This is how bad Tesla’s Autopilot is and how it might greatly improve
While Tesla is burning all the bridges with its Full Self Driving technology, plain Autopilot is lagging.
There haven’t been major updates for several years already, and some significant changes, such as not using radar anymore, have even worsened things.
Autopilot technology is outdated. It still relies on old software and doesn’t take advantage of all the new neural network training that Tesla has been deploying for its FSD program.
The camera-based Autopilot system is falling behind the competition. Other brands can maintain speed on the highway and stay in their lanes with at least the same reliability as Tesla vehicles.
The system designed for highways and roads still lags in some situations and is not outstanding on its own.
Autopilot indeed offers an enhanced package where the car can change lanes and leave highway exits. However, the system is far from perfect, as it still relies on outdated software.
FSD highway stack
A few weeks ago, Tesla released version 11 of its FSD software. Its major update was the merging of the highway and city streets stacks.
Until then, the highway stack consisted only of the old Autopilot, leaving the most advanced system for city streets.
This update has changed the way the car behaves on the highway. Instead of relying on old heuristic programming, the car now follows neural network training.
This results in a more natural and smooth driving experience, improving not only safety but also user comfort while cruising and changing lanes.
The question is whether Tesla will deploy this stack to Autopilot users once it has been adequately tested. Improved highway behavior would certainly be welcomed.
Driver monitoring system
Another issue that needs to be addressed is how long drivers will need to hold the steering wheel physically while Autopilot is engaged.
It is still necessary as we may experience some erratic movements from the system, especially…