Are We Out Of The Woods?

Are we in the clear yet? — These are the big questions in the autonomous car industry today. Week after week we read about new deals and investments that will eventually help cars become driver-less. But until it happens, we will experience lots of trials …and errors unfortunately.

The fatal Tesla accident that caused one man’s death in May is the latest such example. The self-driving mode was not able to distinguish between a white truck and a bright sky in that case, according to Tesla, and resulted in the tragic death of the passenger.

Weeks later, Tesla terminated its partnership with Mobileye — the self-driving software/hardware provider — and announced it will start developing its own autonomous driving technology.

Earlier in the year, GM invested $500 million in Lyft, and the two companies said they’d be working on developing an on-demand, self-driving car system that could be available as soon as 2019. Meanwhile, Uber partnered with Volvo and the two are launching a pilot with the Volvo XC90 SUV in Pittsburgh this month. In a joint venture, Uber and Volvo are also investing $300 million to develop a fully autonomous car that will be ready for the roads by 2021. (Disclosure: GSV owns shares in Lyft).

At the same time, Ford announced it is partnering with Baidu to invest $150 million intoVelodyne, a major supplier of self-driving car technology. Velodyne’s LiDAR is a light-sensitive radar mounted on the top of vehicles, and it provides a 3D view of surroundings of up to 200 meters away. LiDAR’s price tag has come down significantly over the years, from $80K to now $8K, and the new investment should further reduce the cost. Ford executives believe cars with this new technology will be ready for the road by 2021.

But some places already have driver-less cars. Last week, Singapore launched the world’s first robo-taxi service with cars operated by nuTonomy — a Cambridge, MA-based startup, backed by Highland Capital and EDB. The trial allows users to hail a taxi and get a free ride on a 6km street network. nuTonomy is also testing self-driving cars in the U.S. and the U.K., where it collaborates with Jaguar Land Rover. Users in Singapore can now call a ride just like Michael Knight did with KITT 2000… Thirty-four years later, reality meets Hollywood!

The single biggest challenge for self-driving technology is “recognition.” As in the Tesla accident, the software needs to recognize the street, the signs, moving objects, people, cars, what’s approaching, etc. Everything else is taken care of in a manner of split seconds. But if the system cannot identify between a bright sky and a white tractor, it will not compute.

Especially troublesome are street boarders/marks, according to Baidu Chief Scientist Andrew Ng. It is absolutely essential for the car to know its position on the street. But in many places around the world, street marks are not obvious, or even missing, which causes errors in the self-driving system. Think of third world cities, or off road driving… That’s why well organized places like Singapore are among the first ones to now offer self-driving services.

As published in this week’s A 2 Apple: