[Experiment 6] I built an autonomous drone defense game to learn about drones and autonomous AI

I continue my experiment journey. Number 6 now and there is, I believe a net improvement compared to the previous experiments. As I have mentioned in one of my X posts, I do feel that the experiment style works very well for me. It is kind of like if I create my own education. I work on something until I feel that it is enough, then I move on.

For the experiment 6, I stick to the same structure as for experiment 3-5. Use AI LLM, to create a game about a subject so that I enjoy learning about it.

What I feel is that AI can be used easily to go from 0 to 50% learning in a new subject. Maybe even 75%. A person with experience will still know more about something than someone that only talks with AI. However, someone that got experience and then uses AI in the same subject, will go way above 100%.

In short, superhuman skills are achieved when a user with expert level skills in an area decides to take advantage of the AI capabilities in the same area. This expert will then operate way faster and focus on links that are non standard so the AI won’t suggest it, but when prompted correctly, AI will find it.

I feel like these learning games that I have created are kind of AI slop. They are quite good and there are some good functionalities, but there are still some stuff missing. It can be felt that they are not fine-tuned and perfected. That is fine anyway as I am not trying to sell them.

But every game still gets better and in this experiment 6, I kind of moved away from the click, wait and win approach and instead now there is actually some thinking. I still feel that it is not at the level it could be but it is better.

In order to push the motivational driver further, I have pushed through Claude Fable to use the Octalys framework. Anyone can google it (https://octalysisgroup.com/octalysis-framework/ by Yu-Kai Chou) so I won’t go in details about it but I would like to highlight that I did use it specifically to ask for the creativity driver (driver 3).

The creativity driver is that a person that likes to be creative in finding his or her own path, will enjoy this game.

How was the creativity driver added? I made the drone base to have multiple levels and the skill tree is now closer to the Cyberpunk 2077 skill tree (not at that level of course)

The Change

The skill tree that was created was based on investments. In short, the user earn money through contracts and can then by more skills/hardware. Learning was then coming passively from questions asked from customers with a multiple option answer.

Basically the whole game was financially based and still kind of click and win

What I felt was that the learning should truly come by motivation so a levelling system was added and for each level, the user gets a skill point. Customer types were then added:

• Super rich
• Private
• Industrial
• Government

Contract with those are all different as the player evolves. Maybe data encryption is more important for an Government customer than a private house etc. All this leads basically to the fact that each customer type requires unique skills from the skill tree.

I also added a limitation. The total number of skill points are only 65 while the total number of available skills are 84, which means that the user must optimize

Now the learning comes simply from that in order for the user to get a good high score, he must study in detail the skill tree and choose and invest in the path which is aligned with the customer.

This is how it looks:

And if we dive in the details of the Government customer, we can see certain requirements including the encrypted links

We can also see in the skill tree under Flight Software that Encrypted links depends on companion computer, which in itself requires Waypoint Missions. For completion, the encrypted links also requires that Long Range Telemetry Radio is active from the Hardware skill tree.

The skill trees that exist are:

• Hardware: About the actual components that are attached to the drone
• Flight Software: The software installed on the drone computer
• Search and Navigation: The algorithms (and graphs) that runs
• AI Vision: How the drone detects objects like human vs animal classifier
• Swarm Technology: How multiple drones collaborate within the same space

I think this game is better than the previous learning games I have created. I still don’t think it is good enough to work in an educational environment but it is a good initiation

Improvements for a v2

There are several ways in which I could progress.

One would be to make it more like a tower defense game. Currently it is based on customer contracts but it could also be that the player gets a building to defend and can expand in many ways.

For the actual current game, the end game is really bad. Waiting time for contracts to complete is just really long. In the beginning it is a few seconds to a minute, which is perfect because it gives time to the player to think about next move and read about various components.

Waiting time at the later stage will be from 3-20 minutes. Even though there are maybe 10 contracts that run in parallel, it is still very long

The graphics could be clearly be improved, especially a Hangar part (which we haven’t spoken about but it is where the player can see his drone). It may work for educational purpose but it will not win over any users

The game is free to play and can be found here:

https://www.buildlooplabs.com/drones/

Thanks for taking the time

Andreas