Better Maps Are the Key to the Perfect Self-Driving Cars
The technology and the knowledge for creating self-driving cars are already in place. There are countless examples, from Tesla to Waymo.
However, there’s still a final piece of the puzzle needed: maps. In fact, there’s an acute need for more than just a few sources to map the world.
So, in the following lines, you’ll be able to read about why digital mapping is at the center of any driverless car program.
The Internet of Things and Its Monetization
Just like the smartphone has altered our lifestyle in a positive way, the Internet of Things (IoT) will influence the way businesses function and consumers move around the world.
In fact, there’s a lot at stake when it comes to developing high-quality maps to power fleets of self-driving cars like Uber’s or Google’s. It’s all part of fully monetizing the Internet of Things, as it will be the main medium involved.
Car manufacturers are amongst the biggest advertisers out there. Purchasing a car is one’s second most important buying decision, apart from buying a home. Additionally, the smart home is already a reality thanks to the wide array of voice-activated smart devices like Alexa. Adding cars to the same network seems only fit.
So, whatever company will be able to provide autonomous cars with precise mapping technology will benefit from a direct monetization pipeline. Businesses will be able to use the maps for advertising purposes.
Why Developing 3D Maps for Self-Driving Cars Is a Challenge
A fully 360 degrees view is mandatory for a driverless car to safely navigate the streets. Naturally, every inch of public roads needs to be mapped in 3D. And it’s not just the tarmac, surrounding environmental details need to be “recorded” as well.
Self-driving cars need up-to-date 3D maps to be able to ‘see around the corner’ and take a decision based on what’s coming. It involves aspects like icy roads or accidents. It would let the autonomous vehicle know what route to take in almost real time to get to that free parking spot downtown, for example.
But the world as we know it and public roads are constantly changing. 3D mapping of all public roads and surroundings is, in itself, a monumental task, but accounting for each small change makes it even harder.
The Big Players
There are millions of dollars poured into the digital mapping of roads, transportation infrastructure and the vicinity in which autonomous vehicles operate.
Naturally, ‘the big players’ have already stepped up.
Mobileye, an Israeli mapping company, has recently been acquired by Intel Corp for a whopping $15,3 billion. The company uses ubiquitous in-car cameras to gather real-time road data for driverless cars.
DeepMap is just a small and fresh startup joining the race of high-quality maps for self-driving cars. They recently got $25 million funding on top of a previous $7 million from a previous funding round. DeepMap plans to create high-quality maps using imagery from cameras and data from lidar units.
Then there’s Nexar with a different approach. Instead of using dashboard cameras, Nexar offers a free iOS and Android app for your smartphone. The app packs in some useful features like collision detection, warnings and automatic hard-braking detection to record a car accident.
HERE needs to be mentioned as well. The mapping company is currently measuring down to one-inch precision over 30,000 street miles each week and up to a height of 130 feet, all over the world.
Are Crowd Sourced Maps the Answer?
Some of the most well-known map providers, like Google, Apple and HERE, use fleets of lidar-equipped cars which drive around to create high-quality maps. However, the end-result usually falls behind pretty fast in terms of accuracy because the maps can’t be updated daily.
To combat that, mapping developers are turning to crowdsourcing. Just like Nexar does with its smartphone app. Accuracy is of utmost importance if we want self-driving cars to ever be safe and efficient.
Lidar also has a disadvantage over crowdsourced images: size. Vehicles equipped with this technology gather so much data that uploading it into the cloud needs huge data pipelines. As you can guess, the cost would be simply too high.
On the other hand, crowdsourced maps for self-driving cars require no additional hardware installed. The same cameras which detect road obstacles can capture images of the roadside.
Because those camera images are less detailed than lidar images, they don’t generate a big data stream and can easily be uploaded to the cloud. Once there, the data would be analyzed, the map would be updated and the latest revision would be downloaded overnight to each self-driving vehicle.
However, all of it must be done in conjunction with ‘traditional’ lidar mapping. Crowdsourced maps would only account for changes and fresh information.
Wrapping It Up
Naturally, with this high amount of real-life data being poured into the cloud, privacy concerns arise. Luckily, all 3D mapping corporations and startups committed to never tracking a vehicle’s precise location for a lengthy period of time.
The next couple of months will also let us know if the strategies mentioned above actually work and high-quality, high-accuracy maps are being rolled out.
What’s your take on 3D mapping’s status and when it would be ready? Let me know in the comments section below.