Fresh off settling a lawsuit over former Tesla employees taking proprietary information and trade secrets regarding warehousing, logistics, and inventory control operations, Zoox has released a video showcasing their autonomous driving technology.
The video, which takes place in and around the streets of San Francisco, shows a Toyota Highlander equipped with their self-driving technology take a ~1-hour drive through busy downtown streets. Their self-driving technology, which they call their perception system, includes 18 cameras, 10 radars, and 8 lidars.
In the upper half of the video, which shows us the real-world image with objects being identified in different colours, there is a notation underneath each object to show which system detected and classified the object. The notations are small, but from what we can see almost all object were detected by lidar, which as we know Tesla CEO Elon Musk thinks is not the way to go for autonomous driving.
This complex sensor stack allows the vehicle to detect and track about 100 vehicles and pedestrians simultaneously at a range of over 130 meters (426ft) away.
According to Zoox, the nearly 1-hour drive was 100% autonomous, and also with 0 disengagements (there was a driver behind the wheel for safety reasons), claiming it to be “the first autonomous driving video of its kind…ever” with full perception output.
The video is narrated by Zoox Senior Director of Perception and Prediction James Philbin, and Director of Computer Vision Sarah Tariq.
We get to see the retrofitted Highlander deal with some complex and difficult driving situations, such as overtaking a row of three double-parked cars, right turns on red lights, a left turn at a six-way intersection, and navigating through a tunnel without the use of GPS.
To be able to complete tasks like these, Zoox’s perception system identifies things like vehicle lights, parking lane occupancy, road position, and the behaviour of other traffic in the area.
Tariq says Zoox’s perception system can detect, classify, and identify even more behaviours and obstacles than what it encountered during the drive. She claims the system can detect whether pedestrians are standing or walking, which way they are facing, whether they are looking at their phones, or even riding an e-scooter.
Eventually Zoox hopes to move from retrofitted Highlanders to small, driverless, custom electric cars that will operate in a shared fleet.
Their design calls for four-wheel steering, dual power trains and dual batteries, which they claim will have a combined capacity greater than more EV’s today (via VentureBeat).
Check out the full video below, and let us know what you think in the comments.