Actually, based on a talk that one of the Tesla autonomy engineers gave at Stanford last quarter, it seems like they do an approach similar to Boston Dynamics. At a high level, the end to end network interprets cameras feeds and outputs trajectory/path to follow. Then, lower level traditional controllers actuate the steering and drivetrain to follow that path
So you disagree with Teslas end to end learning approach?
Actually, based on a talk that one of the Tesla autonomy engineers gave at Stanford last quarter, it seems like they do an approach similar to Boston Dynamics. At a high level, the end to end network interprets cameras feeds and outputs trajectory/path to follow. Then, lower level traditional controllers actuate the steering and drivetrain to follow that path