Tesla’s Autopilot is able to steer in the snow
The Autopilot wasn’t meant to be used in those conditions and it is still not recommended officially, but it was designed to be able to continue working for brief period of times where lane markings were fading or to give time to the driver to take control.
During the release of the first Autopilot update, CEO Elon Musk gave the example of fading lane markings on the 405 in Los Angeles where the Autopilot was working.
The company has been improving on the system ever since and the latest versions have resulted in some decent performance without markings – even in the snow. A new video posted today by a Model S owner shows Autosteer keeping its path despite the vehicle often not detecting the lanes or even having a lead vehicle:
You can see on the instrument cluster that sometimes the Autopilot is detecting the curb as a lane indicator, but only sporadically and despite that, it is not prompting a “Take Control" alert.
There has been speculation about how Tesla’s first generation Autopilot can manage that without the trifocal front-facing camera of the second generation Autopilot or the ‘Tesla Vision’ image processing system.
As it’s often the case with Autopilot, data is key. Tesla is building high-resolution maps using its fleet of vehicles and those maps can help position the vehicle based on data from the GPS and inertial measurement unit (IMU), but there are restrictions on the use of those maps, especially in Europe and China.
Nonetheless, Tesla can still use the fusion of the data from all those sensors in order calculate inertial dead-reckoning, which is the process of calculating the current position by using a previously determined position, and determine a path based on previous trajectories used on the same road by Tesla’s fleet.
Again, the use of the Autopilot is still not recommended without proper lane markings. That will be for the second generation Autopilot when it will be ready to achieve level 4 and 5 autonomy.