Future of Maps: Self Driving Cars, Satellites & UAVs, Crowdsourcing

2015, as well as the beginning of the 2016, have been quiet eventful periods in mapping services area: we’ve seen some deals, partnerships and announcements that outlined long-term trends in the industry, such as:

  • New growth areas
    + Connected Car
    + On-Demand Economy
  • Increased number of data sources and their quality
    + Crowdsourcing
    + Satellite/UAV imagery
  • New features
    + Offline and predicative capabilities

Some supporting details on these statements are presented down below.

Maps are essential for the future of Connected Car

The biggest announcement in mapping services area was definitely HERE acquisition by a consortium of German carmakers including BMW, Audi and Mercedes’ parent company Daimler for more than $3 billion. This deal has serious strategic impact for carmakers consortium because HERE is “one of the biggest and more valuable mapping assets to come to the market in coming years” that could fall into other autonomous vehicles developers, that expressed interest in this deals, such as Uber + Baidu, Apple and Amazon (despite this, it was announced that “HERE’s management would retain its autonomy with the company continuing to serve rival car makers”).

But why maps are so important for implementation of self driving cars concept — because fully autonomous cars will require maps that are significantly different from the maps that we use today. First, they have to be extremely detailed: if meter resolution maps may be good enough for GPS-based navigation, map for self driving cars has to include data on location of every lane market, curb, the height of traffic light and meaning of every traffic sign at centimeter accuracy. Second, maps for driverless cars have to be real-time ones, updated with information about accidents, traffic jams, lane closures and weather conditions to understand the “big picture” of what is happening on the road. Today’s autonomous vehicles are already pretty good at sensing obstacles and hazards, such as other cars and pedestrians due to progress in sensors’ technology, they can even recognize unmapped traffic signs. But in a complex situation (like at unmapped traffic light at the crossroad) car might slow down and turn on super cautious driving mode that may cause dangerous situation on the road.

Thereby, the main idea is that the basemap is pre-loaded and secondly updated, vehicle can focus its computing power on temporary obstacles — like other cars and pedestrians.

As it can easily be seen from the above, creation of super high-definition real-time map is an enormous task — vastly more effort than what’s needed for Google Maps (that explains the fact that Google Cars were tested mainly in Mountain View –only 2,000 miles of road was mapped by Google, while the whole US road network has 4,000,000 miles of road).

At the moment there are 2 approaches to face requirements that were detailed above.

The first one is Google’s “do everything” approach: the company controls its entire self driving car operations, gathering the map data itself and processing it for the intelligent software that drives its cars. For example, Google may use Street View to get its map data, collected by a dedicated fleet of cars, but it would be expensive and impractical to run Street View dedicated cars for the sole purpose of repeatedly scanning roads to keep maps up-to-date.

Another approach has been taken be HERE Maps. Like Google, HERE is also driving its own mapping cars fleet, but, more importantly, company plans to collect data from the cars themselves and share it through the cloud to other vehicles — thereby realizing the concept of high-definition real-time maps, detailed above.

As an example, let’s say that in some area tire sensors on certain car report that the tires are slipping. At the same time, local weather agency (or probably some similar startup, like Understory) send out an alert that temperatures have dipped below freezing. HD Live Map, introduced this year at CES, aggregates and analyzes all of this heterogeneous information to understand that there is an ice on certain area on the road and send information to other cars that are expected to pass that area.

HERE’s cloud-based HD Live Map concept

Similar concept has been chosen by some major carmakers.

Also at CES, GM announced collaborative project with Mobileye dedicated to “exploring” the use of onboard cameras and sensors of GM’s cars to gather high-definition mapping data by utilizing cellular connection of its OnStar modules, already built into most of the vehicles it sells, to update the database with such data like lane markings and precise location information. According to Mark Reuss, GM’s Executive VP, this technology could be integrated as soon as “later this year”.

GM autonomous driving visualization

Toyota also plans to turn each vehicle into a map-maker by repurposing onboard GPS and cameras. It is planned to use image matching technologies and correct road image data collected from multiple vehicles, as well as high precision trajectory estimation technologies to restrict system’s margin error to a maximum of 5 cm on straight roads.

Tesla also has pretty similar ideas on how to build next generation maps for autonomous vehicles: the company will use cloud technology build into its Version 7.0 software to gather road data for Model S and Model X owners via radars, cameras and sonar equipment that’s installed on Autopilot-enabled Teslas and leverage this date to create high-definition real-time maps. Elon Musk calls this a “fleet learning network” where all of its cars contribute to a shared database. “When one car learns something, all learn,” said Musk.

Certainly, crowdsourced approach to map making looks more viable (huge number of data sources will help to meet high-definition and real-time demands of maps for autonomous vehicles), thereby, crowdsourcing is the traditional carmakers’ very big advantage and, according to Raj Rajkumar of Carnegie Mellon University,

“There’s an interesting competition ahead and it’s an unusual position for Google, whose power has always come from having more data than its competitors.”

Summing up, despite there are different approaches to the mapping for autonomous vehicles, maps are definitely critical feature for implementation fully autonomous vehicles — thereby automotive industry would be a huge driver for developing mapping technologies.

On-Demand Economy would be a significant driver

Talking about various On-Demand-Economy startups, all of them require user location information and, accordingly, mapping services to be integrated through API. Taking into consideration On-Demand startups investments statistics and some general stats (Burson-Marsteller’s and Intuit’s surveys) it can be positively concluded that this new economy is here to stay and availability of such services as Juggernaut, supports the hypothesis that the integration of maps in the on-demand apps is already an important task.


As Nokia HERE deal demonstrated, creation of company’s own maps costs a lot while the value of such maps is pretty questionable (HERE was sold for around $3 bn. while the acquisition of Navteq in 2007 cost Nokia $8,1 bn.) largely due to free data sources such as OpenStreetMap, that are, according to some opinions, are no worse than Google Maps and even better in some use cases, such as walkways. By the way, there may be an interesting niche for super accurate maps for foot-passengers and wearable apps that have been published in mid-2015 by major players, such as Google (App and API) and MAPS.ME.

Consequently, key player began to attract users to work on maps’ upgrading and improving. Google have been pretty active with its Local Guides program and Map Maker integration into Google Maps iOS App. Local Guides program is offering some rewards like free Drive storage, early access to new Google features and a large set of community activities for “writing reviews, uploading photos, adding new places, fixing outdated information and answering simple questions”, spicing it with some gamification (level ups). Map Maker is another example of Google’s crowdsourcing approach: since November 2015 users can easily add a missing business on the map form smartphone.

Regarding Nokia, in 2013 crowdsourcing tool called Map Creator was introduced, first tests occurred in India and in March 2015 HERE Map Creator API Challenge was hold in Indonesia, where the main criteria was that created app to be able to “add new information such as roads, routes of hiking, biking, pedestrian, natural disaster evacuation, POIs and house numbers.

The choice of the countries was not accidental: India and Indonesia have some similarities in terms of mapping, namely, poor maps accuracy because it really depends on the spare wealth of the country: it requires people to have some spare time, computers for non-paid mapping for community from one side and extensive government spendings from the other. Moreover, in many areas, formal street names and addresses don’t exist and the only real alternative is to use address of the form “nearby that haunted house”. Respectively, it is nearly impossible to organize delivery, receive visitors, etc.

However, there is a number of developers that are working on this problem. One of them is what3words, startup (raised $3.5 million round led by Intel Capital in November 2015) developing location reference platform comprises a global grid of 57 trillion 3mx3m squares — each of which has a unique, pre-assigned three word address, such as skins.paused.civil, for example. Company is generating revenue by licensing three word addresses to geographic information systems providers such as Esri and Norwegian National Mapping Authority. At the same time, Google released similar function called Open Location Code in April 2015 and integrated it in Maps in August 2015.

Satellite/drone imagery would be implemented in maps much more actively

Louvre, Paris via Digital Globe Maps API

Satellite-based Earth Observation industry is transforming: imagery is “democratizing” and becomes much more available for non-professional users due to companies significantly changed distribution model: nowadays user can simply purchase integrate satellite imagery through API and integrate data to their apps instead of contacting resellers and negotiating the terms for a long time. This applies well as to VC-backed “New Space” startups like Planet Labs, Urthecast, and many others as to established market players like DigitalGlobe (recently partnered with Mapbox to launch its own API). In case for satellite imagery API becomes a powerful sales tool in addition to being a tool for developers: imagery providers’ content and services (like NDVI calculation) is integrated in other services (like farm management software) and this generates new users and, correspondently, new revenue streams.

Second significant trend is an increase of the data sources quantity: the number of satellites to be launched in coming years (including Cubesats) is expected to grow exponentially and, according to some estimates, more than a half of them would be Earth Observation ones — price for satellite imagery would decline significantly (due to increased number of the supply on the market). Moreover, increased number of sources will change pricing also: instead of paying for the whole shot, it would be possible to pay for interest area instead of paying for the whole satellite image (as an example of such approach, Astro Digital).

Astro Digital satellite imagery browser

Talking about UAVs, their use in maps would be closely connected with crowdsoursed maps concept. Sales of commercial (DJI, 3DR, etc.) drones (which are good enough for making high resolution imagery that are suitable for use in maps) were doubling last 3 years and there are some premises that exponential growth will go on. Thus more people would have instruments for making high-resolution imagery and share it with the community. Moreover, there are already such examples for commercial solutions and as for hobby. Also, there are some examples of interaction between UAV-related companies with Mapbox.

Probably, data from satellites and UAVs would be “mixed” (data fusion) and complement to each other and available for more users and developers through API. Thereby, an exponential growth in the number of data sources and its quality is expected that will greatly improve maps accuracy.

New features

Several new features have been integrated into key mapping services during 2015 and they can be structured into 3 groups: offline capabilities, Points of Interest and predicative capabilities.

Offline navigation and has been introduced in Google Maps in November 2015 (while offline viewing of maps became available in May 2014) and HERE announced iOS app with offline navigation feature in March 2015 (while it has been available for Android users for a while). At the same time, there are some other companies that position themselves specifically as offline navigation apps, such as MAPS.ME.

Offline maps work pretty simple — users just have to download pre-selected area and it becomes available offline at any time. Such capabilities bring significant value for travelers: it allows to cut down roaming costs, navigate underground and expand battery life. Also, as it was stated above, offline maps are important for implement self-driving concept as they provide a failsafe for autonomous vehicles.

While offline navigation is not a really new concept, predicative capabilities in mapping services are much more interesting: new feature implemented in Google Maps in January 2016 suggests user’s destination before it’s been entered based on location and searches history. But the most interesting feature (available through Directions API) allows to predict the time of moving from point A to point B in the future using historical time-of-day and day-of-week traffic data to estimate travel times at a future date.

Predicative Travel Time via Google

Many uses cases (that can easily be monetized via API) can be imagined such as optimizing route planning and delivery times, mobile workforce management and many more, especially if we develop this idea and add feature of assistance in decision-making for users and real-time data from cars that will allow to increase accuracy of forecasts. By the way, HERE also has similar feature called HERE Predicative Traffic that allows to estimate travel time up to 12 hours in advance

Among other interesting announcements trend toward more information about POIs can be outlined: Google now shows gas station with cheapest fuel along the route and suggests best time to avoid lines while visiting public places.


Summarizing this research, the following trends on the mapping services market can be outlined:

Exponential growth in the number of data sources is expected

1.Satellite and UAVs markets are close to its tipping point: the number of satellites to be launched and worldwide drones sales are growing, technical capabilities are improving and distribution models and pricing (in case for satellite imagery) are democratizing.

2. Open data sources such as OpenStreetMap, powered by community-generated data are getting used more widely followed by crowdsorurcing efforts of the key market players that are gaining traction.One of the approaches to implementation of self driving cars concept is utilizing cars’ onboard sensors to increase maps accuracy and provide real-time updating — this approach will significantly increases the number of data sources also.

3. One of the approaches to implementation of self driving cars concept is utilizing cars’ onboard sensors to increase maps accuracy and provide real-time updating — this approach will significantly increase the number of data sources also.

Implementing of new features, especially predicative ones

1. A number of new features have been introduced recently such as, offline navigation, extended information about POIs and predicative features, such as predicative traffic that seem to be the most promising ones due to increased number and types of data sources that will allow to increase forecasts’ accuracy. Also, we should expect that predicative features would be monetized aggressively in many verticals.

The emergence of new market drivers

1. Maps are essential for implementing of autonomous vehicles concept and they would be significantly different from the maps we are using today: centimeter accuracy and real-time updating is required. Due to self driving cars is one of the main technology trends and a lot of stakeholders are involved, these requirements will stimulate market demand and technology advancements.

2. On-demand economy is another global trend and implementation of most applications requires location information and geospatial data in general — that will also stimulate market demand for mapping services.

To summarize, progress in maps area is driven by the exponetial increase of the data sources number, requirements from the maps “consumers” that are involved in some very important global technology trends and new ideas of key market players on how to monetize location information.

Thanks to Yuri Melnichek , MAPS.ME CEO for a feedback on developing of this article.