Why I’m Doing This

I have always enjoyed the challenge of working with constrained environments or limited resources. My technology life journey started in the era of 8-bit home computers, when development was as much an electronics project as a programming exercise, and pushing the limits of a 6502 processor to connect with others on BBS was some weird mix of BASIC and machine language typed in from the back of a magazine. As computer technology moved along it always seemed to me that, rather than becoming faster (though in fairness, it has), systems become more bloated and wasteful of the considerable processing resources available to it. Fast forward to our current state, and AI datacentres are gobbling up tons of global resources, energy, and causing a fair amount of pollution along the way.

There is another way. Not a replacement, but a different research direction that centers on smaller scale embedded, low-power implementations that revisit the sort of constraints I experienced early on. I first came across the concepts of swarm intelligence and tinyML as part of my Master’s research though I did not explore it at the time. I just sort of took a mental note and moved on. Lately, however, as I see larger implementation AI frameworks and paradigms taking centre stage - despite serious questions around the environmental sustainability of these approaches - I have found myself revisiting these concepts more and more.

The Challenge

The research quesion I landed on at a high level was whether I can build robust, scalable machine learning systems using swarms of resource-constrained devices. I want this to combine my interests in constrained computing and programming embedded systems, mesh networking, and electronics; with a desire to reduce the impact on the environment and maybe even self-powering systems. The initial goal is the creation of a wildlife monitoring using autonomous sensor networks, but I am giving myself the permission to pivot as necessary. I have little doubt a year from now I will look at this challenge and smile at how much I did not know in embarking on this journey, but it is as good a place to start as any.

The Plan

As I starting mapping out how this project would work, I sort of split it off into a year-long project that divides things up into machine learning, programming, electronics, and a lot of math. Given that I will be spending a lot of time thinking about swarm technology, I have included some biology resources that focus on studies I hope will be relevant. There is a higher-level breakdown of the plan, goals, and timelines on the About page of this Blog: https://sharris416.github.io/about/ and if you really want to follow the journey I will document everything in the research repository: https://github.com/sharris416/tinyml-study

Finally, feel free to join me! The About pages lists a number of ways I am engaging with the community, and I will add to that as I go forward. Life is a team sport, and if you are on a similar path or just interested in the work here, I look forward to connecting. Week One has been largely setup, aside from a reading about bees and a Linear Algebra refresh, so now I’m off to begin Week Two. Thanks for dropping by!