09 November 2017
Waymo, an autonomous car development company, and a subsidiary of Google’s parent company took some time at Google I/O 2018 and discussed self-driving cars. John Krafcik, Waymo CEO explained first how Google’s self-driving cars project was born.
Later, Dmitri Dolgov, CTO, and VP Engineering- Waymo joined the stage and explained at Waymo that now AI touches everything from Percept to Prediction, Decision-making and Mapping as well. The self-driving cars need a deep semantic understanding of what's around them. The vehicles need to detect the obstacles and avoid them.
Dolgov explained the areas where AI has its huge impact and is helping the self-driving tech to become a secure, sustainable, easier, and accessible mode of transport. The idea is to build driverless cars safer for pedestrians. At the event, Dolgov demoed how the self-driving car with the power of AI was able to detect the pedestrians in obscure situations. For instance, a construction worker peeking out of a manhole with most of his body obscured.
In another example, Doglov showed a man that has his half covered by a plank of wood. Or, something like a pedestrian who’s dressed in an inflatable dinosaur costume and crossing the road. Waymo explains that its self-driving cars are able to detect such instances and classify them as the deep nets have been applied to a combination of sensor data. The cars not just have cameras but also the lasers to measure distance and shapes of the objects and also the RADAR to measure their speed. Waymo applies the machine learning to the combination of sensor data so that the self-driving cars can accurately detect the pedestrians in all forms.
Waymo self-driving cars with the help of machine learning are even able to detect what could come at the intersection point and to stop when a car is crossing the lane when it’s a red light for it.
According to Waymo, it can perform these functions in real-time, even in bad weather. Waymo put an example where the self-driving car was able to perform in a snowy weather. It became possible with the machine learning. But, that’s not it, Waymo is working on such kind of vision problems that could hinder the sensor data. Once they achieve the success in such advanced challenges, the self-driving cars will be made available to more countries and then we may not be too far from the “truly autonomous vehicles” era.