Pittsburgh. Otto. Autopilot. Cruise. Waymo. Self-driving beer trucks. Flying cars. What would’ve sounded crazy a few years ago is now a fixture of the news cycle. And the future that once seemed decades away is suddenly around the corner.

As the sun rises on the age of autonomy, is the sun also setting on today’s mapping and navigation apps?

Many believe so. In “Are Maps Necessary?”, Nicolas Carr writes:

“We’re in a moment of transition, as the automation of navigation shifts responsibility for map-reading from man to machine. It’s a great irony: the universal map arrives at the very moment we no longer need it.”

Meanwhile, the San Francisco Chronicle recently published “As self-driving cars arrive, Google Maps looks for its place”, writing:

“Autonomous cars create new mapping needs, and it’s unclear what role Google Maps will play”

And speaking of Google Maps, I can’t write anything about it these days without getting responses like this:

Many are convinced that Google and Apple Maps are about to go the way of the dinosaurs, and they’re already writing the obituaries.

But what if this is wrong?

PART 1
TODAY

Today’s mapping and navigation software is optimized for human drivers. In order to be successful, it must answer four basic questions:

Before the trip:

   (1.) “What’s the best route between my origin and my destination?”

During the trip:

   (2.) “When’s the next turn?”
   (3.) “What’s it called?”
   (4.) “And is it a left or a right?”

As long as the software accurately answers these four questions, we’ll get to where we’re going. And though some of these products are better than others at lane guidance (Google) and showing traffic jams and cops (Waze), the experience across them is remarkably similar:

Though the apps above all look a little different, they all work the same way – and each has the same two modes: Route Selection and On-Trip Navigation.  

Route Selection answers our first question:

   “What’s the best route between my origin and my destination?”

The screen itself is usually a map with the two or three best routes highlighted. Here’s how it looks in Apple, Google, and Waze today:

Notice that they all follow the same general pattern: they all have a map of the best routes in the middle and a “Start” or “Go” button below it:

Notice also that the maps occupy more space than anything else – around 2/3s of each screen:

Upon hitting the “Start” button in Route Selection, the user will enter a second mode called On-Trip Navigation – where they’ll remain for the rest of the trip.

Compared to Route Selection, far more of the trip is typically spent in this second mode:

On-Trip Navigation answers our three remaining questions:

   “When’s the next turn?”
   “What’s it called?”
   “And is it a left or a right?”

The screen itself has three main parts: an instruction for when and where to act next, a detailed map of the next couple miles, and a set of figures tracking your overall progress:

Here’s how it looks in Apple, Google, and Waze today:

Notice that their basic designs are identical: instruction at top, map in the middle, and progress at the bottom:

And look at how much of each screen is a map – it’s even more than in Route Selection:

As you can see, maps occupy more space than anything else (just as they do in Route Selection) – making them a core part of the experience.

So that’s the basic design of navigation software today. But the Age of Autonomous Vehicles is quickly approaching...

...and the questions that today’s navigation software is designed to answer...

   (1.) “What’s the best route between my origin and my destination?”
   (2.) “When’s the next turn?”
   (3.) “What’s it called?”
   (4.) “And is it a left or a right?”

...all become irrelevant tomorrow as the vehicle takes over the responsibility of driving and answers those questions itself.

So because the car will choose its own route, we’ll no longer need Route Selection:

And because the car will be making all the turns, we’ll no longer need most of On-Trip Navigation:

All we’ll really need to know is where we want to go – and the car will do the rest. And the “driver” will become just another passenger, free to do all the things that passengers do, like:

   - 💻  Work
   - 📚  Read
   - 📺  Watch movies and TV shows
   - 🕹  Play games
   - 📞  Communicate (meetings, conference calls)
   - 🍽  Dine and drink
   - 💤  Sleep

These all sound like things that people do on planes today, don’t they? And it’s true: riding in an autonomous vehicle tomorrow will be much like flying today.

But here’s the thing about flying: we’re so busy doing other things (such as working, reading, and sleeping) that we often lose track of where we are and how much time is left.

And this is useful information.

“Should I eat on the plane? Or should I wait until I land (when there will be many more options – and the food will be so much better)? Should I get in line for the bathroom now? Or can I just wait until we land?”

Flights are long. And the answers entirely depend on how much time is left.

To that effect, airlines have started adding maps to their in-flight entertainment systems – and some even suggest that they’re the most popular feature of these systems:

But what’s interesting here is that none of the people looking at these maps are actually flying the plane. Nor are they navigating it. Yet they still want to see a map; they still want to see where they are.

And the same might be true for self-driving cars. We’ll be busy doing so many other things (like working, reading, and sleeping) that we’ll lose track of where we are and how much time is left.

But unlike planes, cars don’t have flight attendants or kitchens or restrooms. And whereas we never have to worry about refueling the plane, we constantly have to refuel and recharge our cars.

So in addition to the four navigational questions we looked at earlier...

   “What’s the best route between my origin and my destination?”
   “When’s the next turn?”
   “What’s it called?”
   “And is it a left or a right?”

...there’s actually another set of trip-related questions, like:

   “Where are we at now? And how much time is left?”
   “How long until I need to refuel / recharge? And where’s the best place to do it?”
   “I’m starting to get hungry, where’s the best place to stop in the next hour?”
   “I could really use a coffee. Is there a Starbucks anywhere nearby?”
   “Where’s the next bathroom?”

Of course, none of these questions are new – and most are as old as driving itself. And the longer the trip, the more of these questions there are.

Interestingly, those most interested in purchasing AVs are those who drive the most and the farthest:

📊 Source: Auto Insurance Institute

So while AVs change who is driving – they don’t change many of the other realities of traveling. Put another way: cars are changing, but people aren’t. And people are still going to:

So as magical as AVs will be, they won’t be magical enough to make all of these other realities disappear. And we’ll still have the same, old trip-related questions.

Unfortunately, today’s mapping and navigation software isn’t optimized for answering these questions. It’s not that they’re incapable of answering them; it’s just that they’re not as good as they could be at answering them because they’re so focused on human drivers and our original four questions:

   “What’s the best route between my origin and my destination?”
   “When’s the next turn?”
   “What’s it called?”
   “And is it a left or a right?”

Let’s look at some of the issues...

Because today’s mapping and navigation software is optimized for the human operation of a motor vehicle, it’s focused on the next time the driver has to act. As a result, the map is zoomed in very close, in order to give a detailed view of the next turn.

Now this isn’t so bad in cities, where there’s plenty of interesting things to look at on-screen...

...but outside of cities (i.e., 97% of the world), it can be abysmally boring:

...yet even though it’ll be a while before I have to turn, each one of today’s navigation apps locks me in to a narrow (and not particularly useful) view of the mile ahead:

Now as silly as this all sounds, it does make some sense. Today’s apps are (rightfully) focused on turns because that’s where we need to act. (And if we miss a turn, we won’t get to where we’re going.)

But tomorrow, with self-driving cars, we won’t have to do anything. No turns. No maneuvers. The car will do it all for us.

So because we’ll no longer have to do anything, we’ll no longer need such a close up view of the next mile. Our focus will shift from “Where do I turn next?” to “What’s around me?” and “What am I approaching?”

In other words, we’ll start caring more about macro details than micro details. Which brings us to our next issue... the only distance shown in today’s software is the distance to the final destination. And the software doesn’t tell us the distance to the other cities and places along our route.

Now this isn’t so much of a problem today because as we’re driving, we constantly see signs like this:

But tomorrow, we’ll be doing other things instead of driving, and we won’t be paying attention to the signs along the road.

In fact, fewer than half of us expect to watch the road while we’re riding in self-driving cars:

📊 Source: Auto Insurance Institute

So we’re going to miss many of those signs. And worse, those signs tend to be infrequently placed. Miss one and we might not see another for 20 minutes. (And we might miss that one too!)

Worse, because our map is always zoomed in so close, we can’t see upcoming cities until we’re already entering them. So in the future, we might pass through cities without even realizing it.

But why are we so interested in cities anyway? Because of the places inside of them – places like restaurants and hotels and gas stations:

Places are important while we’re traveling because of the services they offer. And many of the trip-related questions we’ll have tomorrow are centered around services and places:

“How long until I need to refuel / recharge? And where’s the best place to do it?”
      ⛽  Gas Stations & 🔋 Charging Stations

“I’m starting to get hungry, where’s the best place to have dinner in the next hour?”
      🍽  Restaurants

“I could really use a coffee. Is there a Starbucks anywhere nearby?”
      ☕  Cafes & Coffee Shops

“Where can I grab a snack?”
      🍫  Convenience Stores

“I have to go to the bathroom.”
      🚻  Rest Areas (and any place with a Restroom)

“Where can I stay overnight?”
      🛌  Hotels & Lodging

It’s for this reason that Tesla positions its Superchargers within a short walking distance of clusters of restaurants and shops. (If it’s going to take 40 minutes to charge the car, why not get something else done in the meantime?)

* * *

We saw earlier that, because we’ll be doing other things, we’ll likely miss those distance signs along the route...

...but there’s an even more important set of signs we’ll be in danger of missing:

Just as we care about services and places – we also care about the lack of services and places. In other words, we’ll want to know if we’re about to drive into a desert:

But that’s yet another problem with today’s mapping and navigation apps: they don’t warn us before we enter corridors of sparse services. And because we’re always zoomed in so close...

...we’re always flying blind.

* * *

All of the issues above can be traced back to a single root cause: that today’s navigation software is designed for the human operation of a vehicle. Put another way: Today we have maps designed for drivers. But tomorrow we’ll need maps designed for passengers.

But there’s a second, even deeper set of issues affecting today’s navigation software...

Up until a few years ago, most online maps were sets of static images:

Each zoom was a separate image, and the images were bundled together to create an interactive, zoom-able map.

But because all of the map’s content was pre-baked into these static images, the map couldn’t change upon our inputs. Instead, things like search results and traffic and transit had to be overlaid on top of these images – often obscuring the map itself.

So pins would mark search results – while sometimes also hiding them:

And red, yellow, and green lines would show us traffic – while also hiding the roads the traffic was on (along with the rest of the map):

And with transit, we’d get two maps in one – with the first map hidden underneath a spaghetti of subway lines:

Making matters worse, screens in those days weren’t as good as they are today...

...so to be legible on low-quality screens, most maps had to be brightly colored and have thicker lines than they have today:

And this, in turn, created an interesting challenge for pins and routes, which had to be easily visible on top of these already busy and colorful maps.

So pins were given dark shadows and distinctive colors:

And routes were drawn as thick, impossible-to-miss, blue lines:

But because routes were also overlays on top of the map, the very things that helped them stand out also worked against them.

It was a great irony: the labels you were most interested in reading, were the ones most difficult to see. And because the labels were baked into the map, they couldn’t be moved to a more legible place.

In those days, the map was interactive in the sense that you could pan it and zoom it and load overlays. But because the map itself couldn’t change, the experience wasn’t much different than browsing a set of a paper maps – paper maps, under glass.

Fast forward to today and now, thanks to vectors, layering, and other modern rendering techniques, maps are no longer static images. Instead, they’re dynamic – responding to our every input and adapting to us.

Piles of pins are relics of the past as labels themselves now come forward as search results:

And notice how well the rest of the map adapts to the search results. Even the street labels shift and scoot to make room for them:

And speaking of labels... today when we tap on them, the map’s content changes to show us other places we might be interested in:

It’s so cool and useful, and it would’ve never happened in the static map days.

And look at traffic today: it’s now right where it belongs – tucked neatly under icons and labels:

But there’s perhaps no better showcase of the dynamism of today’s maps than in how they present transit information. Instead of overlaying a smattering of spaghetti, the map morphs into a transit map:

Notice that two things are happening here:

1. Information that’s unrelated to transit is being dimmed and de-emphasized. Roads are de-colorized and their labels are dimmed. And parks are de-saturated.

2. New transit-related information is being added: lines and stops and stations.

These changes transform the map, allowing us to better see and understand the transit information.

And look what happens when we request transit directions between the New York Stock Exchange and Times Square:

Once again, information that’s unrelated to us is de-emphasized. In this case, the transit lines unrelated to our route have been dimmed and de-colorized, along with all of the unrelated stations – bringing greater focus to our route.

And new information is added: the stations we’ll pass through on our way to Times Square have been brought forward and highlighted:

By highlighting the stops we’ll pass through, the map gives us an easy way of assessing our progress along the route.

Vibrant. Dynamic. Responsive. Intelligent. We no longer have the static images, the paper-under-glass we used to have. Instead, today’s maps feel like living, breathing things – highlighting everything related to our taps and inputs, and dimming and de-emphasizing everything that isn’t.

But there’s a peculiar exception to all this – one that becomes even more peculiar when you consider that it affects the most popular feature of today’s mapping and navigation apps: Driving Directions.

Oddly, when we request driving directions today, the map’s content doesn’t change at all:

Above, I’ve requested driving directions between “San Francisco” and “Los Angeles” on Apple Maps – and it’s as if we’re still in the static map days. Apart from the blue route lines, nothing else on the map has changed. No new information has been added, nor has anything unrelated to the routes been dimmed or de-emphasized – the opposite of the transit maps we just saw.

And we see the same when requesting driving directions between “San Francisco” and “Los Angeles”) on Google Maps:

Once again, a couple of blue routes are added – but the rest of the map’s content (the cities, the roads, the parks, etc.) is unchanged.

But why should the map’s content change? What’s the big deal? It’s surprising that the map’s content doesn’t change, because so little of what’s on the map is related to either route...

Let’s take another look at the routes on Google’s map, but this time we’ll only focus on the things labeled within ten miles of them.

Why ten miles? Similar to what we saw with batching, people want to save time while they’re traveling. So given how much more time it would add to their trip, it’s unlikely the average person would go more than ten miles off-route to stop at a gas station, restaurant, etc.

So let’s highlight the ten miles either side of each route:

Interesting, isn’t it? Almost everything labeled on the map lies outside these highlighted corridors:

And look how little is labeled within ten miles of Google’s suggested route:

We’re driving 400 miles across half of California and just four cities are labeled between the origin and destination: Los Banos, Avenal, Lost Hills, and Buttonwillow. And there are large gaps between them:

   - A 115 mile gap between San Francisco and Los Banos
   - A 85 mile gap between Los Banos and Avenal
   - A 125 mile gap between Buttonwillow and Los Angeles

What’s between these gaps? Are there cities and services – and the map just isn’t showing them?

Or is it more like this?

Unless we go hunting – zooming in and out repeatedly, all along our route – we’ll never know.

So given how few cities are shown – and how large the gaps are between them – why doesn’t the software add more of them along our route? Similar to how it added the pass-through stops along our transit route:

What makes this all even stranger is that we’ve specifically asked the software for routes between “San Francisco” and “Los Angeles”. In other words, we’ve specifically said to the software that this is what we’d like to see in greater detail. And yet the routes aren’t any more detailed than the rest of the map – even though there’s plenty of detail that could be added...

So given the detail that could be added, isn’t it surprising that the map doesn’t elevate points along our driving route in the same way it elevated stations along our transit route?

Isn’t it surprising that the route doesn’t look more like this?

Of course, if we zoom in far enough, we’ll eventually see all these things on the map. But by then, we’ll be zoomed in so close, we’ll see only one or two of them at a time. (And it’s tedious to hunt for them.)

So because we can’t see all these details together, as a set, it’s difficult to plan our stops ahead of time.

But here’s the greatest paradox of all: the longer the trip we take, the more stops we’ll have to make. But the longer our trip, the less detailed our overview will be:

In truth, we’re abusing the Route Selection map. It was never intended to show us all the cities and services along our route. Its purpose is only to help us choose a route.

But now we see the second set of issues with today’s navigation software: we’re still in the static map days. Navigation today is a route added to a map – i.e., we take the map we already have and draw a blue line across it:

But tomorrow, we’ll want a map added to a route – i.e., a map that’s built around our route, showing us the services and cities along it:

So let’s imagine what this map would be like – a map that’s optimized for autonomous vehicles and the questions that come with them, that puts the information we’ll want directly in front of us, and that’s designed for passengers instead of drivers.

To do that, let’s fast forward to tomorrow...

Let’s imagine that it’s a few years from now. We’re one of the first people lucky enough to have a self-driving car, and we’re taking a late afternoon trip between Omaha and Chicago.

Our car has chosen our route and we’re exactly one hour into our trip, somewhere on Interstate 80 in western Iowa:

Here’s what it looks like in front of us…

And if we were using one of today’s mapping apps, we’d see something like this:

As we saw earlier, the longer our trip, the more services we’ll need. So given that we’re on a seven hour trip (with six hours to go), we have at least four things to figure out:

🔋 Re-charging our car
Our self-driving car is electric and gets around 250 miles between charges – so we’ll need to recharge it at least twice along the way. Thankfully, our car’s manufacturer offers a special kind of charger, called a “Rapid Charger”, that’s meant for long-distance trips and offers fast charges. How many are along our route? And where?

🍽 Dinner
We haven’t had dinner yet, and we’ll be arriving in Chicago after 11pm (when most restaurants are closed) – so we should eat along the way. That said, we’re not hungry yet, but we will be in a few hours, and it’d be awesome if we could eat while our car is charging. Which Rapid Chargers have the most restaurants around them? And does anything look good?

🛏 Where to stay overnight
We haven’t booked a hotel yet, so we need to find somewhere to stay. And given that we’re booking last minute, it might be cheaper to stay outside of Chicago – but we don’t want to stay anywhere that’s far from the city or that isn’t along our route. And we also need to make sure the hotel allows pets because...

🐕 We brought our dog.
And we’ll need to walk it a few times... but where? If there are rest areas along our route, they’d be perfect for this...

So we’ve got these four things to figure out. And though it’s certainly possible to figure them all out with today’s mapping software...

...it’s going to take a bit of work, with several steps:

If only this was a little simpler.

And as we saw in Part 1, the maps themselves aren’t much help either – the Overview is too zoomed out and On-Trip Navigation is too zoomed in:

It’d just be so much easier if we could see, in a single view:

- 🔋 The locations of the Rapid Chargers along our route.
- 🍽 The number of restaurants around each Rapid Charger.
- 🛏 The number of hotels in each city.
- 🐕 Rest area locations.
- 📐 And some indicator of our distance from all of these.

So let’s take the On-Trip Navigation map we have today...

...and let’s stretch it and zoom it out:

Now we can see our whole route in front of us:

But all we’ve really done is trade two miles of beige for hundreds of miles of it. So let’s start building a map along our route.

First, we’ll add our destination – and while we’re at it, let’s also add our trip’s halfway point so we’ll know when we’re halfway:

Now let’s add some of the other landmarks along our route.

We’ll add the next several rest areas, the Illinois state line, and a popular state park / tourist area that we’ll pass by later:

Next, let’s add some of the cities we’ll pass through.

In the foreground, we’ll add the next several cities... and further away in the background, we’ll add some of the larger cities along our way:

It kind of looks like a transit map, doesn’t it? All the cities are laid out as if they’re stops along our way. But the map still looks bland, doesn’t it? Too much beige...

So let’s fill it out by adding some extra context. First, let’s add a more detailed picture of the landscape along our route – things like hills and forests and rivers. And let’s also add some of the other roads in the area – but unlike Google and Apple, we won’t colorize them since they’re unrelated to our route:

Next, let’s add some of the other, nearby cities. But similar to what we just did with the roads, let’s de-emphasize them by making the off-route cities look less prominent than the on-route ones:

Now we can clearly see, at a glance, which cities are on-route and off-route.

So with everything we’ve added, here’s what we end up with:

We’re no longer flying blind, and we now have a clear picture of what’s ahead of us.

But how far are we from all of these things? And where will we be in an hour? (When we’ll be starting to get hungry...)

* * *

We saw in Part 1 that fewer than half of us expect to watch the road while we’re riding in self-driving cars:

📊 Source: Auto Insurance Institute

So we’re going to miss many of the signs along the road, including these:

But we can actually do better than those signs because our car knows many other things that those signs don’t, things like:

- Our Current Location
- Our Current Speed
- The Speed Limits of all the roads along our route
- The Speed of Traffic along our route
- Accidents along our route
- Traffic Jams along our route
- Construction along our route
- Weather and Road Conditions along our route
- And the Quality of the Roads that make up our route

Those old signs never knew any of this, so they could only tell us the distance in miles. But really, we don’t care about miles – we care about time. (Miles are only a means of calculating time). And besides that, people generally have trouble visualizing distances: “500 feet”, “275 yards” – it’s hard to picture that. But “five minutes”... “a half hour”... – we all have a good sense of those.

So let’s swap miles for time...

...and let’s add these live travel times to our map:

Now we know how far we are from everything ahead of us.

But all the cities look the same, don’t they? How are they different? And which ones offer more services?

If only we had an easy way of knowing what’s inside of each...

If you’ve ever driven in the U.S., you’ve undoubtedly noticed signs like this:

And this:

And this:

And nearly every U.S. city, no matter how big or small, seems to have them:

But it’s nice that they do, because the population figures on these signs are useful information: they’re a proxy for how many services a city is likely to have.

For instance, a city of 10 or 100 might not have any services. No gas stations. No shops. No restaurants. But a city of 1,000 will likely have a few, and a city of 10,000 will surely have many. And after a certain size, a city will offer just about anything.

So generally, the higher a city’s population, the more services it’s likely to offer. And just as we use miles to calculate travel times, we use populations to guess at how many services a city is likely to have.

But here’s the thing: we don’t actually care about how many residents these cities have – we just want to know how many services to expect. So let’s reframe how we look at cities, from collections of people to collections of services & places:

And when we look at it this way, we realize we already know what kinds of services are in each city...

* * *

It’s been fewer than eight years since Google started adding labels for businesses to its map. But in that short time, Google, Apple, and others have quickly built up massive, global databases of Point of Interest information. And the coverage is becoming so good that we now know which cities have restaurants, which have convenience stores, which have hotels, which have Rapid Chargers, etc.

So let’s add this information to our map:

The city has a total of 26 places.

But we can actually count the city’s places by category, like this:

And we can do it for all the other cities, too:

So we can go beyond population and directly compare cities by the number of services they offer:

But we can do even better yet...

Not only do we have a rough count of the number of services in each city, we know which services will be open as we’re passing through:

So let’s add this information to our map – the number of services that’ll be open in each city as we pass through – along with each city’s population:

And just as we know where all the services are, we also know where they aren’t.

So let’s take our map and highlight the “dead zones” – the corridors of our route where there aren’t any services for several miles:

Now we’ll know when it’s our last chance to grab a snack or a bite to eat... or if our car is about to drive us into a desert:

So our map now shows us which cities offer which services – and how many they offer. But we’re still only getting half the picture, aren’t we?

We know which cities have the most options, but we don’t know what the options are. For instance, where can we grab a decent cup of coffee in the next 20 minutes?

Out of everything in the next 20 minutes, Stuart looks the most promising. It has more restaurants than the other cities, so it likely has something with a drive-through. But what are these restaurants in Stuart?

With today’s maps, we have two ways of figuring this out: Browsing the Map or Searching. But as we saw, both can be cumbersome. If only we could see some sort of list of what’s inside Stuart – something similar to a mall directory, but for a city.

So let’s see if we can make something like that:

And hey, since we’re in a self-driving car... let’s add “Go” buttons next to all the services. That way, if we see something we like, we can just press one of these “Go” buttons and the car will drive to it:

But honestly, everything on the map should act like that – so let’s make everything along the route tappable (including the cities) and then we’ll tap on Stuart and schedule our stop:

And while we’re at it, let’s schedule a couple more stops: we’ll stop and recharge in West Des Moines. And we’ll walk our dog at one of the rest areas further out:

I’d rather have a map like this than no map at all. And that’s why I don’t believe those who say self-driving cars will end map reading...

We can’t apply today’s maps to tomorrow’s cars – but this is exactly what those who think cartography is dying are doing.

It’s not that we’ll no longer be navigating, it’s that we’ll be navigating different things (services and places instead of roads) – and we’ll need new kinds of maps to help us.

Except for Des Moines and West Des Moines, all of the map’s cities are presented in exact same style (i.e., they have the exact same text size, weight, color, and icon).

But even though the cities look the same, they have widely varying populations:

Notice how large Ankeny’s population is, compared to Marysville’s:

Ankeny’s population is more than 50,000, and it has a large number of services: 12 hotels, 20 gas stations, and more than 150 restaurants.

Here’s what Ankeny looks like from the air:

In contrast, Marysville has a population of just 66 and is little more than a group of houses and a post office:

Clearly, these are very different cities, and Ankeny is 781 times larger than Marysville. But because they’re presented in the same style, you’d never know it by looking at the map: