DOOM DRONE ALGO

• 

  Katmai Computing

Inspiration I am working to build my own quantum computer that I can hold in my hand, and it was fun to try solving military problems with different aspects of the quantum ecosystem.

What it does

For 1-1 drone swarms finding the optimal path to destruction and returning back to base in the most efficient way possible is similar to the quantum (TSP) Traveling salesman problem. It is actually a bit harder because the TSP has fixed points, Drones are moving and potentially attempting to evade detection. Trying to alleviate NP hard problems before swarms are large enough to overwhelm systems.

How we built it

Code, with help from AI. DeDrone longitude and latitude data was valuable. I focused on DJI, becuse they are some of the best and most dangerous in my opinion.

Challenges we ran into

AWS account didn't have appropriate privileges to access the Amazon Braket quantum hardware, so started over. Tried IBM, but the sign-in page was bugged and I couldn't access the cloud through them either.

I downloaded and started using Q# but found it hard to compile without a paid Azure cloud platform, before making the final move to using Google Colab and PennyLane, an open-source quantum ML-based software.

Accomplishments that we're proud of

Made a reasonable MVP for showing how a QAOA can be used to determine optimal pathing for drones. Learned a lot about Amazon Braket, Q#, by scraping all of it. Cleaning data in Jupyter notebook, pulling to Colab and with a fair amount of help from GPT.4, PI-AI, LLM studio (using llama 7B) and Colab's on-board AI, along with VS Code Copilot, covered a large amount of coding ecosystems and still produced an explainable MVP.

What we learned

The cloud is great if you pay, I am doing this for free and it can cost 2500 for an hour using an Amazon Braket based quantum computer. If you don't pay for AWS you don't get customer service and Google Colab + open source code tends to work better. Quantum hardware is restricted by region, AWS's 256 qubit system is based in Boston so I couldn't use it. Q# is ok but can be difficult to compile and use on macOS and Colab is still great for learning and Jupyter lab is still great for data cleaning. (crashed the colab simulating 30 qubits, dropped it to 26)

What's next for DOOM DRONE ALGO

Build an NMR based quantum computer for 999 USD, as part of Katmai Computing (my startup) and launch a software program that doesn't suck or milk people for money. The goal is to put a Qubit on a circuit board at room temperature. Quantum offers exponential compute.

Built With

• colab
• dedrone
• jupyter
• pennylane
• python