Using Metromile data to determine a car's gear ratios | Hacker News Reader

Using Metromile data to determine a car's gear ratios

whatsyourpvalue.wordpress.com

73 pointsjeffonabike23 comments

Great article! The gear ratio data would be useful to potentially derive more granular driver behavior model (acceleration) that is less mass dependent than accelerometer sensor. However, I do wonder if there is a way to determine the usage of the reverse gear itself. This data is currently not universally available via OBDII. The reverse gear data would be very useful since large amount of accidents are backing accidents.

From the data provided, it seems likely that OBD is reporting any speed while reversing as 0, so you could probably guess that when the engine speed is changing slowly (while above idle), and OBD reports vehicle speed as 0, the car is reversing.

This may be true - but even if the sensor is reading zero vehicle speed and greater than idling engine speed, it's difficult to assume these are in fact reverse and aren't just "engine revving" moments. Looking into a time-series breakdown of each moment - seeing if the engine speed is 'changing slowly' as you mention - would be key in properly identifying Reversing vs. Revving.

This is why I specified engine revving slowly; engines which are disconnected from the drivetrain tend to rev up very quickly. In addition, cars with automatic transmissions are a majority of the US fleet, and spend very little time in neutral.

> engines which are disconnected from the drivetrain tend to rev up very quickly

So do engines which are in reverse and being controlled by drivers who are imminently about to have an accident :)

This might also be specific to the Metromile hardware (OBDII interpreter). I just checked on my vehicle with STN OBDII interpreter reversing is showing non zero value (I increased the reporting frequency to 1Hz).

What does this look like for a car with continuous variable transmission, like a late model Nissan Altima?

Most cars with a CVT are configured from the factory to have "gears". Even though the transmission is continuously variable, it is made to quickly move between distinct ratios. This is mostly because with a true CVT acceleration is very smooth and difficult for the customer to perceive.

So the graph looks almost identical.

This is fascinating:

   This is mostly because with a true CVT acceleration is very smooth and difficult for the customer to perceive.

Can you provide any more insight or further reading? I am considering a CVT (Subaru XT), and smooth acceleration without jumps is precisely what I would be looking for (which I believe was provided by my Scooter with a CVT as well).

If companies are adding artificial jerks to a technology that is designer for smooth experience, I'm not sure whether i'd laugh or cry, but I feel I'd like to know more :-/

Thanks!

>> I'm not sure whether i'd laugh or cry

Exactly... for me it's the same with electric cars that make fake engine noises. But that one I grudgingly accept, in parking lots, for safety reasons.

I'm pretty OK with electric cars making fake engine noises. The Prius owner in my neighborhood would be semi regularly creaming people if they didn't have any auditory indication that he was about to come cruising out of the alley without slowing down to check for pedestrians.

It doesn't have to be engine noises, it could be playing some sort of crazy "woop woop woop" sound for all I care, but I'm guessing the owner would prefer it be something a bit less embarrassing.

My 2012 Subaru accelerates smoothly without jumps unless I switch it to 'manual' mode. Throttle position determines engine RPM and the CVT smoothly tracks acceleration, no artificial jerks. It was an eerie transition from shifting transmissions.

I wouldn't mind seeing similar graphs of CVT cars, where there wouldn't be distinct lines but patterns of typical driving behavior may still be obvious.

People have come to associate things like the kick of shifting gears or the sound of the engine revving very closely with acceleration. So much so that if they are missing, they infer that the car is less powerful. Accelerating with a CVT, the car has the option of placing the engine in the max torque range and adjusting the gearing as speed increases to keep it there. If you take your cues from the first set of behaviors, this might not give you the same sense of acceleration.

(I have a 2008 Altima)

I don't think this is accurate.

Most cars with a CVT don't have this but tout gearing as a special, unique feature. Often costing more than the same vehicle without.

Case in point: Subaru Forester. All but the most bare bones models come with a CVT. But it's only on the XT turbo charged model that you get a CVT with fake "gearing" and it is touted as a special add-on feature.

Other cars that come to mind with a CVT without gearing: the rest of the Subaru line, Nissan Versa, Nissan Sentra, etc.

I'm not sure I'd say most anymore. Ten years ago that was the case, but not anymore. People are just use to the idea of CVTs.

A few have the option for virtual gears or a manual shift mode, but default to normal CVT behavior. But there are several that have no option for shifting beyond normal and "low" gearing for engine braking, just like traditional automatics.

130 km/h is a pretty good clip. Does Metromile charge clients more for high speeds?

Their fee structure is a base rate plus a fixed rate per mile. From what I remember, those rates are determined by the typical demographic info + driving record/past claims. There are other companies that I believe try to adjust their rate based on driver behavior data, but metromile doesn't (at this time).

75 mph is about the average highway speed in LA - assuming you're not in stop-and-go traffic of course. Speed limits seem to be more of a recommendation than an absolute.

California prohibits pricing based on driving habits like speed. Other states do permit it and Progressive offers snapshot device which does what you describe - offer discounts for safe driving behavior.

Wow really, that is amazing and unsurprising at once. I can only imagine the debate that led to this regulation.

Actually a pretty good way to understand manual transmission. Very cool.

It would be neat if you could enter a tire size (e.g. 165-65R13) and estimate the overall diameter from that.

Extra credit: figure out a plausible count of gear teeth for each ratio.

Neat

Reply on news.ycombinator.com