From my narrow experience as an R&D engineer in the auto sector:
You need to understand the duty cycle of the platform you're developing for (ex. Tires for a vehicle). A garbage truck is going to have a different loading cycle from an EV or a CAT morning truck. If you know the right conditions your product will fail on durability.
Fleets are ok with collecting this data for R&D as long as you don't bother them with extra work and give them useful analytics. Data collection devices need to be easy to install, collect for a few weeks and remove.
You can also use this as an R&D platform to develop IoT data-enabled products like predictive maintenance, route analytics, insurance and warranty claims etc. ML has a lot of potential here but you need to collect data.
Eventually when you have an IoT product you can get a vendor to make an optimized ASIC to collect and process exactly what signals you want.
There are many more applications in the auto industry for these kind of niche products.
There's an existing product called the SlamStick that's used to record high-shock/high-vibration environments. They released a paper on using one to find the source of a troubling vibration in a Navy helicopter. I came across them when doing research for a customer in the Vehicle Dynamics space.
It's one of those little niches that no one knows about and seems to support quite a few small companies.
