It’s no secret that amongst its many cooky and futuristic projects, self-driving cars are something Google hopes to make real within the next few years. Late last month, Google’s fleet of autonomous automobiles reached an important milestone. After many years of testing out on the roads of California and Nevada, they logged well 0ver one-million kilometers (700,000 miles) of accident-free driving. To celebrate, Google has released a new video that demonstrates some impressive software improvements that have been made over the last two years.
Most notably, the video demonstrates how its self-driving cars can now track hundreds of objects simultaneously – including pedestrians, an indicating cyclist, a stop sign held by a crossing guard, or traffic cones. This is certainly exciting news for Google and enthusiasts of automated technology, as it demonstrates that the ability of the vehicles to obey the rules of the road and react to situations that are likely to emerge and require decisions to be made.
In the video, we see the Google’s car reacting to railroad crossings, large stationary objects, roadwork signs and cones, and cyclists. In the case of the cyclist — not only are the cars able to discern whether the cyclist wants to move left or right, it even watches out for cyclists coming from behind when making a right turn. And while the demo certainly makes the whole process seem easy and fluid, there is actually a considerable amount of work going on behind the scenes.
For starters, there are around $150,000 of equipment in each car performing real-time LIDAR and 360-degree computer vision – a complex and computing-intensive task. The software powering the whole process is also the result of years of development. Basically, every single driving situation that can possibly occur has to be anticipated and then painstakingly programmed into the software. This is an important qualifier when it comes to these “autonomous vehicles”. They are not capable of independent judgement, only following pre-programmed instructions.
While a lot has been said about the expensive LIDAR hardware, the most impressive aspect of the innovations is the computer vision. While LIDAR provides a very good idea of the lay of the land and the position of large objects (like parked cars), it doesn’t help with spotting speed limits or “construction ahead” signs, and whether what’s ahead is a cyclist or a railroad crossing barrier. And Google has certainly demonstrated plenty of adeptness in the past, what with their latest versions of Street View and their Google Glass project.
Naturally, Google says that it has lots of issues to overcome before its cars are ready to move out from their home town of Mountain View, California and begin driving people around. For instance, the road maps needed to be finely tuned and expanded, and Google is likely to be selling map packages in the future in the same way that apps are sold for smartphones. In the mean time, the adoption of technologies like adaptive cruise control (ACC) and lane keep assist (LKA) will bring lots of almost-self-driving cars to the road over the next few years.
In the meantime, be sure to check out the video of the driverless car in action: