This year's CES had plenty of self-driving concepts from traditional automakers, but it also saw non-automotive tech companies start inserting themselves into the autonomous vehicle game, like Nvidia lending its self-driving computing setups to both Audi and Mercedes. But by focusing on in-house development instead of using lent tech, Waymo can more tightly integrate its sensor hardware, software and image recognition. The company also claims its network achieves better resolution and performance than the hardware they were previously using.
There are other benefits to dedicated internal tinkering, not least of which is their team's experience looking beyond current sensor setups. While most autonomous vehicles rely on lone roof-mounted medium-range LiDAR, Waymo innovated new long- and short-range sensing units for their Pacifica minivans to make a more complete awareness window around the vehicles. And while initial R&D is always pricey, the company's research efforts have slashed the cost of a single high-performance LiDAR by 90 percent from its initial $75,000 price tag when Waymo first started its self-driving development.
The first Pacificas equipped with the new in-house sensor suite will hit California and Arizona roads later this month, extending Waymo's lead in self-driving distance driven. The company already has 2.5 million miles under its belt, and will hit 3 million in eight months, Krafcik said.