Motion Prediction Algorithms Enhance Safety Features for Automated Vehicles

Source: Communications of the ACM – Artificial Intelligence

By Southwest Research Institute
March 5, 2020
Comments

Pedestrian detection.

A motion prediction system developed at the Southwest Research Institute enhances pedestrian detection in autonomous vehicles.

Credit: blogs.nvidia.com

Researchers at the Southwest Research Institute (SwRI) have developed a motion prediction system that enhances pedestrian detection in autonomous vehicles.

The computer vision tool relies on a deep learning algorithm to predict motion by observing real-time biomechanical movements, focusing on pedestrians’ pelvic area.

The researchers compared optical flow algorithms to other deep learning methods, including temporal convolutional networks (TCNs) and long short-term memory. The team tested several configurations and optimized a novel TCN that outperformed competing algorithms, predicting sudden changes in motion within milliseconds with a high level of accuracy.

The SwRI algorithm optimizes dilation in network layers to learn and predict trends at a higher level.

The team used a markerless motion capture system, which automates biomechanical analysis in sports science. Then, the system used computer vision and perception algorithms to provide insights into kinematics and joint movement.

From Southwest Research Institute
View Full Article

 

Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA


 

No entries found