00:00Welcome to the video from the Auto Magazine Tuning Blog.
00:05This time, we're talking about Autonomous Driving Systems and NVIDIA's role in developing such technologies.
00:12NVIDIA is driving the development of Artificial Intelligence for autonomous vehicles with a combination of the Omniverse platform and the capabilities of the Cosmos system.
00:20Cosmos supports AI-centric autonomous vehicle systems with prediction and inference capabilities that enable a fully trainable end system.
00:33This involves new methods such as model augmentation, closed-loop training, and the generation of synthetic data.
00:42Model augmentation begins with transferring the driving knowledge of a slower, more intelligent teacher model to a smaller, faster student model.
00:51In the car, the teacher model shows the optimal driving path, while the student model learns iteratively until it performs at almost the same level.
01:01This process provides a basis for the control model.
01:05However, fine-tuning is required for more complex scenarios.
01:09This is where closed-loop training comes into play.
01:12It uses driving data logs to reconstruct physics-based 3D models with the help of Omniverse, in which driving behavior can be trained.
01:20These scenes are specifically varied to test the model's ability to generate driving trajectories.
01:29Cosmos evaluates the AI's behavior in these scenes to measure the model's performance.
01:33The resulting new scenarios and evaluations form a comprehensive data set that improves training in closed simulation loops and makes autonomous systems more robust in the face of difficult traffic situations.
01:47In addition, the generation of synthetic 3D data strengthens the adaptability of autonomous systems to different environments.
01:58Omniverse creates detailed 4D driving environments from driving data by combining maps and images to create a digital twin of the real world.
02:06Each pixel is segmented to support Cosmos in scene recognition.
02:13On this basis, Cosmos generates realistic and diverse training scenarios, further closing the gap between simulation and the real world.
02:23Omniverse and Cosmos thus enable the autonomous vehicle system to learn more intelligently, adapt better, and drive more safely.
02:30A clear step forward towards safe and intelligent mobility.
02:37As always, you can find the link to read the article in our magazine in the video description.
02:43Did you find this information about NVIDIA's autonomous driving system interesting?
02:47Then please leave a subscription.
02:48The link to the description.
02:49The link to the description.
02:50The link to the description.
02:51The link to the description.
02:52The link to the description.
02:53The link to the description.
02:54The link to the description.
02:55The link to the description.
02:56The link to the description.
02:57The link to the description.
02:58The link to the description.
02:59The link to the description.
03:00The link to the description.
03:01The link to the description.
03:02The link to the description.
03:03The link to the description.
03:04The link to the description.
03:05The link to the description.
03:06The link to the description.
03:07The link to the description.
03:08The link to the description.
03:09The link to the description.
03:10The link to the description.
03:11The link to the description.
03:12The link to the description.
03:13The link to the description.
03:14The link to the description.
