Inspiration
Living in a city can be an amazing experience, but congested streets can make it difficult for people to move around and products to get where they need to be. While considering different ideas, we thought drones could be a solution to the problems some cities experience. For example, this past Fall, a man in San Francisco clasped of a cardiac event. While a defibrillator was located in a nearby building, the manager would not give it away to help the man outside. Many times, people don't have access to life-saving equipment, and even if some are located nearby, there is no guarantee they will have access to it. This is where
What it does
Our smart delivery drone can be equipped with various payloads, all aimed at bringing aid and assistance to those in need. A drone delivery system completely bypasses the traditional problems associated with cargo transport in congested cities. One key example we focused on during this hackathon was delivering medical supplies.
For instance, in Manhattan, emergency services response times have risen this year, directly correlating with an increase in casualties. According to the American Heart Association, the chance of survival during a cardiac emergency decreases by 10% with each passing minute while waiting for emergency medical services.
How we built it
For this hackathon, we built a miniaturized proof-of-concept for indoor use. We used a budget off-the-shelf quadcopter as the flight vehicle that one of us already owned. We had three main problems to solve to create our project.
-Find a way to control the drone from our computer.
-Track the drone in three-dimensional space and use a control loop to position it.
-Create and attach a lightweight and functional payload delivery system.
To control the drone from our computer the simplest way was to pass the outputs through the RF transmitter via the trainer port. The trainer port is designed to allow a new RC pilot to fly under the supervision of an experienced pilot. This was perfect for our very unskilled control loop especially during the first testing phases as it allows you to take over manual control at the flick of a switch ensuring safety. For the camera tracking we magnificently implemented 4 channel feedback (X, Y, Z position as well as Yaw angle). We achieved this by placing 3 IR LEDs on the drone in an isosceles triangle to easily find the heading by locating the point with the greatest distance from both other LEDs. The upwards facing camera is able to detect Z (and thus compensate for it's FOV at different distances) by calculating the perimeter made by the triangle and comparing it to a calibrated value. From here, X and Y is easy, as we can simply find the center point and compensate for distance.
The camera preprocessing is also intense involving a physical IR filter derived from a floppy disk. This is then further filtered by software to isolate the blue channel as the camera sensor responds with a deep blue hue for any intense IR source seen through the filter. This makes it sufficiently noise-resistant and robust enough with great accuracy, making tuning and stabilizing the control system easier.
The control loop is complex as it involves giving acceleration inputs from position inputs with lots of external factors that can affect the system as well as inherent drift. We went with nested PID controllers as the outer loop allows us to output an instantaneous velocity vector setpoint which is then passed along to the inner PID loop which controls acceleration (the absolute pitch and roll) of the drone.
Challenges we ran into
One of the challenges we experienced was reverse engineering the radio transmitter training mode protocol. We connected the radio to an oscilloscope and were able to analyze and recognize the type of signal. Another issue was creating the passband filter for identifying the inferred light. To overcome this, we repurposed old floppy disks as an IR pass filter. This, combined with digital filters in Open-CV, created a clean image that we could process.
What we learned
For this project, all of us learned and tried something new. During this project, we all gained experience in tools such as Open-CV, how to use hardware timers and interrupt service routines, and the complexities involved in programming low-level hardware.
What's next for Untitled
If this were to be developed into a real product, we would recognize that there would be different challenges to overcome. The main challenges we face would be gaining public trust and possible noise pollution.
We envision using a SLAM algorithm with a Lidar module to control the drone. This allows the drone to know its location and navigate safe paths. We would also implement many failsafe features, such as a landing routine and a parachute as a last resort. Addressing the potential noise pollution concern, modern propeller designs allow for high thrust and low decibels, not being a nescience when flying overheard.
This product has the real potential to benefit and save people's lives, and we believe a trail in a city would be a worthwhile venture.
Log in or sign up for Devpost to join the conversation.