPS location of the RGV.
My team chose to address this design problem by building a custom hexrotor frame with two RGB cameras as our onboard sensor. One camera had a larger field of view (FOV) for initially spotting the RGVs. The other camera had a smaller FOV for precisely localizing the RGVs.
My specific role in this was designing the custom hexrotor frame and modelling its flight endurance. The main goal of this design was to minimize mass, improve flight efficiency by centralizing mass, and protect batteries from puncture in the event of a crash.
On the topic of crashes, we had four. Each time however, we were able to replace the broken parts within 48 hours and continue testing. This is a direct result of our very strong, modular frame design. If any other airframe had crashed the way ours did, it would have been toast. Our airframe was strong enough not only for me to fully sit on, but to withstand three crashes from an altitude of over 30 feet. When our airframe did crash, it did so exactly how I intended. The "hip" component shown below was designed to act as a crumple zone for the rest of the airframe. In all four crashes it did exactly that. Afterwards, the broken component could be very quickly swapped with a spare.
Ultimately, we were able to complete a fully autonomous mission, as shown in the video below. We were also recognized by the department with the gold medal award for our excellent modelling and simulation.
The gold components are 3D printed from PLA pro. They consist of landing gear, gear/arm mounts (hips), and arm mounts (shoulders). The hips and shoulders support the motor arms, but also provide structure to the rest of the frame. Each hip and shoulder is secured to each plate with four M3 bolts, making the frame easy to service, but still very strong. The hips in particular were designed to act as a crumple zone for the rest of the arframe in the event of a crash. Since they were so cheap, fast to manufacture, and easy to install, I designed them to break first. Unfortunately, we were able to validate this function several times, and the hips failed as designed.
The black plates are made of 6061 aluminum. These are simply sheet pieces with no bends or 3D features, so they needed to be as stiff and light as possible. Carbon fiber was not in the budget (and is also difficult to machine), so 6061 aluminum was chosen for its rigidity. The cutouts were necessary to reduce the plates' weights. Members connecting each motor arm were left for strength.
The tubes shown are made from carbon fiber. They were on clearance. Initially we left them at their full 12" length. However, to reduce vibrations and weight we eventually cut them down to 8". This was the minimum length needed for 1" of clearance between the propellers.
Ultimately the frame was much stronger than anybody anticipated. Following completion of the class, we decided to get a bit more risky with our structural testing. This led to us discovering that the frame could fully support my weight while sitting on it.
To model the endurance of the drone two key pieces were needed: a detailed mass budget and accurate thrust modelling.
Building our own drone gave us the ability to precisely estimate the mass of components using Solidworks mass properties. Where other teams had to take manufacturers at their word for masses, we could simply find the volume of our parts and multiply by the known material densities. For 3D printed parts we used the masses of 10, 20, 40, 60, and 80 percent infilled components to develop a model to predict the 3D printed mass density as a function of infill percentage.
When choosing motors for the drone, we were very particular about finding components with detailed and available thrust data. Using this data, we created functions for current draw (A) as a function as thrust (N) at the voltages tested (15.4V, 11.6V). Then, we linearly interpolated to find this function at the nominal voltage supplied by our batteries (14.8V).
Once we had these two pieces, we could easily determine the thrust required for hover from the mass budget. Then, we input that thrust into the motor model to find the current draw per motor at hover. Finally, we simply divide our battery capacity (Ahr) by the combined motor draw (A) to find the endurance of the drone.
Hip - Landing gear and motor arm mount
Shoulder - Motor arm mount
Top plate
Bottom plate