Welche Strategie verfolgt Amazon beim Autonomen Fahren? Interessanter Beitrag dazu.
A recent article covers the details of the Zoox acquisition, and posits that an important driver was Zoox’s expertise in computer vision. More broadly, it’s the ability to use information from multiple sensors like optical cameras, LiDAR, radar, and GPS, along with artificial intelligence and contextual information to perceive the environment and make decisions.
Computer vision and perception are critical for safe AV operation. AVs are potentially big business, but Amazon’s business objectives are larger. Although the initial announcements are all about Zoox continuing on their path to provide people movement, the goods delivery business is probably a bigger and more immediate driver for Amazon – automating long-distance and last-mile trucking and delivery significantly improves profitability in their core business. Computer vision and perception enables this, as it does other Amazon initiatives like drone delivery, robotic movement in warehouses, home security products like Ring, and un-womaned stores like Amazon Go. Potentially, a 3D perception-enabled Alexa could add even more excitement to your home life than it does today! I think Amazon shopped well – the $1.3B acquisition of Zoox (with their talent and IP) looks like a bargain.
Like its competitors in the AV space (Waymo, Argo, Cruise, Aurora, and Yandex), Zoox uses LiDAR as one of the key sensors in their perception suite. However, there is a difference – Zoox procures their LiDARs from external suppliers. Waymo (Google’s AV arm) and Yandex have their own suite of internally developed LiDARs (Google even fought an exciting legal battle with Uber over LiDAR not too long ago). Argo, GM-Cruise and Aurora bought LiDAR companies (Princeton Lightwave, Strobe and Blackmore respectively). They realized that 3D sensing was intimately connected to the quality of the perception suites that they were playing a bigger game in. Direct control over the design, architecture, and features of LiDAR was critical for higher performance and lower latency of the AVS software stacks. I think AmaZoox will need to do the same, not just to support the AV effort, but also the larger efforts discussed above.