Today, Google stated that their self-driving cars are now ready to navigate streets in major cities. This is a much more complex challenge than the interstate-centered autonomous driving project Google has been developing.
Since its initial launch in 2009, this long-term research initiative has logged around 700,000 miles on their autonomous cars. The vehicle Google has preferred to use is a stock Lexus RX450h, albeit with 24 motion sensors mounted on the outside. Since the last time Google reported its project status, the vehicle has been accident free.
Google stated that the shift from highway to city driving one year ago was due to the fact that urban areas are where many reside and work. A statement issued from the company explained, “It’s really important to be able to drive on city streets if we’re going to make meaningful progress toward making self-driving cars a reality.”
The company stated that researchers have driven thousands of miles on the suburban streets of the small community of Mountain View, California – 35 miles South of San Francisco where Google is headquartered. Currently, Google’s autonomous vehicles rely on a network of video cameras, radar sensors, laser-guided mechanisms and an expansive database of information composed from normally driven cars to aid the vehicle in navigation.
To compensate for the much busier city streets, Google upgraded the software within the vehicle – adding more recognizable objects like cyclists and pedestrians. Google stated that the improved sensors can now respond better to external stimuli and understand complex maneuvers. The sensors on the Lexus now have the ability to “read” street signs when they appear unexpectedly, such as a crossing guard raising a stop sign or a lowering gate at a railroad crossing.
To create the car’s “sense” of driving, Google added thousands of driving situations to its already complex computer model. Its vehicles can now recognize situations, comparing what is occurring in real-time with what should be happening in a normal scenario to respond in an appropriate manner. In other words, if a car is approaching a four-way stop intersection at a higher than normal speed, Google’s models can calculate the probability that the approaching vehicle will come to a normal or abrupt stop or speed through the intersection.
Even more fascinating, Google stated that the vehicle’s software includes defensive driving habits. With defensive driving maneuvers, the vehicle can detect whether it is in another vehicle’s blind spot as so to stay out of it and to move away from larger trucks and motorcycles and scooters weaving through traffic. Furthermore, the cars also pause for 1.5 seconds while stationary at a stoplight when it turns green to avoid other drivers who may run the red light.
While it may seem that Google is on the fast track to creating the first fully functional autonomous passenger car, they face a plethora of legal and regulatory obstacles to getting permission for their self-driving cars to remain a permanence on city streets and highways. Moreover, Google is still tasked to tackle tougher driving scenarios: more complex urban streets, construction zones, and rerouted traffic. Google also noted that they have to calculate for scenarios involving more human notions, such as social signals for changing lanes and merging.
“We still have lots of problems to solve,” stated Chris Urmson, director of Google’s Self-Driving Car Project. He explained that the company has made much headway in the last two years. Driving situations that would have dumbfounded researchers hitherto recently, have been easily navigated using the new software in Google’s autonomous vehicles.
Google’s co-founder Sergey Brin stated in 2012 that this technology wouldn’t be made readily available to the public for at least five years. Nonetheless, it seems that Google is leading the automotive industry’s push to create autonomous vehicles by taking its project to the city streets.
By: Alex Lemieux