From the traditional road atlas to the Waze app, maps today are an essential part of a trip. Specifically, they help drivers plan – or simply follow – a visual route from A to B. But take away that driver from the equation, as autonomous technology promises to do, and maps suddenly need to play a different role. It is a role contingent upon smart and connected technology, and without it, autonomous vehicles will never find their way to market.
TomTom specialises in maps and increasingly in maps for automated vehicles, but it knows that maps alone won’t build an autonomous fleet. “We see three core building blocks to a driverless future,” said Tomaso Grossi, Senior Product Marketer at TomTom Automotive. To start with, the vehicle needs sensing capability, consisting of what it can see through the sensors, such as cameras, radar and LiDAR. A second pillar is what Grossi refers to as ‘driving policy’, the instructions to the vehicle regarding how to behave in different driving situations. The third is mapping, but the maps are far from average.
“With automated driving, the map has a different function. You are no longer looking for a map to get you from A to B; it’s more of a map that helps you with localisation, perception and path planning,” he told Megatrends.
Individuals tend to take it for granted that they know exactly where they are, i.e. in the middle of the road, in a specific position within the lane, at roughly a specific latitude and longitude. An autonomous vehicle does not necessarily know where it is located on the road. GPS positioning, with an accuracy level of about 5 metres, won’t cut it for autonomous driving. “A high definition map allows you to enable accurate positioning at decimetre level, whereby the vehicle can see where it is in the lane,” said Grossi.
With automated driving, the map has a different function. You are no longer looking for a map to get you from A to B; it’s more of a map that helps you with localisation, perception and path planning
A handful of different techniques exists on the market for localisation, some of which involve using sensors to correct GPS measurements. TomTom tried and tested this approach but found it very expensive and not necessarily scalable. “While it might be scalable for a small track or a quick demonstration, these technologies are difficult to scale for global implementation in vehicles on the road. We strongly believe that providing an HD map and localisation layers represents a much more scalable approach,” he explained.
This is a question that OEMs will need to address for many different technologies. As Grossi added: “Sure, some technologies might be feasible, and might even be the right technology for short-term implementation, but when you are looking at global implementation, your mind needs to be focused on the scalability of the technology.”
Perception and path planning
Perception is the second key role maps will play in autonomous vehicles, as highlighted by recent white hat hacks. Researchers at the University of Washington, University of Michigan, Stony Brook University, and UC Berkeley figured out a way to trick the vision systems used by an autonomous car by putting stickers on street signs. The stickers disrupted the way in which the machines read and classified signs. For example, a stop sign could be misinterpreted as a 45mph sign, leading to disastrous safety consequences.
While it might be scalable for a small track or a quick demonstration, these technologies are difficult to scale for global implementation in vehicles on the road. We strongly believe that providing an HD map and localisation layers represents a much more scalable approach
“As a human, if you see a stop sign with stickers on it, you still know that it’s a stop sign and that you should stop there. A vehicle could interpret it as something different,” Grossi observed. “If those stop signs were also represented in a high-definition (HD) map, it would help address that issue. It aids perception by also providing redundancy to the car sensors.” Essentially, explains Grossi, the vehicle thought-process would run as follows: This looks like a stop sign but it’s not clear – what does the HD map say? The map confirms that it is indeed a stop sign, and so this vehicle must stop.
Path planning is the third main role for maps of the future. This tackles the question of how the vehicle knows exactly in which lane it needs to be in order to travel from one point to another. “This is not just navigation as in, ‘take a right at the next exit’. It is also recognising that the vehicle needs to turn into the right lane now if it wants to be able to turn right off the highway, for example,” said Grossi. “This is the difference between the use of maps for navigation and the use of HD maps for autonomous or automated driving. This is the pillar where TomTom is specifically active.”
If these are the building blocks of autonomy, what does the navigation experience of the future look like with such maps supporting it? TomTom recently conducted a user experience exercise to examine ways of bridging normal navigation and autonomous driving.
As a human, if you see a stop sign with stickers on it, you still know that it’s a stop sign and that you should stop there. A vehicle could interpret it as something different. If those stop signs were also represented in a high-definition map, it would help address that issue
In general, as automation increases, so will the navigation options. Drivers already have the choice of the shortest route, the fastest route and the most economical route, but there could be more. Specifically, TomTom envisions an ‘effortless route’, which is a route that includes the most automated driving time possible. “The route might take longer but the occupant would be able to be more productive during that time,” Grossi explained.
This sort of experience will require a leap of faith from vehicle occupants – faith that the vehicle knows what it should do and that it will do exactly that. Drivers have already been building that trust through the automated features available at the moment. “When we introduced the portable SatNavs, the key there was to build trust in a small box, a small piece of hardware that was telling people to go left or right,” said Grossi. “We were building trust in innovation in this little machine.”
Trust in the machine
The next step involved building other services like traffic, parking or weather for automotive clients. “Drivers have begun to trust a system that tells them to go in a different direction than usual because there might be traffic or there might not be parking available. That was another level of trust. Now with our autonomous driving products, we need to build trust into this new technology,” he said. “You could see it as a path towards building trust in innovation at different stages – every time there’s a bigger gap.”
Drivers have begun to trust a system that tells them to go in a different direction than usual because there might be traffic or there might not be parking available. That was another level of trust. Now with our autonomous driving products, we need to build trust into this new technology
One way to build trust is to share the information that the car has with its occupants. “When an automated vehicle changes lanes, slows down or accelerates, occupants are not always aware of why it is doing this. We thought of having different screens that show what the vehicle is seeing,” he said. Notably, what the vehicle is seeing doesn’t necessarily mean only what the sensors are telling the vehicle, but also what the on-board map is telling the vehicle. This includes such information as the number of lanes and the road geometry of these lanes. “We thought of it as overlaying the TomTom HD map onto the road or, at the same time, overlaying what the sensors are seeing onto a screen, for example,” he added.
This technology has not yet been implemented, but TomTom is conducting such user experience exercises with an eye towards the future. “It was more of an exercise of understanding how we would want to portray our technology to make the passenger feel safer and more informed,” Grossi clarified.
While his team is specifically focused on product marketing for TomTom’s unit on autonomous driving, the realisation of Level 5 autonomy won’t come any time soon. As Grossi cautioned: “What we have now is more driving automation than autonomous driving. The very end goal of driverless autonomy, of robo-taxis, is still very far away in the future.”
This article appeared in the Q3 2017 issue of Automotive Megatrends Magazine. Follow this link to download the full issue