Richard's Blog
Project start
A note on realism. The story below is freely invented, and I have deliberately dramatised how the project began. IGNIS-8 is a hobby project with a real reference point — working around battery production — but nothing more. There was no Saturday phone call and no swollen pack in a safety cage; there was an idea, a spreadsheet and a lot of reading. Read the narrative as a story; the technical posts in this series are the record.
Key points. On that day there was nothing but an idea: no vehicle, no parts, no code, not even a plan beyond a paragraph. IGNIS-8 is a hobby project in the planning phase, with a real reference to battery production and nothing more. The prototype will come together slowly over the next months.
The question that started it#
It was a Saturday, and somewhere a battery pack had decided to be the interesting part of someone's shift. By the time I heard about it, it was already over — contained, cooled, no injuries — but the pack was sitting in a cage with a swollen cell, and the question put to me was short: can you look at it?
Not "can you fix it." Look. Look closely enough to decide whether the pack is scrap, whether the cells around the failed one are damaged, whether it is safe to move. On such a floor the usual answer is a long pole, a thermal camera on a tripod, and a careful walk. Nobody wanted to walk.
Lithium-ion cells fail fast when they fail: a cell can release enough energy to take its neighbours with it. But most of the time a pack does not run away — it swells, it heats, it vents a little, and it sits there undecided. In that window there is real information to gather: where exactly the heat is, how fast it is rising, whether the neighbouring cells are following. That decides whether the rest of the pack is salvaged or quarantined.
A person cannot gather that safely. A fixed camera cannot move to the other side. What would be needed is something that approaches, looks, listens, measures — and gets back out — while the operator stays at a distance.
The brief#
That evening I wrote the idea down, because an idea that lives only in the head is a mood, not a plan. The brief was short and surprisingly hard: approach and inspect damaged packs from a safe distance; deliver a live picture — thermal and optical, plus the sound of the pack; be small enough to move around a production floor; be controlled from a handheld terminal; fail in the direction of stopping; and be boring to operate, so the operator's attention stays on the pack.
That day, nothing existed#
It is worth being precise about the state on that day. There was no vehicle, no parts on a bench, no motors, no camera, no code, not even a folder of notes. The entire inventory was a paragraph and a head full of questions. Everything else — every axle, every sensor, every line of code — was a decision that had to be made before it could exist, and none of those decisions had been made yet. The plan was to let the idea grow slowly against reality, over the next months. No prototype in a week, no promises.
The idea, roughly shaped#
Three things were clear from the start.
A ground robot, not a drone. A drone sees the top of a pack; the interesting heat in a battery is on the sides and underneath. It stirs up air that confuses a thermal image and runs out of battery at the worst moments. A ground platform can carry a camera at the right height, hold position indefinitely, and move on the floor — where the packs live.
All wheels steered. If every wheel can steer, the vehicle can drive straight, crab sideways, rotate on the spot, or any combination — exactly what you need to approach a rack, sidestep a pallet, and leave the way you came. Eight wheels, on four axles, also spread the weight so a heavy robot does not dent the floor panels battery halls like to use.
Thermal first. The primary instrument is the thermal camera — the image that shows a cell two degrees warmer than its neighbours, or a terminal already heating. The operator looks at that first, and everything else is a dial.
The same evening, two more principles were set, and they never changed. Stopping is instant and easy; restarting is slow and deliberate. And the operator's screen and the vehicle must always tell the same story, so a warning on the display and a light on the vehicle can never disagree.
The name#
IGNIS — fire. The number 8 for the wheels. It sounded like a designation from a manual nobody reads until they need it, which felt right. Named with a straight face, on the very day it was still only an idea.
Closing#
That day had exactly one deliverable: the idea. No prototype, no parts, no code, not even a repository. Just a paragraph, a name, and the decision to let it grow against reality over the next months. The sentence that existed on day one was: yes, we can look at it. Everything it needs to be true is the work of the months after it.