Predicting Pedestrian Movement

Predicting Pedestrian Movement
Feb-19-19
A new predictive system for self-driving cars anticipates pedestrian movement for better defensive driving.

In development by a team from the University of Michigan and sponsored by a grant from Ford, the system relies on a catalog of human movements observed near a busy, urban intersection. Video was recorded from a parked self-driving car using a combination of cameras, lidar and GPS to study how people move in three dimensions and determine the motions that indicate the pedestrian’s next move (such as a distracted person playing with their phone). According to the team, the Michigan team’s system offers a median prediction translation error of just 80 cm after six seconds, compared to up to seven meters of error for other methods.



More Info about this Invention:

[DIGITALTRENDS.COM]
Next Invention »
Share on      

Add your Comment:

[LOGIN FIRST] if you're already a member.

fields are required.



Note: Your name will appear at the bottom of your comment.