As Tesla boss Elon Musk recently posted in his blog, the autopilot function will be available rather soon – by end of next year, it will be possible to cross the United States without the need to touch the wheel.
According to Tesla, the vehicles will get eight cameras for a complete 360°surround view – hitherto, the vehicles had just one. In addition, they will get a more powerful forward-facing radar and twelve short-range ultrasound sensors. To process all these data, the Tesla cars will get a new computing platform, 40 times more powerful than the existing one and capable of running neural network algorithms and thus be self-learning (according to sources familiar with the topic, the hardware will be provided by Nvidia which will replace Mobileye). Before activating the systems, the car will have to learn – in learning mode it will be able not only to learn from its own driver but likewise from all other drivers equipped in the same way; their data will be stored and kept ready in the cloud to be combined to a huge data base.
The new model 3, scheduled for production start in 2017, will already come with the new hardware. Likewise, the existing models will be equipped with the more powerful platform, effective immediately. However, the installed base cannot be upgraded, Musk wrote.
Though the cars will have the hardware, owners won’t be able to use it for the time being. This will only be possible after the learning process. Actually, the cars with the new platform initially will have less features than with the old one; they will lack automatic emergency braking, collision warning, lane holding and active cruise control. The good news: Once the learning process is completed and validated, these functions can be enabled, along with a number of additional yet unspecified functions that can be downloaded over the air interface. And yet another bad news: Though