When electric-car maker Tesla Motors (NASDAQ:TSLA) launched its next-generation suite of Autopilot sensors in October, there was one immediate caveat in the big update: Tesla vehicles with the new hardware would temporarily lack key Autopilot features already in cars with its first-generation sensor suite. Missing abilities even included important safety features like automatic emergency braking and collision warning. Tesla said it would need to further calibrate the system before it caught Autopilot 2.0 up with Autopilot 1.0. But the new system is almost ready, according to a tweet from Tesla CEO Elon Musk on Wednesday.
"Tesla Autopilot vision neural net now working well," Musk said on Twitter, referencing the new system's onboard computer, which has 40 times the computing power of its predecessor and runs the company's Tesla Vision neural-net software (developed in-house) for sonar and radar processing. "Just need to get a lot of road time to validate in a wide range of environments."
In late November, Musk said Tesla would begin rolling out updates for its new Autopilot sensors to catch the new technology up with the capabilities of its existing sensor suite in December. The updates, Musk tweeted, "will get rolled out incrementally in monthly releases."
Tesla's first-generation Autopilot still includes a long list of driver-assist features that Autopilot 2.0 doesn't have yet, including parking assist, automatic emergency braking, blind-spot detection, auto-steer, auto-park, summoning, side-collision warning, and more.
Earlier this month, Musk said the "Tesla software team is working seven days a week to complete testing and validation" and that the team was "[g]etting close."
How Tesla goes about validating a totally new set of sensors using a new computing system will make an intriguing story. If the company really can begin catching up its new sensor suite starting in December, it would highlight Tesla's ability to quickly generate and organize significant real-world data and rapidly deploy improvements to its fleet.
When Tesla began shipping its vehicles with its first-generation Autopilot sensors in 2014, it took a full year before the company was ready to beam an over-the-air software update to its fleet for automatic steering. Bringing significant features to Autopilot 2.0 this quickly, therefore, would put the spotlight on the company's growing prowess in generating useful data from its sensors, and deploying software updates to take advantage of them.
While Musk's reference to needing "a lot of road time" is vague, the CEO's previous promise to start catching up Autopilot 2.0 to 1.0 in December implies he doesn't expect it to take much longer. With fourth-quarter deliveries -- all of which include new Autopilot sensors -- expected to be up about 44% year over year (and equal to about half of the company's total deliveries last year), Tesla's higher sales volume undoubtedly helps by giving the company more vehicles to collect data from. But the company's growing experience in organizing vehicle data, and Autopilot 2.0's more powerful computer and software, are likely also key catalysts in helping Tesla refine the system.
Over time, Tesla expects its Autopilot 2.0 capabilities to far exceed Autopilot 1.0. Tesla says the updated cars have all the sensors they need to totally drive themselves in the future. Of course, to get to this point Tesla will need to continue to validate and refine the system. In addition, Tesla will need to obtain regulatory approval to enable its cars to drive themselves.