Tesla CEO Elon Musk stated that the company’s autopilot systems are being upgraded to 4D, which will improve the performance and capabilities of the semi-autonomous driving feature. Currently, the autopilot is operating with “~ 2.5D,” Musk said.
The developments stemmed from a question asked by a Musk follower on Twitter who is having trouble using Tesla’s invocation feature in his driveway. Summon allows owners to retrieve their vehicles using their smart phones. By holding down the “COME TO ME” button within the Tesla app, the car will use GPS vectoring to travel to the phone’s location.
Summon is part of Tesla’s complete autonomous driving suite. But, the rework of the autopilot dimensional upgrade is apart from something much bigger. Perhaps this is a complete rewrite of the autopilot that will expand the company’s FSD capabilities.
We need to finish updating the autopilot to 4D vs. ~ 2.5D, then it will go up steep slopes
– Elon Musk (@elonmusk) July 22, 2020
However, the owner stated that the slope of his driveway is slightly steeper than 10%, preventing the vehicle from traveling on roads with slopes. The steepness of the grade, along with the normal transitions from a street to a driveway, can present problems for the Tesla autopilot. This could be due to the lack of information that the Tesla Neural Network has to navigate in these environments.
That said, Tesla is developing a 4-dimensional system for autopilot. Developing new elements to help the autopilot understand the vehicle’s road environment and surroundings could lead to more drastic improvements and an increasingly accurate understanding of the roads a vehicle is driving.
Tesla’s current autopilot package uses ~ 2.5D, Musk said. Now, use two-dimensional images along with tags, which could explain the around 2.5 dimensions that Musk talked about in the tweet.
Adding dimensions to the autopilot system will simply increase the precision of how the car reacts in certain situations. Tesla currently uses images from autopilot cameras that are tagged with information. Tesla could use three-dimensional stereoscopic scenes that are reconstructed from video, along with time stamps to improve accuracy.
Some members of the Tesla community put their two pennies into what the 4-dimensional autopilot system might involve.
Reddit user u / __ TSLA__ stated that curating a massive series of scenarios and traffic objects that a car could encounter while driving could improve the accuracy of the autopilot and Tesla auto-driving capabilities.
However, another Reddit user, u / Semmel_Baecker, said 4D could mean that autopilot cameras could build a 3D environment in real time and then predict the movements of 4D-tagged objects based on past behaviors. from other vehicles. This strategy would effectively use the Neural Network to learn the reactions of other controllers or objects.
Tesla continues to develop autopilot behaviors to eventually launch a “complete” full self-driving suite in the future. The electric car maker continues to release patents that aim to build a more accurate autopilot system that will accelerate the company’s journey to Tier 5 Autonomy, which Musk says will be arriving soon.
More recently, Tesla filed for a patent entitled “Enhanced Object Detection for Autonomous Vehicles Based on Field View”, which would crop important objects in images and increase the resolution of those images. If pedestrians, vehicles, or other objects are available in an image, they would be available at a higher resolution to improve the accuracy of the autopilot.
Tesla’s exact plans for an autopilot upgrade to 4-dimensional images are unknown. The highest precision is necessary for the company’s cars to drive in any environment. Tesla will soon launch FSD’s “Driving on City Streets” feature, which will complete the suite.
