Autonomous vehicles to leverage HD maps from space
Satellite imaging technology is an important next step for the autonomous drive space. Freddie Holmes speaks to Maxar to find out why
October 28, 2019
It may not be common knowledge, but the automotive industry is in deep discussions to find out how aeronautic technology can benefit the next generation of road vehicles. More specifically, satellite imaging firms are using their expertise to assist the creation of high-definition (HD) maps, which can optimise autonomous vehicle (AV) navigation, ride-share operations and last-mile delivery services.
Maxar is based in Westminster, Colorado. From here, it runs a global business you’ve never heard of, but will almost certainly have used. Its constellation of satellites circles the earth once every 90 minutes—that’s 16 revolutions per day—on what is called a sun-synchronous orbit. These satellites are roughly the size of a minivan; each satellite is essentially a giant camera floating in sky, with the lens aperture spanning over a metre in width. Currently, Maxar has five of them operating on a continuous low earth orbit, but the company’s imaging capacity is set to triple with the introduction of a new constellation of satellites in 2020. It is serious business: each satellite takes several years to build and requires a huge initial investment.
Back on earth, Maxar has dedicated teams that send commands up to these satellites every ten minutes or so. Rather than taking ‘snaps’ like a camera would, these satellites perform a thorough scan of ‘targets’—a city or a stretch of road, for example. This process forms the basis of Maxar’s Earth Intelligence business, which essentially involves mapping the entire world. “We’ve been doing this for a couple of decades. If you’ve ever used Google Maps, Apple Maps, Snapchat or Facebook, you’ve no doubt seen data from Maxar,” explained Kevin Bullock, Director of Business Development at Maxar. “They’re all using data from us.”
Space for a new supplier
In the automotive industry, there are a handful of players developing so-called HD maps. TomTom and HERE Technologies are perhaps the most prominent players, but various AV developers are using the sensors on their vehicles to help generate a detailed view of the world. In practice, HD maps can help to prepare an AV as it enters the unknown; if a long-range sensor cannot accurately judge how sharp an upcoming corner may be, for example, the map can ensure the vehicle slows to an appropriate pace.
If you’ve ever used Google Maps, Apple Maps, Snapchat or Facebook, you’ve no doubt seen data from Maxar
At Maxar, the team is leveraging advances in machine learning to optimise the next generation of these maps. “We can collect imagery on a global scale and create extremely accurate maps for AV precision requirements,” said Bullock. “We are working directly with automotive companies around the world, building out new datasets to fuel the autonomous technology revolution.”
Interestingly, this is not an entirely new space for Maxar. The company has served as a Tier 2 supplier in the automotive industry for around 15 years, and recently became a Tier 1. “We’ve been in the automotive supply chain for a long time, and that’s mostly due to our relationships with companies like HERE and TomTom,” said Bullock. “We have used satellite data to help them build out their street maps of the world. Today, we are gaining interest from the industry’s biggest automakers who want to use satellite data to build upon the existing data from their mapping partner.”
Cars vs satellites
The benefit of satellite imagery is that it is extremely accurate, and can be generated far quicker than by manually driving a fleet of vehicles around. “Most HD mapping has required fleets of vehicles that scan the streets with LiDAR sensors,” said Bullock. “That is time consuming and expensive, and when you look at roads all over the world, it’s a tough problem just in terms of scale.”
We can collect imagery on a global scale and create extremely accurate maps for AV precision requirements
In April 2019, Maxar announced it was working with Toyota Research Institute-Advanced Development, the automaker’s autonomous driving arm, as well as IT services specialist NTT DATA. The partnership aimed to create a digital HD map of the Tokyo metropolitan region using high-resolution satellite imagery. NTT DATA trains algorithms to label images, detecting road markings like lanes, stop lines and medial strips, which can all be recognised from space with Maxar’s high-resolution satellite sensors.
“Leading automakers want an autonomous vehicle to be able to navigate safely, whether it is on the freeway, on a residential road or even a dirt road heading into the mountains,” said Bullock. “We know the precise lane width and curvature of every road in the world, and so in addition to traditional HD mapping, manufacturers will leverage satellite technology to fill in the rest of the world.”
The new world map
It is not just the autonomous vehicle of the future that may benefit from satellite imaging. Ride-share and parcel delivery services could also be optimised by better understanding how the last few metres look when collecting a passenger, or finding an optimal drop-off location. In September, a popular ride-share company explained that it was using satellite imagery for this very reason.
Our satellites are not just mapping the major interstate freeways here in the US; we are literally mapping every paved road, every alleyway and every dirt road across the entire globe
With a large Minneapolis shopping mall as a case study, it found that most passengers were picked up and dropped off at the storefront, but that the maps only showed drivers how to get to the main road. “When a driver is on the main road, how do they find their passenger who’s waiting in front of the store? Those issues can cause ten-minute delays,” said Bullock. “The ride-share company figured out that if it could cut that down to one or two minutes, the operation becomes significantly more efficient; their drivers make more money and their passengers are happier because the pickup happened sooner.”
But even these operations are expected to eventually become driverless, and those vehicles will need to be able to find their way around without the safety net of an in-vehicle operative. Satellite imaging is by no means the answer to such headaches, but is another vital layer of data that should not be ignored.
“Heading into 2020 there is an awesome convergence of technology from space centres and AI that will ensure autonomous driving technology gets rolled out safely,” concluded Bullock. “Our satellites are not just mapping the major interstate freeways here in the US; we are literally mapping every paved road, every alleyway and every dirt road in the entire world.”