Curse or blessing? AI drone


Jun 07, 2018

Autonomous Intelligence for Drone Flight Control made during NVIDIA Jetson Developer Challenge

photo: Autonomous Intelligence for Drone Flight Control ,

The talk of Artificial Intelligence has been trending in the last couple years. There has been a lot of talks going on about this technology, with doubts of its benefit to mankind. The debate is yet to end. Therefore, we should be prepared for a lot more controversy when Artificial Intelligence starts being implemented in unmanned aerial vehicles (drones).

The word machine learning seems to be thrown into every statement by a lot of people. But the paradox here is, only a few truly understand what it really means. As humans, it is only natural to want to go against anything being predicted to take the place of humans. The question now is:

What is machine learning?

Machine learning is basically the learning of computer systems from lots of data, so predictions can be made on future occurrences. With the believe that future occurrences would follow the pattern of the past data. However, the questions remain:

Is it a blessing or a curse? Is the world really ready for this?

These questions can’t be answered in one go. However, something I know for sure is that the world has seen drones improve over the years. One very notable improvement we have seen is its affordability, decrease size and weight. In recent times, drones have become affordable to the average man, with the availability of small drones known as quadcopters. Quadcopters go for about a hundred dollars depending on the needed quality.

Where does machine learning or artificial intelligence meet an unmanned aerial vehicle?

With Artificial Intelligence, we may not need to have someone behind the controller anymore. This is because models may have been trained to move around obstacles using computer vision, self-supervised learning, reinforcement learning amongst a host of other techniques. Computer vision, helps in ensuring that computers can actually interpret what objects are present in images by converting the pixels into vectors.

What are the advantages of this marriage between artificial intelligence and drones?

1) Improved Drone Safety:

One may ask, “Why would a drone driven by artificial intelligence be safer than one driven with someone behind the controller?” Well, for people who have been into driving drones for a while. They would agree that when someone starts learning to drive a drone, a lot of mistakes would be made, leaving the drone to suffer a lot of crashes.

Artificial intelligence can be implemented in a drone, even with a controller so as to ensure crashes are avoided when possible. Considering drones are quite fragile, nobody would want to see their precious unmanned air vehicle get damaged after a collision.

2) More time for other tasks:

In cases of videography, an artificial intelligence drone would be very beneficial. Time needed to control the drone can be invested into making the stage great for a nice shoot. In the case of e-commerce as well, drones can be made to go deliver products without having to invest time in the driving of the unmanned aerial vehicles.

3) Cost Effective:

During construction, drones can be used in lifting lightweight materials to the top of a building. This is more cost effective than using a crane, or some other alternative in moving materials. This is also useful in a lot of other industries.

4) Improved risk-taking:

In cases of disaster, when it may not be advised for journalists to be at the scene due to possible dangers. Artificially intelligent drones can be used to take coverage of the situation of things, considering they would be better in reaching far areas. These drones would serve as great tools for carrying out tasks that could endanger the lives of humans.

5) More jobs:

This may sound weird, however it is of no doubt that drones driven by artificial intelligence would create more jobs. More people would be needed to create machine learning models, more people would be needed in embedded systems to implement the models as well. Artificially intelligent drones can create a lot of jobs, instead of taking them away.

6) Real time updates:

With the use of artificially intelligent drones, there would be more opportunities for real time data to be provided. Mundane tasks such as having someone to hold the cameras, or controlling the drones would be automated. Computers do not get fatigued, same level of concentration all through unlike like an human. Inasmuch as there are a lot of advantages, there are many disadvantages as well.

7) Cyber Security Issues:

With drones, there needs to be some form of connectivity to a point of control. By the time people become very dependent on drones driven by artificial intelligence, cyber criminals would have more reasons to hack into them. This means in essence that drones can be hijacked while at work. So while trying to make things automated and faster, AI drones are also opening greater possibilities of cyber crime.

8) Political Security Issues:

There is a lot of insecurity going on the world right now, and a lot of countries/states/parties/organizations are using drones as means of spying on their opposition. What this means in essence is that people would become skeptical of drones, and would most likely shoot them down the moment there is any sign of danger with a drone in sight. A lot of drones would be shot down, regardless of the purpose. It‘s better safe than sorry as they say.

9) Unpredictable crashes:

Drones have advanced a lot in recent times, however there has been little advancements in comparison when looking at developments made to the batteries. Since even the best of drones can‘t even stay in air for more than thirty minutes, what this means in essence is that drones would fall out of air the moment they run out of power. Asides the issue with power, machine learning models aren‘t perfect. Therefore there are going to be some unexpected fails during use. If there were to be some expert driver behind the controller, a lot of those crashes would probably be avoided. Except AI drones know when to return “home”, in cases of low power, there would be a lot of unpredictable crashes.

10) Lack of human factor:

In as much as AI drones can do a lot of things on their own, they wouldn‘t be able to do a lot of things with the creativity that comes with the human control. In videography for example, the human touch would be needed to take creative pictures which wouldn‘t be so much possible when done by Artificial Intelligence.

Which industries would benefit more from Artificially intelligent drones?

This technology would be very useful for those in the:

  • Real estate industry, to display housing properties to people in a new way.
  • Entertainment industry, to get live footage of events.
  • Farming, to check the situation of crops at far distances and carrying out land survey.
  • E-commerce industry, for the transportation of products.
  • Disaster recovery, to search for lost properties or get to places too difficult for humans to reach.
  • Filmmaking and photography, for recording events and taking classic pictures.

To show that the concept of adding artificial intelligence into drones is not new, there were a couple of projects developed on this at the NVIDIA® Jetson™ Developer Challenge.

First is TITUS, this project is an Unmanned Aerial Vehicle project, with machine learning implementations by the programmers to help the drone navigate its way in the forest. The team behind this project set out a goal do build a drone whose aim was to discover possibilities of wildfire in forests. Using a single camera, the programmers were able to create a drone to develop real time depth maps in forests.

Another project called On-UAV realtime detection of storm damage submitted to the same hackathon was made to help with the Unmanned Aerial Vehicle real time detection of places that have suffered possible damage from storm. This is quite similar to a drone used in cases of disaster, as this project was developed to search for survivors of a storm and carry out an assessment of damaged properties. The programmers made use of convolutional neural network which was trained on the dataset provided by NVIDIA‘s DIGITS software.

These amazing projects are created by programmers with little resources. Imagine what would happen when one of the big four Apple, Alphabet, Microsoft, Facebook and Amazon gets to invest big in this technology. Amazon has already begun testing artificially intelligent drones in the delivery of products, so there is a lot of potential to be achieved and milestones to be reached.

Artificial intelligence in drones has a long way to go as there are some hiccups with the technology. However, there is no turning back and a lot of research and work is being put into perfecting the Artificial Intelligence in drones. In the end, it‘s only wise to accept that Artificial Intelligence in drones is here to stay and would become a very reliable technology in the nearest future.