Find Your Machine Learning Mojo

12 Trials and Tribulations in the Life of a Mini Drone

Drones and mini-drones are pretty cool, aren’t they?

Have you ever spent time thinking about the tech behind an artificially intelligent drone though?

There are so many different questions the drone needs to answer.

It doesn’t even need just to answer these questions once either – it needs to be able to make these decisions, and get them right, consistently in real time.

In this article, we’re going to be looking a little deeper into the trials and tribulations in the life of an artificially intelligent mini-drone.

To help us really get into the mind of the drone, we’re going to be looking through all the questions from their perspective.

Meet Mini Drone!

meet mini drone
This is mini-drone!

Ok drone, your mission should you choose to accept it, is to get this undisclosed, top secret package to the other side of the city.

Do you accept?

‘Hell Yer!!’ – said drone (trust me :D)

Off you go little drone – see you on the other side!

Thinking about logistics

Where am I going?

where to go

The first question our pal needs to ask itself is where is it going? To answer this question he needs to understand several things, where he is located, what potential route he can take, obstacles, where the destination is.

Then comes the weather

What is the weather going to be like?

Can I fly in the rain?

If it is windy, how does this affect my flight performance and fuel requirements?

If it is windy and I transition from “free air” (windy conditions), to the shelter of the building, how will I behave? Do I have sufficient control margin to cope with the transition, or will I crash, or fly into the building?

can I fly in this?

How heavy am I, and can I reach that destination?

If Drone is too heavy, he won’t be able to fly properly, and he may run out of energy – not good!

What about other things that might be in the way of Mini-Drone?

I am tiny and hard to see. How do I make sure other drones and human-operated vehicles, do not fly into me?

Do I transmit a beacon to other vehicles (if so, how can I be sure that they are all listening)

Do I electronically file a flight plan that tells other vehicles where and when I will be (but what happens if I am late or early)?

dodging things with mini drone

How will drone get out of the way?

How do I make sure I don’t fly into other vehicles?

Do I listen out for their beacons?

What about aircraft that don’t have to have beacons (some light aircraft, gliders, and balloons)?

Do I use visual means to detect them?

Can I do this in all light conditions, including directly into the sun?

Do I use the air traffic control radar data in real-time, fed into my control system?

What about fast-moving military jets and helicopters?

How do I make sure I don’t fly into obstructions?

Do I use a terrain map? If so, what happens when a tree grows a little higher?

What happens if someone erects a whip antenna, or puts a balloon up (completely legal at low altitudes, often without telling anyone).

What happens if my map is wrong?

How do I ensure I have enough flight performance to clear an obstruction in the future part of the route? Do I assess in real-time, or have predictive performance planning?

There are whole areas of airspace which I am not allowed to fly (e.g., near Gatwick 😊), so how do I make sure I stay away from these locations.

mini drone and planes

What happens if my brain fails mid-journey?

Do I have reversionary systems which take-over to make sure sensible decisions are made?

What is my center of gravity with and without the package and can I survive a motor failure in both conditions. Would I have enough thrust and maneuver margins in a degraded mode?

Without a good understanding of this and be prepared for a failure this Drone would be pretty f****d if something did go wrong!

mini drone break

If something goes wrong, then what do I do?

Return to the base?

But then, what happens if I get another failure on the way home which means I crash?

Land immediately?

How will I know that there isn’t a building/dog/child/petrol station directly below me?

Go to a safe pre-arranged landing spot?

What happens if Mini-Drone gets another failure on the way?

Intentionally crash somewhere to avoid a school and minimize further damage deliberately?

Do I tell someone that I am about to crash and where?

How do I tell them?

When I reach my destination, how do I know it is safe to land?

Are there powerlines?

Is the ground level?

potential obsticles

Am I in the correct location?

Are their people or animals around?

Once I have begun my landing process, what do I do if my situation changes?

I am just about to land, and a child or dog runs towards me – do I?

Cut my engines to avoid cutting them with my propellers, but risk hurting them?

Fly away, but risk hitting them or cutting them?

How do I ensure that the leaves in the trees blowing in the wind are not incorrectly detected as a threat so that I have a reasonable landing rate?

testing for drone

How do I prove that I am intelligent enough to make sure all of these decisions are correct.?

I must make sure I don’t kill people (or kill sufficiently few that I am no worse than existing safety standards). I (and all of my drone friends) should only kill someone every 1×10-12 times per flying hour. For a quadcopter, which essentially will fall out of the air as soon as a single motor fails, this is a challenging target to achieve.

As you can see, there are many, many questions little Drone has to navigate to make his journey.

So how does Mini-Drone start to make these decisions?

Where does artificial intelligence come in to help Drone make these decisions?

Many different techniques go into making an artificially intelligent drone.

Two that I will cover (very) briefly are computer vision and reinforcement learning.

What is computer vision?

Computer vision is technology that allows our friend mini drone to see. Without computer vision, little Drone wouldn’t know if it was going to crash or if it was safe to land.

The drone will be able to look at the different features of an image and understand what it is seeing and what it needs to avoid.

You can read more about how computer vision works here.

What is reinforcement learning?

Reinforcement learning is where the computer learns from experience. It does this through trial and improvement.

The way the mini-drone learns is by reinforcing the correct answer. The more experience it has, the more it is able to understand.

As the mini-drone builds up, an understanding, it can react to different situations.

You can read more about how reinforcement learning works here.

Are you ready to get started with your own mini-drone?

Advertising Disclosure: Artificially Intelligent Claire may be compensated in exchange for featured placement of certain sponsored products and services, or your clicking on links posted on this website.

Here are 3 of the best-rated mini-drones you can purchase on Amazon!

1.Potensic Drone with HD Camera, Mini Drone with Induction Mode of Gravity

From the reviews, the screen is great to see what your doing and this one is good for photography. Can be a bit unsteady though.

Rated: 4.8 Stars from 38 Reviews

2. DROCON Traveler Beginner Drone with Optical Anti-Shake HD FPV Camera 

The reviews say it’s great for photography and easy to use for beginners

Rated: 4.6 Stars from 82 Reviews

3. SGOTA RC Drone Foldable 2.4GHz 6-Axis Gyro Remote Control Drone 

From the reviews, easy to use and great for kids. Simple to set up and has a good battery life.

Rated: 4.9 Stars from 39 Reviews

Not everyone likes mini-drones though!

You can’t fly them everywhere.

Airspace Regulations overview

So you don’t get caught out here are some of the rules for flying your mini-drone.

Please note these may have changed since the time of writing and are the UK Airspace laws – local laws outside of UK may differ.

Basic rules for operating drones in UK airspace based upon CAP722:

Without specific permission, the drone must remain within visual line of site (usually 500 m) of the operator, who must be able to control the drone actively.

Drones must be lighter than 35kg in weight.

It cannot fly within 50 m of any building, road, vehicle, person, unless I own it, and it is under my direct control (reduced to 30 m for take-off and landing).

As a business, even if I do follow the regulations above, I need specific approval from the UK CAA to fly a drone.

This requires you to prove how the drone plan to mitigate risks and provide evidence of the qualification of my drone operators.

Drone can break the above rules, but only with specific approval from the UK CAA.

This needs a full risk assessment and more stringent safety procedures. I have to prove how I can intervene in the event of a failure.

So, If my drone is 2km away, how will I know it is safe to fly across the road and that there are no power lines?

How will I take control if something goes wrong and what will you do? How will I make sure I don’t fly near another aircraft?

This post was proofread by Grammarly

Advertising Disclosure: Artificially Intelligent Claire may be compensated in exchange for featured placement of certain sponsored products and services, or your clicking on links posted on this website.

Love it? Pin it!

mini drone life

Leave a Reply

Your email address will not be published.

%d bloggers like this: