Chapter
The Power of Learning-Based Perception for Autonomous Driving
57:45 - 1:02:30 (04:44)
Turning autonomous driving into a learning problem allows for a more efficient and effective approach, rather than relying on hand-coded hacks and modules.
Clips
Hand coding the output of perception systems for self-driving cars is insufficient because it will miss important features in the real world that are not present in simulated environments, making the feature vector look fundamentally different.
57:45 - 59:26 (01:40)
Self-Driving Cars
Summary
Hand coding the output of perception systems for self-driving cars is insufficient because it will miss important features in the real world that are not present in simulated environments, making the feature vector look fundamentally different.
ChapterThe Power of Learning-Based Perception for Autonomous Driving
EpisodeGeorge Hotz: Comma.ai, OpenPilot, and Autonomous Vehicles
PodcastLex Fridman Podcast
Will Drevno explains how lane change component are trained through a learning problem and solve it with scale.
59:26 - 1:02:30 (03:04)
Self-Driving Cars
Summary
Will Drevno explains how lane change component are trained through a learning problem and solve it with scale. He believes that self-driving cars need to rely less on hand-coded hacks for navigation and integrate more adaptive cruise control in all driving situations.