E-scooter startup Spin apparently uses RasPi 4s inside their scooters
abolish.social
abolish.social
A benture capital subsidized micro mobility startup pulls out of a major city ostensibly because threshold for potential profit has been crossed and they’ve determined they cannot adjust their pricing to get the right numbers on their spreadsheet (note: this likely part of the reason why so many of these companies have trouble converting people to yearly plans, why bother if your market can be dropped with such swift indifference?). After pulling out of the market they leave their trash, that were assets a few days previous, scattered amongst the city as technological blight strewn across the landscape, left to rust. When opportunists go to crack them open, they find a raspberry pi, an SBC created for educational and hobby purposes but has been infamously out of stock because larger companies want to vacuum them all up to use in their own products. Then you wonder where all the engineering cost for these scooters went. After the presumably thousands of hours of labor that went into designing this, they went with a consumer grade, off the shelf product for an application that would have required a fraction of the power it was capable of? Not to mention that Spin can be identified as one of offenders of why the raspberry pi is so goddamn hard to find.
This all makes me irrationally irritated.
What a strange complaint. RPIs are out of stock because they're useful - this company found a use for them. Really the issue isn't one of who's buying, but rather an issue of the Foundation not making enough of their wildly successful product. Seems like a high-quality problem.
I can't imagine the Foundation being up in arms over someone finding a use for their product. The more users, the better the economy of scale, the cheaper the product is for everyone.
> Then you wonder where all the engineering cost for these scooters went.
Well, into the backend, the integration, the mechanical engineering - the myriad other things that mark the difference between a fun thing you made at home and a product you sell to the public.
> After the presumably thousands of hours of labor that went into designing this, they went with a consumer grade, off the shelf product for an application that would have required a fraction of the power it was capable of?
Again, economics of scale. One product that's more capable than any one person needs - but has a bigger audience - is likely cheaper than a niche one that's 'right-sized.'
Your remote control doesn't need a Cortex M0 but they're cheaper than an 8051 now.
> Not to mention that Spin can be identified as one of offenders of why the raspberry pi is so goddamn hard to find.
So can anyone who hit 'add to cart.' Especially since they only have like 500-1000 scooters per city in which they operate. That's not exactly Apple-scale orders.
The issue is supply, not demand.
While I sympathize with this sentiment, it should be noted that their use case could have been most likely accomplished with something like an ESP32, which is only slightly more difficult to work with but boasts a bulk price of $1-3 a pop.
Using a Raspberry Pi 4 for this is a bit like flying a private jet to get to the grocery store.
esp32 is embedded enough that you have to develop for it like you develop embedded software - you need to do things like think about how much ram you're using that you just don't have to on raspi.
Solution: Put a way overpowered computer in the product so they can program it.
Ridiculous problem with a ridiculous solution. Are they allergic to any level of correctness? Sometimes I'm ashamed to be in this industry.
The only product that matters to a customer is one that exists, and if this approach saves you money, lets you prove product market fit, or allows the product to keep existing longer - isn't that more important than the fact the Pi has more parts than you need?
What if the jet ride cost 25% less than the drive? What if they're only using the jet to prove demand exists on the route? Optimization only makes sense in the face of scale, but you need to get to the scale first.
The unnecessarily wasteful hardware is just one of the (ultimately more benign) side effects of that.
This is exactly what happens in other industries as well, it is not a software only problem.
You really don't. And in most cases, making it harder means that you're probably making it buggier.
Pick any random set of Software Engineers from any country. I guarantee you that you are more likely to find people who can develop fast on an RPi than on an ESP32.
Now if you're a startup needing to build very fast and iterate quickly, it's obvious you're going to use the RPi4. Then, once you've grown, you can make the system more efficient, port it, etc. It'll be a growing pain, but a sustained one. And without being completely immersed in ridiculous amounts of VC-funding (hopefully)
I'm the CTO of a company who helps startups go from the idea to the scale-up phase. We helped build a tiny startup to one of the most promising startups in europe using this idea. We're very proud of what we've achieved. Could we have built something with an ESP32? Sure, and we might have even made it in the same timeframe given the amazing team we have, but it was less risky to bet on the RPi than the ESP32.
In fact, the first iteration had an ESP32 and we were brought in because they needed to move faster and it was becoming hard to develop for and iterate at their pace. Again: perhaps we could have built the same with the ESP32, but it's a matter of risk analysis!
Sometimes I feel like people talk about these things in a kind of vacuum of perfection and utopia, outside of the real world. In the real world things get messy and move very fast. Supply-chain issues arise, hardware has to be swapped, people move cities and quit their jobs. Where we have excelled as a partner is in dealing with very very very fast change and using an ESP32 for that would have made things harder. Now instead of a dead product there's thousands of the thing we helped build out there, and growing. And we're finally moving away from the RPi4.
Technology is actually a very small part of what we do. Most of it is creating the logistic and the "system" around the technology to get everything to work. Sometimes, as engineers, we forget that.
We also don't know anything about the quality of the software running on those devices. I've seen some awful crimes against computers committed in embedded software - and I've seen some pretty well written software on more capable machines. And v.v. of course.
Then again, I wouldn't say that it is way more difficult, and I would wager that most of those who have the skills to build the brain of an electric scooter with a Pi 4 could learn how to port their application over to the ESP32 platform within at most a week. Even if you went with a more pessimistic estimate than that, the enormous price difference between the two boards will always make up for the cost of engineering at scale.
Consider a few other advantages of using the ESP32 here: faster replacement times of faulty hardware, and a way smaller attack surface. How much do you want to bet that those Pi 4s have "root" set as their username and password?
Driving the 3-phase motor yes, that can (and is usually) done with a dedicated real-time MCU/ASIC. But the interface with the backend, the actual brains? You need to deal with GPS, authentication (unlock via app/bluetooth/...), ride logging, debug data. And at that point an ESP32 or heaven forbid an Arduino won't cut it either, and you don't want to spend hundreds of hours on wrangling with some custom higher-power embedded board because that is real nasty to get right. So you use an RPi because the extremely broad user base has already solved every issue you might have run into.
Besides, the innards of e-scooters can be had as whitelabel solutions, no one engineers these themselves. All people do is design the outer shell based on a design kit, do whatever needs to be done to pass legal certifications, and integrate some sort of brains if it's for a fleet.
Why? ATMega board is surely a questionable fit, but ESP32, or nRF52, or some STM32 board with Bluetooth all should be fully capable of doing the basics. I can't imagine those scooters do some fancy signal processing on-board or handle some particularly heavy traffic. ESP32 is capable of wrapping and transmitting a decent video stream (and even do some basic detection while it's at it) - surely it's much more data than some route information and various on-board telemetry. GPS and cellular are just talking over some bus to a module (pretty much the same as with rPI4), Bluetooth is available and not resource-expensive at all. So is talking to whatever relays run or block the motors - I imagine that's probably just a bunch of GPIOs. All that remains is need for some RAM to store/buffer the traces (ESP32, or one can throw in some external storage) and a basic TLS and MQTT implementation (or whatever tech would make most sense there) for the command and status message queue.
And most importantly, average ESP32 energy consumption is way smaller than rPI4's (it's pretty hot even when it idles). While I understand that scooters have large batteries, it still matters.
I've shipped ESP32-based boards and written firmware for it that really tries to cram in as many features as possible, but you do hit limits. While I almost certainly wouldn't make the choice of shipping a high-volume product with an actual Raspberry Pi in it, I do see why trying to cram all of this into an MCU is tough — especially if you're trying to get to market quickly.
Perhaps this would have been a good place to use the RPi CM4 though?
The RaspberryPi is not a real time computer, it uses a conventional OS that is inherently not real time. Regardless, you don’t need anything RealTime(tm) because the throttle produces a signal to the BLDC controller, and the MCU only turns on/enables its use.
The benefits from certification are minimal of using an off the shelf product, as you are still required to have the custom stuff l certified, too. They are likely using a cellular module that carry’s its own cert which is the biggest hurdle anyway.
My point is that using a hobby-grade Linux system designed for a much different use-case shows clear deficiency in the skillset of the team, management, or hiring, this is very obviously not the right tool for the job. What makes this upsetting is the fact that the market this was made for (and still is exerting demand pressure) cannot get their hands on it because of the pressures in the industry that values something that just works over correctness. If they would have used the Rpi compute modules, it would have given them a lot of flexibility in what they chose to include while reducing costs all around. This whole thing is a masterclass in gross incompetence characterized by waste and poor engineering.
A me-too copycat startup like Spin who got kicked out of Seattle a year ago ( https://www.seattlebikeblog.com/2022/05/13/city-announces-ne... ) doing it this way doesn't seem like a great endorsement to me.
You have to spend $$$ up front on thousands of vehicles that cost hundreds of dollars each, and your marginal revenue per ride ain't great. Putting a dent in your vehicle theft rate becomes a big deal fast.
Given that it's a routine sport in many cities, particularly those with bodies of water, to damage or destroy the scooters, it's imperative to make them as cheap as possible and as easy to repair as possible. Both conflict with stuff like custom boards or barriers to opening them up.
The biggest theft use I see locally these days of the modern Limes and Birds is using the batteries for charging shit at homeless encampments.
But there were a million copycat companies like Spin who launched with off the shelf stuff after Bird and Lime who seemed to copy the "light money on fire theft-wise" problems of off the shelf stuff.
Okay, I'm German, so no idea about homeless encampments because we simply don't have these in large scale, but how does that even work? Like okay, you rip out the batteries, but then what? You'd still need some sort of voltage regulator just to power a device, and that hasn't even included recharging the batteries.
But it's a lower-demand type of theft than "spend 30 bucks on alibaba and have a free scooter forever thanks to the off-the-shelf-nature of the first rental scooters" that people in the US market were going nuts with in 2018/19. It's a bit more wild west than Germany. ;)
I've no idea about those scooters, but I won't be surprised if it's all behind some plastic panel held by few screws.
Besides, why would you want to steal them? You can grab new ones for a couple hundred dollars. Not worth the risk.
This was already all over the internet 4 years ago when Lime/Bird got started and were using off the shelf vehicles.
AFAIK, somehow going after scooter thefts in cities that often already had a grumpy relationship with them for GTA never really happened. And come on, there's a LOT of people for whom "a couple hundred dollars" is a plenty big enough score. Why would anyone steal a bike in that case either?
Given how little prosecution there is around bike theft, I doubt any police department or prosecutor is interested in spending the resources to prosecute the theft of an even cheaper mobility device.
Another angle is paying contractors. I doubt that you need a person-year to make an electric scooter, so that pushes the break-even point down a few units. But, your Python engineer can also make the backend when you're done with the hardware. It really depends on how expensive your recuriting/hiring/onboarding pipeline is. I am sure many people you run into will say that they know how to design e-scooter control systems, but you have to find the ones who are not lying. That's a cost.
So all in all, it seems like they spent their investors money wisely. You do the cost reduction when you want to expand, not for the "yeah, this business can't work" test phase. It seems like that's what they did; the business doesn't work in Seattle at any scooter BOM cost, so it's not a business.
The dumb part is not paying someone to round up the scooters and eBay the electronics and batteries, though. They could easily make a lot of their money back, probably even selling the used Pis for more than they paid!
It was never meant just for education and hobbyists, Raspberry Pi Foundation absolutely supports the commercial use case, and their newer products like CM4 even more so.
I’d wager you never tried to launch an embedded Linux-powered product, because if you did you’d quickly realize that building custom images from obscure Linux and U-Boot forks is just not fun. Raspberry Pi solved that for everybody, this is why it’s popular among startups even though it’s not the best hardware around spec- and cost-wise.
I'd wager most people go with a RTOS
Is it? $35 seems pretty cheap for an SBC that does what you need it to, with you having to do zero of the engineering work yourself. I think most industrial SBCs cost way more than $35.
> ... requires too much power ...
Pi 4 consumes 2-6W. [1] Scooter batteries are 400-600Wh based on a cursory glance, so we're talking up to two weeks of idle time. The scooter only goes 12-15 miles, and at 10mph, that's going to be a significantly bigger issue.
> ... and has a vastly higher rate of problems than many other off the shelf, far cheaper and much more widely available products.
Which products would you recommend, and also, do you have any data on the failure rates of the Pi? I'm curious.
Is it prefect? No, probably not. But if you only need a few thousand units, and you're trying to prove out the product market fit, why isn't this a great solution? Think of all the millions in capital they didn't waste building a custom solution. Personally, I respect the hustle.
[1] https://www.pidramble.com/wiki/benchmarks/power-consumption
I really don’t think you have any idea of not the costs of these things, every penny counts with hardware products because any amount increases the marginal cost by at least 1.5x, vs software where it is so low it is effectively zero. I can get a very beefy STM32 for sub $10 at low quantities (sub $5 at > 1000) and toss all the passives and connectors needed for an extra $2-$4 the PCB may cost $1-$5, and you might be at $10-$15 depending on volume. You also can pre buy components that are known run out or design the board to accept different MCUs incase of shortages. With the RPi, you are totally SOL if you cannot find them, with your company being at the tail end of a, now, know volatile supply chain with a company you cannot get guaranteed fulfillment contracts with.
Your assumption about power is also very divorced from the reality of the situation, battery performance is not so easily calculated when your battery is not in clean room conditions, these scooters are subjected to extreme conditions and abuse, all of which can affect the battery. Additionally, It is disingenuous to consider only the operating parameters because it does run a full Linux system and software you load into it creates non-deterministic (at least with current models) situations where new software can skyrocket the power consumption. They also ideally have a separate battery from the drive motors so it can call for help and a tech can collect the scooter for maintenance.
using a very cheap off the shelf linux board with great software support is a completely reasonable choice for corporations, too
I am not formally trained in electronics, but I managed to learn enough in the last year to design and manufacture my own PCBs for respiratory breath analysis with Atmel microprocessors. Essentially, they're like Arduinos but I also attached an off-the-shelf SD card reader and a bluetooth module. After that experience, I learned enough that I can simply build virtually everything into the board itself if I were to try again (which I'm still sort of in the process of doing).
Point is, the fact they're just using Raspberry Pis tells me that they barely know what they're doing. No reason why they can't design their own hardware to do the same thing but consume way less energy. What else do they truly need other than a basic microcontroller and a GSM module? GPS maybe?
Make-versus-buy decisions abound in product development. If you aren't a full stack electronics and firmware designer, then you have to manage the design project with its uncertainties, including schedule. A buy-in gets you moving right away, to address other project issues that require a prototype. You can always design it out later.
Dealing with complex chips often involves discovering bugs in the docs, and a second board spin to fix them. Then you'd better have a good scope and some real knowledge.
Despite its hobby orientation, the RPi is legit.
Not to mention the time penalty - as a rule of thumb it takes about minimum 6m average 12m to go from concept to physical units in people's hands at scale. These Pi's are shipping today. As a startup, that's hard to pass on.
A volume like 10k puts you in a dreaded "in between" zone. You're not making a one-off widget that a customer is willing to pay an unlimited amount for because it solves a unique problem. And it's not enough volume to make something that just rolls off an assembly line. All of the simplistic mental models for how to proceed, don't work.
I work in such an industry, making specialized equipment. Every component is an expensive buy-in. It's ridiculously hard to find cost savings every component is up against the same math. Then again the competition faces the same problem.
How did you learn? I'm not formally trained in electronics either and would like to be able to do this.
> Computing for everybody
> From industries large and small, to the kitchen table tinkerer, to the classroom coder, we make computing accessible and affordable for everybody.
I would say that's exactly what we're seeing here.
If I wasn't personally involved in building something I'd generally assume the engineers were reasonable people making reasonable decisions.
Nothing in the source suggests these are “strewn across the landscape”. Your comment is basically made up outrage bait. I live in Seattle and walk a few miles a day and haven’t seen any abandoned Spin scooters.
The availability of something everybody uses is entirely up to Broadcom right now. The pi foundation fixed the problem on the 2040, and maybe they will fix on the mainline some day, but the situation isn't optimal right now.
Now, I would agree that a Rpi 4 in an e-scooter seems overkill but if they are small and cheap enough, and easy to work with then why would you go to the trouble of finding a less powerful solution?
Because making a lot of money is more satisfying than supporting education and hobbyists. —- ftfy
You should get a job at the New York Times
> This all makes me irrationally...
I can tell
It's also worth noting that the rapid selling out of hobbyist inventory happens in part because much of the supply is earmarked for industrial customers; we don't actually know if that would still happen if all Pis were sold directly to hobbyists.
I'm still looking for 3 pi zero 2s and trying to be fiscally responsible and not get scalped or further contributing to this mess but wow it's been way too long and it's maddening.
I'm not a company, but I've spent a ton of money on pis (including professionally). I was a big champion and supporter convincing everyone I knew to use them for their projects. Now I'm just mad and looking for any alternative I can find. I know I'm not the only one. How can this not have negative long term consequences??
I know they are just a business at the end of the day but I was all in for years. Rational or not I feel completely betrayed.
When the supply issues happened, they were concerned about not being able to manufacture the devices. That's a multi million dollar investment at stake not to mention the employees and customers that would be impacted. Luckily they figure out a supply.
Compared to that, my as a hobbyist having to be patient to get another pi is completely incosequential.
Between "retool" and "hobbyist wait" one seems worse.
You may remember car manufacturers had to shut down assembly lines due to chip shortages in the last few years even for non-critical chips. Redesigning around missing components is non trivial though obviously doable
And in one sentence you've nailed its value proposition perfectly. That's what is so special about it.
I think the argument is that there shouldn't be a Pi at the core
In many cases, meeting a development schedule is far, far more important than cost or other constraints. I will happily throw a $25 ESP32 module at something that a $0.50 Atmel chip could handle, if I only need to build one unit. The $25 is a bargain compared to the amount of work I'd have to do to get the Atmel chip up and running.
That's what Spin didn't like - competition.[3] "Open permit markets are places where multiple scooter companies can run businesses, with no caps on fleet sizes. Bear (Spin CEO) said they "create an uncertain operating environment" with "race to the bottom pricing." Apparently, Spin gets double the revenue in (non-competitive markets) compared to locations with a free-for-all market.
Er, that's how capitalism is supposed to work. They can't handle a free market.
[1] https://dallascityhall.com/departments/transportation/Docume...
[2] https://www.seattle.gov/transportation/projects-and-programs...
[3] https://www.engadget.com/ford-e-scooter-company-spin-leaving...
Really context pointing toward intent is king here.
They responsibly and quickly gathered up (almost) all of the scooters in one weekend.
Maybe 0.1% of the scooters remain, likely due to the battery being dead and tracking being difficult.
So in my experience, I believe they are acting in good faith when exiting a city.
Not that long ago it seemed the tradeoff was either use a cheap but unreliable hobby platform (RPi or Arduino), or a stable but expensive (and usually somewhat custom) platform that involves picking the right components, board design, etc. Espressif sounds like they might be filling this gap, but it’s been a long time since I’ve looked seriously at any of this.
It’s cheaper to buy Pis and make your startup than it is to make something yourself, which is equal parts impressive and sad.
I think this might have been an early dev security unit because all of the ones I saw had a much smaller single pcb and iirc was a third party all-in-one board or custom board sourced later. Also, this looks like an ES2 and Spin switched to ESMaxes early on.