Korina

It's an AI that runs on a Raspberry Pi computer. Looks pretty interesting. https://sf.streetsblog.org/2022/12/1...fight/#new_tab

greatbasin

The cellular providers have all the data. Every single one of their customers is tracked by GPS and the record of those tracks is stored indefinitely. They aggregate that data and sell it through consultancy services. A query like "how many people traverse this roadway," is a simple thing. Of course, it costs something, but the data is there. From what I've heard, they won't sell access to the data outright, but sell consultancy services so the questions the data answers are mediated for their fees. It is frequently applied for issues like traffic and parking. If a proposal is going to bring a lot more traffic into an area, there's analysis about where all those cars will park. Planners need to know things like how much of the traffic is parking and for how long.

The RPi solution is cute and cheap, but scale it out to the whole city or state and it's impractical. The GPS tracks of every cell phone are already in the database sitting at the query prompt.

FBinNY 

Yes the data is there, but it's sometimes difficult to tease out critical details.

Discounting idiots like myself who keep location turned off, the problem is that the data only says where the cell phone is, which isn't always the whole story, because we don't know how it got there or what kine of vehicle (if any) is involved.

While it might be possible to tease it out by speed, ie 3mph implies walking, 30mph is driving, etc., it's lots of work and not reliable in many instances. For example, in dense metropolitan traffic, my bicycling speed is closely matched to general traffic. In parts of Manhattan, so is my walking speed. If I'm moving at 25mph in the burbs, am I driving, riding as a car passenger, on a bus, or a strong bicyclist?

Businesses measuring traffic for marketing purposes don't have a reason to care, and urban planners also don't care about small skews. But if one's purpose is to analyze vehicle use, the raw data isn't nearly as usable as implied.

greatbasin

Turning off location does not stop the GPS tracking. All that does is stop location tracking by software or app providers like Google. The cellular provider cannot not know where the phone is. Unless the phone is off, it's location is being tracked. Without location data, you could not connect to the nearest cell tower. But that brings up the point that the location tracking data is available from multiple sources. The cellular network providers like Sprint, Verizon, AT&T, T-Mobile, Spectrum, etc. and then also app providers like Apple, Google, Microsoft, Facebook, etc. Users can opt-out of location tracking for the latter, but not the former. But with many people not choosing to opt-out, the data is available from several sources. The data set doesn't need to be complete or perfect. Margins of error can be calculated and the data presented with measures of center and standard deviation. The raw data isn't what drives decisions anyway, but the summary metrics of the aggregate.

The street in question was closed to motor vehicle traffic, so that was sufficient to determine non-motorized traffic, pedestrians, bicycles, unicycles or whatever. Either the slow street was being used or not. To get more sophisiticated use data like to split bicycle from pedestrian traffic, it could be estimated by rate of speed as you say (available from GPS data) or just interpolated from sample data. It's a lot easier if a targeted survey only needs to determine a ratio than if it bears the entire burden of data collection.

FBinNY 

Thanks for that. FWIW, my reference to turning off location was simply an intro to thr real issue of teasing out vehicle type.

On that point we seem to agree that some kind of visual census would be necessary to estimate the mix ratios, after which info could be extrapolated from the totals.

In any case, I get grief (what if there's an emergency, etc.) because l generally don't take my phone, or keep it off when riding.

Now I have another excuse, so thanks again.

Korina

I'm not sure it's meant to be scaled up to large city or state; it appears to be a neighborhood scale solution. And not everyone carries a smart phone. Now if our COVID vaccination rates were higher we could easily be tracked.

greatbasin

I don't carry a smart phone. Even so, what is the portion of traffic that does not? I'd guess less than 1%. Even if you don't have a smart phone, virtually all late-model cars are equipped with GPS and cellular. Even if a subscription of any kind isn't activated, the radio is still contacting the cell tower. It isn't denied service until several more hops on the network, and even then it isn't denied emergency service. It's increasingly an "internet-of-things" and if you have anything, you're probably being tracked. We could say that the data missing from the people who don't have anything isn't as significant as the data from the things.