Tesla robot cars aren’t coming quite yet

Tesla’s plans for a fully driverless ‘cybercab’ in 2026 is optimistic to say the least. This is not because the technology is not advancing, as Honda and Mercedes are both demonstrating.

Tesla is also planning an unsupervised version of its Full Self Driving software next year for its existing vehicles, and Waymo and Cruise have been offering commercial ride hailing services with driverless cars for the last year or so.

Nor is it about the price. The target price of $30,000 is welcome, as it indicates the falling price of electric vehicles and is largely driven by the predicted reduction in the cost of the battery pack.

Even the higher cost of a robotaxi with multiple cameras, lidar laser and ultrasound sensors is affordable on the basis of 18 hour a day operation 7 days a week, and the cybercab will have wireless charging to automate the charging process for fully automated operation.

The issue with the cybercab is that it will not have local controls. There will be no steering wheel or pedal(s), so there will be no options if there is a problem. This is currently solved with remote operators, which will be vital for a company such as Uber using the cybercab.

This same remote operation will have to be behind the proposed robovan which can carry up to 20 people.

The industry has actually seen many of these on private roads already. Autonomous shuttles have ferried passengers across Paris for tennis competitions and on public roads in Norway and Hamburg, so this is not new. The styling of the robovan is straight out of the 2005 film Robots, which is appropriate given the launch on a movie set in Los Angeles by Tesla on Thursday night. But the robovan actually has a better chance of rolling out commercially but faces more mature competition from Amazon’s Intel-based Zoox shuttle.

These will undoubtedly share the same Full Self Driving (FSD) AI inference chip developed at Tesla with a focus on performance per watt, although the benchmarks are not available.

But the limiting factor are human. The driverless shuttles are obvious, and will be run by large organisations, including Renault.

Offering vehicles without steering and pedals to consumers is a regulatory nightmare, and the regulators are not yet equipped for this. It will take several years for the cybercab to gain regulatory approval, even if the vehicles shown are the production version rather than protypes, which is unlikely.

Regulators are already wary of Tesla’s tendency to release software that needs improvements, and coupled with the lack of controls there will be significant regulatory interest.