Driverless Cars' Pedestrian Defense
Teaching Self-Driving Cars to Predict Pedestrians' Next Moves
Source: University of Michigan |
Ford-Funded Predictive System Developed at University of Michigan
Researchers at the University of Michigan are developing a new predictive system that enables self-driving cars to anticipate pedestrian movements. The purpose is better defensive driving by driverless cars. Ford is underwriting the research.
Catalogs of Video
The system starts with videos from a combination of cameras, lidar and GPS shot from a self-driving car recording how people move in a busy city intersection. It records in 3D and determines motions that indicate the pedestrian's next move, such as a person focused on texting not crossing a street.
High Predictive Success Rate
According to the Michigan team, the system's results are spectacular. It offers a median prediction translate error of only 80cm after just 6 seconds, as opposed to 7 meters of error with other methods.
Comments
Post a Comment