Navigation has been a core component of the automotive industry for decades. While the market today has progressed far from its paper map-based origins, for all their benefits and practicality today’s connected, even real-time maps may not cut it in an autonomous future.
Any commuter well versed in the nuances of GPS is likely to be aware of the flawed nature of existing navigation methods. Postal addresses can often leave commuters a block or more from their intended destination. Ongoing construction work can make certain locations entirely inaccessible. Road collisions and freak weather conditions can force last minute detours. While a human driver can react spontaneously and improvise their way to their 10 o’clock meeting, will an autonomous vehicle (AV) that is reliant on millimetre by millimetre detail to simply stay in lane be able to navigate with such freedom?
Another factor to consider is the scale of the task at hand. AV pilots today operate largely in small, geo-fenced locations which are scrutinised and recorded in minuscule detail prior to any AV making its way onto the road. Scaling this effort up to an entire nation’s worth of road networks, all of which can change on a daily or even hourly basis, is a monumental task. While on paper many of these changes will be filled in by data shared between increasing numbers of connected vehicles, this is still by no means a guaranteed process, especially when considering that the density of connected vehicles is likely to be lower in less built up areas.
In an industry where players have traditionally gone to great lengths to keep new developments secret, the sudden desire to cooperate highlights the value placed on having competent AV software to build upon
On top of this comes the conundrum of whether an AV needs a map at all, a question that, to a degree, has split the industry. While many believe in the need to have high definition (HD) mapping others, perhaps most notably Elon Musk, argue such dependency will result in brittle AV systems. Instead, the naysayers stress the value of artificial intelligence (AI) and neural networks. Certain experts also follow a similar motto. For example, researchers at MIT are opting for a more minimalist approach, where AVs use information gathered from sensors to spot the differences between their current surroundings and previous topological map knowledge.
For those in the majority, however, the need to revolutionise the map is rapidly becoming big business. Mapping companies new and old have already embraced the challenge of building a map for the AV with technology companies and traditional automotive players beginning to play their cards. The German trio of Daimler, BMW and Volkswagen, for example, have already placed their trust in HERE via a joint majority share-ownership of the company. In Japan, Toyota, Nissan, Honda and Mazda have snapped up HERE rival Ushr, under the guise of the Dynamic Map Platform consortium. In the US, GM and Ford’s data collection programmes are also well under way.
While a human driver can react spontaneously and improvise their way to their 10 o’clock, will an autonomous vehicle (AV) that is reliant on millimetre by millimetre detail to simply stay in lane be able to navigate with such freedom?
It is these collaboration efforts that underline the importance of the task at hand. In an industry where players have traditionally gone to great lengths to keep new developments secret, the sudden desire to cooperate on such a potentially valuable product highlights the importance placed on having competent AV platforms to build upon.
Indeed, many players have already concluded that mapping quality could very well directly correlate to AV competency in the future. As discussed in Automotive World’s special report, ‘Mastering the autonomous vehicle map’, AV maps are likely to be one of the most exciting and innovative areas of automotive industry development over the years to come.
