Last month, one of Google’s fleet of Lexus RX450h autonomous self-driving cars hit the headlines when it had its first autonomous crash, crashing with a public bus along El Camino Real in Mountain View, California. The car, which had been driving in the right-hand lane as it approached the Castro St. intersection, assumed the bus — which was one lane over — would wait for it to drive around sandbags positioned around a storm drain in the right-hand lane.
It didn’t, teaching Google’s autonomous drive system the lesson that most of us learn very quickly in our driving careers: give vehicles that are larger and heavier than you a wide berth — and expect the unexpected.
Since then Chris Urmson, head of Google’s self-driving car program, has said Google has been working hard to ensure that its autonomous driving software now understands a little better what to do when they come across a bus. Given the coverage of the incident, you’d be forgiven for thinking that Google hadn’t ever thought about how it would interact with a bus on the public highway — but it turns out that Google has been working on the program for many years.
Last week, nearly two years after it submitted a patent with the U.S. Patents and Trademarks Office specifically designed to help it detect and react to busses, Google was granted said patent.
Fate, it seems, has a terrible sense of humor.
As Future of Transportation reports (via Re/Code) Google’s just acquired patent was first applied for back in March 2014, and was principally to help a Google autonomous vehicle to correctly identify and react to a school bus. Since every state in the U.S. requires drivers to stop and wait for a school busses to pick up or drop off children and pull back out into the flow of traffic before they can move, it’s important that an autonomous car does the same.
Consequently, the patent discussed the ways in which Google’s autonomous driving system could identify a school bus, ranging from size comparisons and vehicle behavior analysis to image recognition and of course, color matching the bright yellow paint that every school bus wears.
Sadly, it’s not clear if Google had already implemented its (now patented) school bus detection software in its cars at the time the collision occurred between one of its cars and a public transit bus. Either way, we’re guessing further patents will be seen in the coming months as Google works hard to implement its latest experience into new software designed to ensure it doesn’t happen again.
There’s one more thing that Google’s new patent reminds us. While the majority of driving duties are relatively straight forward to us humans, teaching artificial intelligence systems to drive is a particularly tough thing to do. While autonomous vehicle may spot things we wouldn’t and be generally far safer behind the wheel than humans in most situations, there are a whole lot of subtleties that autonomous cars have yet to learn. From understanding the importance of human eye contact at a crosswalk to dealing with the four-way stop, things that we might think of as easy are anything but for autonomous vehicles.
To that end, we’d like to hear suggestions in the Comments below of things that you think of as easy behind the wheel that autonomous cars will find difficult in the future — and how you think companies like Google, Nissan and Tesla will teach cars to master them.
You can also support us directly as a monthly supporting member by visiting Patreon.com.