Biped.ai Uses This Automaker’s Self-Driving Tech to Guide Pedestrians

• Designed to help visually impaired people walk safely down a busy street, the Biped would not be as capable as it is today without using hardware and software from Honda's self-driving cars.

• Software replicates the way self-driving cars detect static and moving obstacles, but on a walking level instead of at highway speeds.

• Biped.ai is releasing its first units in Europe and expects to offer the device in the US soon once it gets FDA approval.

Self-driving tech isn't just for cars. In fact, some of Honda's autonomous vehicle technology appeared at CES this year, far away from the cluster of automotive companies in the transportation area. Instead, it was found at a booth by the Swiss company Biped.ai, which returned to Las Vegas with the commercial version of its assistive technology device for pedestrian mobility.

The device, called Biped, was designed to help visually impaired people walk safely down a busy street as a supplement to the traditional white cane. Looking like an internal support piece from an unusual piece of armor, the current Biped design hangs around a person's neck and has two long arms with sensors in them that drop to the chest area. It's not the smallest device, but it's come a long way since Biped.ai brought a prototype unit to CES a year ago.

The Biped would not be as capable a device as it is today without using some hardware and software from self-driving cars. The Biped uses cameras and LiDAR sensors to scan a 170-degree field of view in front of the wearer, at ground and head level, day or night.

Software replicates the way self-driving cars detect static and moving obstacles, but on a walking level instead of at highway speeds. It turns out that sidewalks are a tougher nut to crack than regulated roadways.

"On the software level, there are lots of things that have been figured out in the last couple of years in the self-driving car industry that we're still facing," Biped.ai's Mael Fabian told me. "Pedestrian level is more challenging because you have more movements of the shoulders—you have more stochasticity. Path planning and obstacle detection are both systems that are very well developed in self-driving cars, but we still have to fine-tune them."

Fabian said the European arm of the Honda Research Institute approached Biped as part of an initiative from the automaker to spread its technology to new applications. Self-driving cars are already good at detecting static and moving obstacles, which is why Biped has now integrated the automaker's trajectory prediction system into the device.

"It works amazingly well compared with what our team was able to put up last year," he said. "It just overcame lots of the limitations we had." The bone conduction headphones now also transmit short beeps to the user, similar to the audio alerts that work with backup sensors in cars.

"Just like parking assist in a car, if there's an obstacle coming from the left, the person will hear beep-beep-beep," he said. "It's a very immersive and simple experience."

It's possible that future Biped units could communicate with autonomous vehicles. Fabian pointed to Waymo's accessibility program as a place where Biped could help a vision-impaired person safely make their way to a waiting AV.

"It's the pedestrian-level last-mile problem," he said.

Biped.ai is releasing its first units in Europe and expects to offer the device in the US soon once it gets FDA approval. Biped.ai offers the unit for free for a month-long trial and then charges $129 a month. A swappable battery provides over three hours of operation on a charge. Future models could be equipped with communication tools to create a live map of places that might be difficult for vision-impaired people to walk, but for now, each unit is self-contained.

"What we want is to have full integration with fleets of vehicles in general and also have crowd effects when it comes to the blind community using this device," Favian said. "If someone encounters challenges in the street somewhere, and another blind person comes by a couple of days later, we probably want to inform the next user that there are potential hazards on the street there. So that's definitely the vision here, to make the whole system smarter over time."

Beyond helping pedestrians, can you envision autonomous technology being useful in other ways? Please comment below.