Fogbow

https://speedcamanywhere.com/

From an anonymous on SlashDot:

The app has also faced other difficulties in getting off the ground. Google initially refused to allow it on the Play Store, claiming it wasn't possible to estimate the speed of a passing vehicle using AI alone, however this claim was later proved wrong. An iOS version has also been developed, but it has not yet been approved for distribution by Apple, who have not given a reason for the delay. "We're not sure why they would block a useful piece of technology, something that could save people's lives," Sam said. [...] Currently, the app cannot lead to drivers receiving speeding tickets, as the algorithm is yet to be vetted by the Home Office, meaning it is not legally a speed camera, although drivers could still be charged with 'dangerous driving' offenses if their behavior is deemed to be sufficiently negligent. Sam says he hopes use of the app will alert police to speeding hotspots, encouraging them to take more action against dangerous driving.

https://tech.slashdot.org/story/22/04/1 ... s-speeding

One big problem with this is relative movement. If you are in a car going 30, and the other car is doing 40, the App would either read 70 (approaching) to 10 (both cars going the same way.)

Maths is hard, I'm thinking some vaguely competent programmer could take that into account?

John Thomas8 wrote: ↑Tue Apr 12, 2022 1:29 am Maths is hard, I'm thinking some vaguely competent programmer could take that into account?

Yeah. Your phone's GPS is much more accurate than your car's speedometer.

Suranis wrote: ↑Tue Apr 12, 2022 1:12 am One big problem with this is relative movement. If you are in a car going 30, and the other car is doing 40, the App would either read 70 (approaching) to 10 (both cars going the same way.)

Similarly, how can the app provide the other guys speed without providing yours as well? I'm not going to use it (is what I imagine some other driver would say).

A bit off topic, I've always felt the fines for speeding should go down as these systems come online and the odds of being caught go up. The embarrassment and the insurance premium hikes are still there and if you know you are going to get caught, you might not speed in the first place.

I've always thought speeding was an indication of mental defect. There's a sign, with legal repercussions, that denotes the speed you can travel on a given road. There's a dial or display on your dashboard that shows (mostly) how fast your vehicle is going. If one doesn't match the other or is over, you're liable to get a ticket.

It's not rocket surgery.

Well, on Kauai we drive at 50 mph. Everywhere, except when there are too many cars congregating on roads. I was stopped at one point, but I told the police officer that I live here.

Position relative to others is not considered ... normally ... someone someday somewhere may want that information. The phones are reporting autonomously without referencing each other. In fact, with the current state of technology, they have no idea that other are in the vicinity. These are just individual particles updating their positions as they progress blind to what others are doing. Likewise from the server side. Inbound packets are treated individually before being accumulated to derive inferred traffic conditions.

The speed calculation improves as their speed increases, but is very dependent upon gps reception. They require a clear view of sky with two or more birds visible for a 2D lock. Urban canyons are well documented to limit accuracy creating many snap-to-position errors. IE: a series of position updates will be missing making it appear someone has stopped. Then they send in a new update some distance away. The instantaneous speed calculation to traverse that distance will be impossibly wrong.

I've seen tow trucks, cop cars, and buses move around downtown like chess pieces completely ignoring physics on a CAD screen that only worked with literal data. Once hard limits and dithering are properly applied, a bus will not show up on the rooftop of the Hilton, and the police cars won't be in the river anymore. They will actually stay on the roads, but may appear twitchy. Likewise with forward speed. Limits and averaging are applied to compensate for lost packets. An instantaneous change from a normal and expected speed to something much greater will be massaged out until it can't be ignored anymore when a different threahold is reached.

Signed - someone who spent far too much time working with gps imprecision, inaccuracy, and fallibility, though will admit it has greatly improved in the last two decades. Just not ready for a robot in my pocket to tattle tale on me.

Mmmm, I was assuming the phones were taking a photo or the other car or something and calculating speed based on that.

Gps precision is also improving because base reference grids are being adjusted periodically.

Australian trains actually run on their tracks instead of some farmers field since they did a correction a few years back. It helps self driving cars too.

EDIT : The words I'm looking for are "geocentric datum", which were updated in 2020 and they'll need to do it again in a few years.

What is changing and why?

I'm not ready for that either. But, my truck (ute to you) has manual roll up/down windows too. It does however have an automatic transmission and power steering. I couldn't take clutching in commuter traffic anymore, and the upper body strength ain't what it used to be.

ETA: and air conditioning too. I know - turn in my man card. Sorry. Not going to happen.

https://www.treehugger.com/ford-app-not ... ns-6741086

Ford's New App Is an Exercise in Victim Blaming
"The dead pedestrian wasn't carrying a phone with the app."

Ford Motor Co. is developing a smartphone app that could warn car drivers of people walking or cycling, even if they are out of direct sight. According to the company, the app on the pedestrian's phone uses Bluetooth to broadcast their location. The vehicle calculates the crash risk and alerts drivers with graphics on the in-vehicle screen and audio alerts.

"Newer Ford vehicles already with Ford Co-Pilot360 Technology can detect and help warn drivers of pedestrians, bicyclists, scooter riders and others—and even apply brakes if drivers do not respond in time," said Jim Buczkowski, executive director of research and advanced engineering at Ford. "We are now exploring ways to expand vehicle sensing capability, for areas drivers cannot see, to help people drive even more confidently on roads increasingly shared by others using their two feet or two wheels."

Ford is part of the Vulnerable Road User Safety Consortium that was formed by vehicle, bicycle, ridesharing, and tech companies to find technological solutions to rising crashes with pedestrians, bicyclists, and others. They have all been working on this for years; we have seen this movie before.

For it to work properly, everyone has to have the app so the vehicles will know where everyone is. So first it might well be compulsory for people on bikes, and then at some point, people on foot if they want the traffic light to change for them. As Bez noted, "A V2X app on a smartphone is easy enough, but people will still try to roam freely. So a jaywalking law will cover off those tricky incidents where cars fail to detect people in unexpected places: if the worst happens, at least the vehicle maker won't be liable. Cross at the V2X-enabled crossing, or on your head be it."

Cool, this technology is critical for me to be able to watch a movie while I'm supposed to be driving.

/ffs

...and all of the great YouTubes!

These days an anonymous GPS position (like from a cell phone) is only accurate to +/- 30 feet. In the past when spoofing was turned on it was +/- 100 feet.

With the Wide Area Augmentation System (WAAS) you can get positions within +/- 1 meter. That was developed by the FAA to obtain accuracies sufficient for the commercial airline industry. The Australian train system may have developed a similar system, with fixed reference stations.

In my land surveying practice, survey-grade GPS equipment can attain accuracies of +/- 1 centimeter down to a a couple of millimeters. My equipment is a wee bit more expensive than my cell phone. I use a rover made by Leica Geosystems and it is approx. $35000 for the equipment and software. I pay a monthly subscription (~$200) to access via the internet the local GNSS reference stations. Without the subscription, my fancy GNSS rover is only accurate to +/- 30 feet.

BTW....GPS (Global Positioning System) is one of several GNSS (Global Navigation Satellite System) constellations. My survey equipment receives signals from the GPS, Glonass (Russian), Galileo (European) and BeiDou (Chinese) constellations to obtain centimeter-level positional accuracies within 2 or 3 seconds. That is because my GNSS receiver usually receives signals from 25 or more satellites to quickly determine its position.

ETA: My survey gear also has a terrestrial camera which for example allows me to walk around something I need to survey. Back at the office, my software processes the georeferenced images to create 3D point clouds of the feature(s) I am mapping.

COOOOLLLLL!!!!!!

I just drove down the main drive by Northwestern University today. Any sensing system would have a meltdown from all the people and apps to identify, give up and go have a beer.

They added a separated bike lane at some point. My peripheral vision and ADD was causing me some difficulty as they would be flying down that lane and I would automatically react, despite there being a barrier between us.