The Future of Driverless Cars

Attribution: Automobile Italia, (CC BY 2.0)

You might have noticed more and more mention of autonomous cars recently—big companies like Tesla and Google are working on taking advantage of what many hope is the next big technological leap. Driverless cars come with many potential benefits including safer driving, but there are still issues to work out before making autonomous driving mainstream. One of them is basic reasoning.

Humans are remarkably good at navigating roads we’re unfamiliar with; even when using a GPS, we rely on observation and our senses to determine where we need to go. Driverless cars, however, can’t reason in the same way.  Part of the problem is due to how these cars navigate now: in every new area, they first use complex maps to analyze all new roads, then generate 3D scans to rely on. This process is time-consuming and not always reliable in more rural areas.

Researchers at MIT are working on this shortfall by creating a system that uses only simple maps and visual data to help driverless cars navigate routes in new, complicated environments. By creating an end-to-end navigation system, offering new maps available for cars to download, and creating an easier way to store and process maps, these cars should be able to navigate more easily in unfamiliar areas. Since car-to-infrastructure communication is essential for driverless cars to function smoothly, researchers are hoping this new system will make the process run more efficiently.

It turns out that even smart, driverless cars need to depend on human intuition to fill in the gaps that advanced technology can’t match! For more on MIT’s project, see the article here on Science Daily.