Tesla Cars Use Supercomputers to Develop Self-Driving
You may not currently buy a fully self-driving car. But Tesla car manufacturers around the world are racing among themselves on who will be the first to produce this type of car commercially, yet two companies do not agree on the same technical approach to achieving this goal. Some employ remote sensing techniques such as LiDAR or LiDAR, while others rely on radar-based sensors to help identify obstacles that are difficult to see on the road.
What Tesla Car Company is Working On?
Companies working on self-driving car technologies use a combination of lidar, radar, and cameras. Tesla is on a different path. Believing that sensor systems based on camera-based visual image recognition technologies are the way to achieve more independent and less expensive self-driving.
But we have a problem with Tesla’s self-driving method based solely on vision very difficult because it requires the use of a continuously improved system that can quickly adapt to new and changing road conditions. So it must be able to share this information with other cars on the road. This type of learning requires more capabilities than is possible for a single vehicle, or in other words. It requires a supercomputer.
Speaking at the joint international conference on computer vision and pattern recognition in June 2021, andré Carpathia. Tesla’s chief artificial intelligence officer. Revealed that Tesla was working on a supercomputer project to do just that.
Tesla’s supercomputer, which has not yet been publicly named, is a cluster of 720 individual computers called nodes. Each with eight GPU a100 80GB GPUs that enable it to perform high-resolution. And highly complex calculations with up to 500 times the capacity of standard desktop processors.
In total, the computer cluster contains 5,760 fee processing units. Or let’s move enough hardware to reach a 1.8 xaglob (supercomputer processor speed measurement unit),
so Carpathia believes this capability makes Tesla’s theoretically supercomputer ranked fifth in the world among the most capable computer environments.
The Advanced Driver Assistance System
Tesla’s modern cars use an advanced driver assistance system called the Automated Driver System. This feature package allows the vehicle to take advantage of eight external cameras to see the vehicle’s perimeter. Then, when activated and where possible, it controls the steering side, longitudinally by acceleration or braking, under the driver’s supervision. This system differs from the self-driving system developed by Waymo. A transition phase or intermediate loop that uses partial automation to bridge the gap between manual driving and complete self-control.
Automated Driver System
We go back to the automated driving system that uses information collected from all tesla cars on the roads to improve its driving decisions. At the same time, Tesla drives through the streets; its external cameras constantly collect data on the external environment. And then the computers inside the car study this data. And make predictions about how to behave in any possible scenario without sending any actual control orders to the vehicle itself.
This information is then shared through an automated learning mechanism called the neurones and then records these predictions. And is sent back to Tesla to determine whether the supposed decision is correct or that some data has been incorrectly identified. In which case the data is constantly operated using the supercomputer. Adjusting its behaviour and decisions until it gets error-free treatment, leading to continuous training of Tesla’s automated driver model.
We can effortlessly conclude that this method not only consumes a large amount of processor capacity but requires an enormous amount of space to store 1 million sections for 10 seconds each. Which represents Tesla’s automated driver training database. These sections alone occupy 1.5 bits (1 petabyte equal to 1,000 terabytes). While the vehicle system can store approximately 10 bits of data on the ultra-fast NVMe flash memory.
Since the Carpathia group uses nevia graphics processors, it appears unrelated to the dojo project. However, it still plays an important role in achieving Tesla’s ultimate goal of being the first car manufacturer to put a fully self-driving vehicle on the roads.
Tesla Car Goal
tesla’s ambitious goal has been somewhat people question on the ground by automakers and by the rejection of computer-only self-driving cars.
Especially after Tesla refused to add high-resolution lidar technology even as part of its entire self-driving car project.
None of Tesla’s currently operating cars uses lidar technology. Elon Musk said in 2018 that the lidar was merely a significant obstacle to Tesla’s computer-based vision system only before all supplementary radars were eliminated earlier this year. Unfortunately, this decision cost Tesla additional security and safety endorsements by Washington’s national highway traffic safety administration.
On the other hand, Volvo has chosen to make lidar a vital feature of the next generation of its xc90 SUV.
The current generation of tesla supercomputers will help train its auto driver model. However, the dojo project is likely to go much. Further, only time will tell whether Tesla car computer-based technologies will outperform its competitors or vice versa, which means that the technology is an adventure that may keep Tesla a leader in the field of full self-driving automation or remove it from its throne. Who knows what hiding the days for musk and Tesla?
Review Tesla Cars Use Supercomputers to Develop Self-Driving.