First of all, sorry for the long post, but there is a lot to be said here.
I am a researcher from a Brazilian University. My team and I work with the development of online machine learning algorithms applied to car related scenarios. For the development of this research, we investigated a number of OBD dongles, including yours.
The thing is, we were so optimistic with your hardware that we bought not only one, but two devices (with the possibility to buy even more further down the line). However, I can say that the journey so far has been very frustrating and time-consuming. So far, it is not showing promising results.
We bought the devices intending to use it with our own Raspbian image and searched the community to check on whether that was possible. We found the following topics that corroborated with our intentions, some even with positive responses from your own team, saying it was as simple as “remove the SD card from the device and flash whatever you want to it”.
On the other hand, on some topics, you guys say it’s not possible as you would “need to create an installable package with all the low-level power management services”.
So yeah, there are some inconsistencies.
But regardless of that, we decided to try it either way and, to our surprise, it did not work at all as the device kept shutting down every 10 minutes (presumably due to the lack of the power management services) and didn’t even let us access the OBD port.
After that, we reflashed the AutoPi-Core image and tried to modify it to our needs. No luck there either. We realized that to get access to the port we would need to stop the salt-minion, but that leads to the device shutting off every 10 minutes, so we were back at step 0. We even opened some topics here on the community that only confirmed our fears:
As a last resort, we tried working with the provided dashboard and cloud, even if that was not ideal, we were not ready to give up yet. As you can imagine, unfortunately, that didn’t work either. We could not get access to the data locally on the device and the cloud loggers never were recorded at consistent time intervals (even as we set them on the dashboard). On top of that, there is no API endpoint to get all the data recorded for one trip, as stated by Malte’s answer to the following post:
So, if the data is not collected at consistent intervals, leading to some sensors having more recorded data than others for the same time period, and all the data have slightly different timestamps how are you supposed to merge this data to perform even the most basic analysis? We couldn’t, at least not easily and reliably without using aggregation methods.
So yeah, as you can see, I’m out of options here and will most likely have to try other solutions on the market.
This situation has set our research back several months by now. Surely, with the time we invested trying to figure out your platform we could have developed an equivalent in-house solution. That being said you can picture our frustration, as we would very much prefer to work the AutoPi.
Thank you and I would appreciate any response or feedback you guys may have,