Harnessing the Potential of Drones to Solve Big Problems.
At a very young age, I got introduced to the world of flight, specifically the field of remote control aircraft. My brother and I started off with buying a prebuilt racing drone, which we got pretty good at flying, then an airplane, which was fun, but buying these aircraft never got to satiate our hunger to tinker around and build something. After this, we started to build our own aircraft and eventually got one to fly.
However, one thing that we noticed was that it was extremely hard to actually build and fly a drone from scratch, with difficulties in both hardware and software. A typical quadcopter drone requires precise control within each of its 4 motors, and it requires complex control systems in order to do so.
A drone needs to convert its current angle, other forces it’s facing (wind) and weight into a desired angle and velocity by changing the speed of each motors. This gets even more difficult considering that the drone’s motors are rotating in different directions to counteract the torque produced by each one.
So, for example, if the drone wants to turn to the right, it will increase the speed of some motors and decrease the speed of others in order to do so. Keep in mind that a drone can move in 3 possible axes, not just yaw.
I was once scrolling down the arduino forums and found a guy who was trying to build an arduino quadcopter in a few weeks and asked the forum where to get the resources. The responses were actually pretty negative towards his question, most of the people saying that he would be better off buying his own flight controller. This is where the inspiration for my idea came. There are probably so many people who are interested in drones but are daunted when building their own drone, that it would be probably be difficult for them to get into the field of programming drones to complete more complex tasks (e.g. autonomy, collecting spatial data). Imagine we could create an environment where someone could get a drone and immediately have an interface where they could easily start programming on it to complete tasks.
My idea was to create a developer drone that can be easily interfaced with, in order to make the process for creating an application for a drone much easier from a software perspective. It should also be able to facilitate input sensor data and handle computation from neural networks.
The main features I am trying to implement are:
- Easily programmable input/output features
- Live, low latency video feed from a drone to a laptop
- Allowing a laptop to control the drone using a TCP connection
- Allowing space for expansion from a hardware perspective
- The use of object detection and other ML features (e.g. gesture tracking) to be implemented
With limited programming knowledge, this is probably one of the hardest projects I’ve tackled so far, and implementing all these things will be extremely fun and help me learn a lot.
The process.
Being one of the most difficult projects I’ve tackled, I’m building a whole bottom-up learning process for this in order to make sure that I get the most out of it. I am currently learning about IoT and how I can use it’s amazing potential in order to make communication much more efficient and allow for better data collection from the drone.
I’m also learning how to use unity and its ML agents in order to create a much faster reinforcement learning training process so that I don’t need to go through multiple drone crashes in order to implement some machine learning model in the drone.
Currently, I’m working on controlling a drone with the keyboard on a laptop as a start to a control feature system that I can work on, and I will start working on an AI control system very soon.
I will be sending in updates on this project very soon so stay tuned!
For now, check my website out and get ready for an epic journey!!