Some comments/history from my side:
When I joined the tumanako project in 2010 there was no working open code base, only fragments of some FOC and vehicle control code. They did operate a vehicle based on some proprietary ST library.
I joined the project and was introduced to libopenstm32 (nowadays called libopencm3). That was quite important because up to that point I ran all my tests using a linux PC and parallel-port-attached PWM card.
From now on I had a good foundation. As a matter of fact I developed the hardware and software architecture as it stands today all by myself (feel free to browse github logs and the mailing list archive) with the first running prototype
in early 2012 and a stable version
in middle 2013. I started offering the kit
in late 2013. All this was an after work-project while attending an 8h day time job.
Nowadays it's not only me working on the project but a whole crowd, some full time.
The (obvious) reason why I'm saying this is to contradict the "another 10 years" guesstimation. It did not take 10 years to get to a good result for ACIMs but only 3.
The development that has been going on since then was mostly motor-agnostic. Things like cruise control, charging and other tweaks.
And, of course, some words on the used MCU. There are better MCUs out nowadays with FPU, higher frequency, more pins, programmable logic etc. But running a motor, doing some comms, doing some digital/analog IO is not a very challenging job. The current software is nowhere near the limits of the stm32f103. Only 20% flash use, 30% sram use, 30% cpu use at the highest PWM frequency. That is, without any dubious optimizations like using assembler.
From a requirements point of view, what are the reasons to upgrade?