Communications of the ACM
More Efficient LiDAR Sensing for Self-Driving Cars
A team from the Massachusetts Institute of Technology has been working on an end-to-end framework for self-driving cars that can navigate autonomously using only raw three-dimensional point cloud data and low-resolution GPS maps.
Credit: Bosch
A new machine learning system for driverless cars uses an end-to-end mapless driving framework that taps raw Light Detection and Ranging (LiDAR) data for autonomous navigation.
Researchers at the Massachusetts Institute of Technology (MIT) engineered new deep learning elements which harnessed modern global positioning system hardware more efficiently to enable real-time vehicle control.
MIT's Zhijian Liu said, "We've optimized our solution from both algorithm and system perspectives, achieving a cumulative speedup of roughly 9x compared to existing [three-dimensional] LiDAR approaches."
Tests demonstrated that the system reduced how frequently a human driver had to assume vehicle control, and was resilient against severe sensor malfunctions.
From MIT Computer Science and Artificial Intelligence Laboratory
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found