Project sites
24.0621
ZiptieAI overview
- Sites
- Motivation (for creating ZiptieAI)
- Phases (~6 phases of this project)
There are 4 main sites for the ZiptieAI project:
- This wiki. The most up-to-date info.
- Gdrive. Contains the core step-by-step (draft) documents (.docx) that provide the detailed info required to DIY.
- ZiptieAI webpage. A short overview page.
- Docs website. Drone/AI process documentation site. For now this is just a skeleton site (its out-of-date).
I had to change jobs a few years ago because of the DoD "vaccination" mandate. I then worked on 2 6-month contracts (both extended to a year) for 2 different companies where my work was slowly micromanaged into something I could not claim as my own work.
I needed a change. I started focusing on AI in October 2023. I was impressed with API's that made it possible to "easily" use some very complex algorithms (that worked amazingly well).
I was in Ukraine in late 2023 and was impressed by Ukrainian drone development. I started exploring drone simulation while in Ukraine. I assumed that building an actual drone (back in the USA) would be the easy part (wrong). The basic concept was:
-
Ziptie drones: Front-line Ukrainian soldiers often use zip ties to configure a drone package. My focus is on learning the components (HW and SW) that can quickly be "ziptied" together for any mission. Below: A zip tie drone on the cover of The Economist.
-
AI: Lost connection to the drone pilot? Carry on without him. Want to swarm the target? No problem, send an intelligent group of drones to do the job. You implement the AI routines on a companion computer (a PI4/5 or Nvidia Jetson mounted on the drone that sends flight commands to the drone flight controller). Below: AI object recognition running in a simulator (part 1 Total simulation shows how to this step-by-step).
But building/flying the drones has taken much longer than expected. Its really difficult to get things working unless you are really deep into the tech. AI is much more complex, but the AI APIs are well done. By contrast much of the drone tech is
- Open source with
- Chinese components
My activities for the first half of 2024 were were 80% drone and 20% AI.
I started working on ZiptieAI in December 2023. Great experience. But my hope has been to create something that will help me land a good job. That might be a challenge.
The author of the excellent "Intelligent drone simulation" video series (Total simulation) said (5 years ago) that he stopped for working with drones for a while because he ran out of money. Apparently there was not a lot of money to be made in open-source drones (drones were not a profession, just a hobby).
The reason for that is probably because most professional applications of drones require a lot of re-development of many of the open source components. There was a period during the early stages of Ukraine war where there was a big market for "ziptied" drones, but that period will soon be over. The Ukrainians have formed private companies that are doing what western companies did ~5 years ago, namely switching to more customized development.
Which leaves me where? Its my hope that (with ZiptieAI as a personal project that demonstrates my qualifications) I could land a job (preferably working with drones) as a
- Product manager/owner
- Tech writer
(just a rough plan; the future phases will undoubtedly differ from this plan)
I have reorganized the wiki probably once a month for the first half of 2024. I would find something in the internet I could get to work, document my steps, and as a result I understood one more piece of the puzzle that was the big picture concepts. And how to organize this wiki. It was a long process.
But as of June 2024, the wiki reorg was more or less complete. It was then I started to focus on phases. I came up with 3 main phases for the ZiptieAI project. Phase 1 is nearing completion right now (24.0621).
Phase 1 (epics 1-3): Basic build/fly/document at least 1 platform variant for epics 1-3 (2024.01-06)
What was originally the first part (phase 1) of this wiki (what I worked on in late 2023) is shown below in grey boxes (what is now epic 3 content). This happened because
- I could not get my original plan target systems to work.
- I understood better what a newcomer needs to focus on (for example, FPV simulators and actually flying FPV drone, NOT total simulation (SITL, Gazebo and ROS).
- Clean up the step-by-step directions in the Gdrive docx's (take what are sometimes basically messy lab notes and distill into clean docx's)
- Update concepts (you have to have hands-on experience to write the concepts)
- Clean up references (concepts, tech refs, etc)
