Hacker News new | past | comments | ask | show | jobs | submit login
Tesla Owners Can Edit Maps to Improve Summon Routes (teslamotorsclub.com)
66 points by AlexTrask on Nov 11, 2019 | hide | past | favorite | 34 comments

This is great news. OSM might not have gotten as much public awareness like Wikipedia, but it deserves it. Having a global map data source, which isn't only freely available, but to which everyone can contribute, is a great piece of information infrastructure. The level of map details, especially off the roads open for car traffic, is excellent around the place where I live.

With Tesla pulling map information used for routing in the summon feature from OSM, this gives not only an overall incentive for Tesla owners to enhance the OSM database, but also gives them the ability to add the data they personally need, making the feature more useful in places which didn't have good data yet.

I wonder though, if Tesla could use the data collected by the vehicles in general to improve OSM. As Tesla seems to track special driving situations for building and enhancing the Autopilot features, they could also automatically detect wrong or missing map information, whenever a Tesla drives "off" the map. Based on that information, they could contribute to OSM a lot.

Anonymized gps traces are a common way to contribute, and routing errors as well. They could build a "Report an issue" feature which can be shared with the community after being validated by their data team.

It's that hard to walk to a car we need to reinvent the wheel? It's getting ridiculous

One way in which this could (I am being very speculative) be useful is that when we pass a tipping point and most cars are autonomous enough for summoning, this could change the architecture of public buildings like malls, condos, and offices:

Parking could be deliberately placed far away: Cars could drop you at the front doors, then drive off and park themselves, returning to pick you up when you are done.

This could revitalize downtown shopping districts, for example.

It may not be a hardship for able people to walk to their car in today’s malls, but perhaps in the future your car could be parked a few kilometers away.

That could be a big win for liveability: Imagine malls that didn’t need to be surrounded by acres and acres of asphalt. Shuttles could carry people driving “legacy” vehicles between the parking and the mall, but those with autonomous vehicles wouldn’t need to wait for and share a shuttle.

> That could be a big win for liveability

Somehow having a lot of traffic created by empty cars is not a win in my liveability book. Unless you're sharing the car and it's picked up by someone 5 minutes away as soon as you get out you earn very little besides some driver convenience. Instead of acting like taxis that you drive yourself, they act like limos and everyone gets to feel like they have a chauffeur to get dropped off exactly at the spot.

The whole premise of autonomous cars is that they behave like limos. I don’t see how that is any worse than people driving their own cars, I see it as better in many ways, including this.

Not as good as people sharing cars, or sharing a Big Urban Service vehicle, we could build it to hold dozens of people... I guess it would be known by the acronym “BUS...”

> The whole premise of autonomous cars is that they behave like limos

Those would be the fully self driving cars. Cars that can somehow navigate short distances to a parking spot where the next person takes over would be the middle step. They sort of act like a valet service or a "self driven taxi" (comes to you where you need it).

> I don’t see how that is any worse

A FSD is not worse. Just not better for anyone but the owner unless it's a shared vehicle. The only way we all benefit is if the number of cars goes down because they are shared or people move to more public forms of transport.

>This could revitalize downtown shopping districts, for example.

Downtown shopping districts already solved this problem...by putting parking lots in the back or shared multi-level parking garages off to the side.

Self driving cars aren't magically going to give them money to rebuild if they haven't already.

In Toronto, parking space within walking distance of shops is expensive. People would rather build condos with it. Being able to park cars five or ten minutes away would make parking more feasible.

Of course, we have a crazy thing called public transportation, but our populist Premier hates it.

> Parking could be deliberately placed far away: Cars could drop you at the front doors, then drive off and park themselves, returning to pick you up when you are done.

> This could revitalize downtown shopping districts, for example.

Or we could have that today by limiting car access to downtown shopping districts, having large parking lots outside of town and investing in public transit. This is being enacted and it's working in many middle-sized European cities.

Well of course, but that’s a different conversation, one in which I’m in agreement with you. Public transit is a win.

But as long as people have cars for whatever reason that is outside of my control, cars that can park themselves and then return to pick us up could create less unliveable cities than the ones we have today.

That system isn't getting rid of any asphalt. It's just wasting more time and energy unnecessarily.

Not true!

Acres of asphalt exists because we need to physically park our cars and then walk safely to and from the mall.

If we don’t need parking in close proximity, or parking that a human must drive into and walk out of, we can build much more efficient storage for cars than acres of asphalt.

They do it in Japan, we just don’t do it in North America.

I won't use this feature personally much beyond playing around a bit with it, and in any case it has not arrived in my country yet.

However I can see it will be beneficial to people with disabilities and the like. It's not all about laziness.

It is also a (small) step towards the fully self driving features they aim for.

Yes it sounds unreasonable to me too.


It’s this constant battle between those who can’t imagine and the unreasonable man.

In today's world, is such a marginal progress worth the ecological sacrifices being made in its name?

Think of this as another step towards full automation.

This isn't necessarily just a "nice to have" to have your car come to you in the parking lot. This is also potentially to have a semi-truck get itself out of the parking lot one day, or your taxi to show up from a remote stand.

Reminds me of Bill Gates trying to explain the value of the Internet to David Letterman.

"Does 'radio' ring a bell? 'Tape recorders' ring a bell?"


I'm waiting for the Tesla with the mobility scooter built in like the batcycle in the batmobile. /s

Uhm, anyone else worried about the security here? What if someone were to draw a “parking lot” polygon that resulted in a path to water?

Not a tesla owner, but AIUI you have to hold the button down on the app while it comes to you - so you'd let go if it looked like it was going in the water.

This is exactly why having open source mapping information is so useful - if you see an issue you can go in and correct it rather than hoping an anonymous corporation will do it for you.

I guess the key here is, that Tesla knows that that OSM isn't "trustworthy"[1] and consequently treats the map information as it should be: just as the source for the routing algorithm. So it uses the maps as the basis on how to get from A to B, but the actual driving is based on the sensors, not trusting the maps. So the worst case of false mapping information would be of trapping the car in a place where it doesn't know how to get to its destination.

[1] Due to its open collaborative nature, there is an obvious risk for bad or even malicious edits submitted to OSM. On the other side, "professional" and closed mapping services can have errors too, especially they run the risk of being outdated, so shouldn't completely be trusted either. Due to the community efforts, overally the OSM data should be better maintained and errors can be fixed resulting in better map quality. But the big thing is, that you cannot trust any map data 100% - and if only it has been invalidated by a tree falling onto the road.

Tesla can collaborate with the community and other business users to take the feedback from their fleet and fix the issues upstream in the data. Facebook, Apple, Amazon and others are already doing so. https://wiki.openstreetmap.org/wiki/Category:Organised_Editi...

Describing what those companies do as “collaborating with the [OSM] community” might be a bit of a stetch. At best some companies have been dragged kicking and screaming to actually talking to the OSM community.

Hi Rory, better than nothing it seems, they were all in Heidelberg for starters :)

It seems they use OSM as guidance about the structure and layout of the parking areas and then still rely on the environmental awareness of the vehicle to actually navigate.

Is the Tesla system prone to driving into water?

I'm guessing OSM is only used for routing not for actual navigation and that they filter out clear vandalism using edit criterias and ingestion delays.

I read a Facebook blog post a few weeks back about how they integrate OSM updates. I forget all of it but they use satellite images and CV to detect if a road exists, as well as some other automated tests. I think they open-sourced it, although I can't seem to find the post now

they didn't open source the code that produces those road suggestions, but they have released the data output

The water ways are also maped on OpenStreetMap and this kind of errors can be validated using querys. Also there is a big comunity checking the changes.

That of the thousands of engineers at Tesla, who focus on problems like these full time, would never consider byzantine failures?

Boeing has thousands of engineers and yet they all allowed a single sensor system able to take out a plane to exist ;)

Yeah but in this case imagine the first thing said in a feature threat model incorporating unverified user generated external data. In Boeing's case it was poor use of systems they controlled.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact