Self-driving cars have been a hot topic among technology experts and consumers alike in recent months. While many have been skeptical about the safety and reliability of autonomous vehicles, a new study shows how researchers are going the extra mile to ensure that pedestrians stay safe.
Researchers from the University of Michigan are working on advancing the technology for self-driving cars so that they can better predict and recognize pedestrian movements.
“Prior work in this area has typically only looked at still images,” said researcher Ram Vasudevan. “It wasn’t really concerned with how people move in three dimensions. But if these vehicles are going to operate and interact in the real world, we need to make sure our predictions of where a pedestrian is going doesn’t coincide with where the vehicle is going next.”
Power of video clips
The researchers started by switching from still images to short video clips of humans so the system could learn how people walk and how arms and legs are symmetrical on both sides.
Rather than just processing a still picture, replaying the video over and over several times allowed the system to first predict where and how pedestrians will move and then assess the prediction for accuracy. The system can also see a three-dimensional view of humans and watch them move in real time.
According to researcher Matthew Johnson-Roberson, the system goes beyond just “making predictions of...one single thing,” and will instead be capable of determining “where that pedestrian’s body will be at the next stop and the next and the next.”
The researchers parked a car at various intersections in Michigan and had it record several days worth of data to better train the system to understand pedestrians’ habits and tendencies in real time.
Additionally, the researchers created the system to be in line with actual human capabilities, including things like realistic walking speed, so as to allow for the most accurate predictions. Based on early tests, the researchers are excited about the potential for the system, as it showed a very high accuracy for predicting what pedestrians would do next.
“The median translation error of our prediction was approximately 10 cm after one second and less than 80 cm after six seconds,” said Johnson-Roberson. “All other comparison methods were up to seven meters off. We’re better at figuring out where a person is going to be.”
The researchers are hopeful that this technology will “contribute to a safer, healthier, and more efficient living environment,” said researcher Xiaoxiao Du.
Autonomous vehicle controversy
The auto industry has seen mixed results when it comes to self-driving cars. While Volvo, Ford, and Honda seem to be all in on putting their models on the road in the next few years, a Toyota executive thinks the technology has received undue hype.
The technology has certainly raised eyebrows, if nothing else. Uber has had its fair share of issues with self-driving vehicles over the last several months.
Earlier last year, a self-driving Uber vehicle killed a woman in Phoenix, and the rideshare giant shut down its self-driving operation in the city at the beginning of the summer. By mid-July, the Pittsburgh branch also closed its doors.
However, by year’s end, Uber reported that its self-driving cars were once again ready for public roads, and the company began the process of reapplying for testing privileges in Pittsburgh.
“Our goal is to really work to regain that trust and to work to help move the entire industry forward,” Uber executive Noah Zych told The Washington Post. “We think the right thing to do is to be open and transparent about the things that we are doing.”