Edge AI - An Inspiration for Future Drone Network

Last updated on Nov 10, 2023

[Article as first Published by Sanjay Kumar for Nerd For Tech here]

Consider yourself as one of the fish in a fish of school. Do you feel uncomfortable? Aren't fishes smart enough to observe and decide quickly where they should be swirling next? Why don’t they hit each other so often?

Not just fishes, look at this flock — How do they synchronize so well?

Even locust swarm knows the trick to synchronize themselves in this chaos. Locusts usually fly with the wind at a speed of about 16–19 km/hr. Isn’t this amazing?

Now, look at us — The Superpower Humans !!

We have conquered space but we don’t know how to organize this everyday chaos.

Thanks to Wright Brothers, who invented the airplane in 1903, and in the last 117 years we have taken their invention from creating a plane to satellites (3372 in numbers) which are administered only by super-secret government agencies.

⚡ My life wish to travel space would go in vain and I would never know if Earth is flat or round.

In the era of drones, what stops us to commoditize the super-secret ways to handle air traffic?

The system which was developed decades ago is slow and can’t handle the drone traffic efficiently.

But before we get into the solutionizing this, why do we need infrastructure to manage such dense air traffic?

Understanding Why?

55 g of CO2 / passenger/km for an average car. Small drone delivery can save emissions. Read more here.
Electric power generation is getting cheaper and clean every day — drones work on green energy. Consider noise pollution also when you think of it.
Faster & Timely Operations (medical help, online surveillance, monitoring, product deliveries, large farm monitoring, etc) as we can think of building layers over layers and make it more dense but organized like a school of fishes.
Low damages of product, as roads quality index, is not good everywhere and packages are being manhandled which results in damages.
Less Traffic on Roads — Online Shopping Was Supposed to Keep People Out of Traffic. It Only Made Things Worse.
Making the life of Traffic police easier, reducing road rage, rash driving by delivery executives, honking, etc.
Think about air surveillance, medical help

Why Not?

Let's understand the counterpart of it

Less space for the growth of natural inhabitants. As of now, I love waking up every day with birds chirping.
Lack of security in a drone could be disastrous. It can create problems as small as using your open WiFi to as big as planning targeted attacks.
Privacy Concerns: I hate drones flying over my house.
Hardware or software malfunctioning could lead to serious and catastrophic problems.
Drone Thefts — Hit it with a rock and it will retreat, now you take the control.

With all these counter arguments in mind, nobody would agree that we should build air traffic controller for drones. Drones are evil.

But we are humans, who always look forward to new challenges every day and we are smart enough to solve problems via communication, setting up protocols, government, laws, etc.

If we don’t have laws in place, the strongest person would have it all. The money wouldn’t have any value.

We will introduce more and more strong laws to grow and make ours lives convenient. These counterarguments can be taken care!

Going from 2D to 3D Traffic Controller

Ask any programmer, they will know how to build a Traffic system — but air traffic controller is a slightly harder problem and it is known to few.

Let's start with visualization — Can you visualize?

Let's start with 2D visualization first, Consider there is one car and it has to go from Point A to Point B. Draw it on a page (physical).

Now draw the same drawing in 1,000 pages and stack up all the drawings. This is what 3D Traffic would look like.

It is like ’n’ layers of roadmap stacked up.

Let's design a data structure to solve this.

Solving the hard part

Does it mean we just need a 3D way to represent the coordinates and time-based series data that can solve this problem?

YES, if you consider your space as a rhomboid.

But wait, how would you know the coordinates of your drone?

It is a hard problem because the earth is not flat (go here to read something interesting). Can we use GPS coordinate? It provides data with 1m accuracy.

Oh No, GPS elevation coordinates are highly inaccurate. It might be having 1m horizontal positional accuracy but elevation accuracy is worse. It has 2 to 3 times accuracy error than horizontal position accuracy.

So until we solve this location problem, we can’t build a transport system anything as close as a school of fish.

A quick solution

We need to reinvent the GPS system to create school of fish transport system.

But one quick solution which we can try is that we can create elevation bands and drones should transmit the elevation band continuously with its GPS coordinates.

But how do we create elevation bands?
We can develop local on-ground infrastructure to continuously track the drones and the elevation band of each drone.

Building , maintaining and monitoring such infrastructure is expensive. This is very heavy on communication side as everything resides in cloud which requires network (which is always flaky in certain areas).
So, what is the solution?

Introducing EDGE AI

Edge AI - Machine Learning algorithms to process data generated by a hardware device at the local level.

Make drones which is self serving exactly like fish. We need to build smart drones which understand traffic and take decisions in real time by themselves.

List of things which drone needs to perform in real time are :

Understand surrounding and take quick decisions.

That's it there is no more work.

Fast, Local, and Autonomous

Data is not required to be shared on the cloud so it may restrict us to store to serve only specific areas. You won’t need a build an infrastructure to make it happen as it will be “Autonomous”.

Welcome to the future - Autonomous Drone with Edge AI.

If you like the post, please share. We cover the future - stay tuned.