Meet Maël Fabien, Co-Founder of biped

Can you introduce biped and share what was your inspiration behind creating an all-in-one mobility aid to cater to people with visual impairment, spatial neglect, hearing loss, and wheelchair users?

Biped.ai is a smart harness that can help guide blind and visually impaired people. The harness uses cameras and AI to detect incoming obstacles and provide GPS navigation to the end-user. The audio feedback helps the person navigate smoothly, bringing independence in daily mobility. Of course, biped can be adapted to pretty much any specific need, and easily adapts to people who have a spatial neglect (impact half of the field of view), for example.

This whole project started when I met a blind person struggling to cross the street in summer 2020. He was making a FaceTime call to a friend who helped him avoid obstacles such as traffic signs or electric scooters.The idea of using AI to help detect these obstacles automatically occurred to me in that instant.

Assistive technology often requires a user-centered approach. How did you involve users in the development and testing phases of biped?

In the summer of 2020, we contacted the Jules Gonin Ophthalmic Hospital in Lausanne. We knew nothing about low vision at the time, so that was quite a bit of a challenge. They provided deep expertise and guidance, and also access to testing facilities and to beta-testers This really accelerated the progress of biped and what we wanted to achieve.

Over the next two years, we developed six prototypes, from a belt in the very early days to the version we started commercializing in December 2022. For each version, we reached out to a large number of beta-testers all over Europe and in the US. That was the perfect way to get exposed to lots of feedback and a large diversity of environments. Over 250 beta-testing sessions later, our 7th prototype became our first market product.

Technology advancements have opened up new possibilities for assistive devices. Could you shed light on the AI and other advanced technologies incorporated into biped, and how they contribute to its overall functionality? E.g. what made you choose the self-driving technology from Honda Research Institute?

What biped is doing was not possible even five years back in terms of algorithms. For two reasons: 1) edge computing has gotten a lot better as phones become real AI-powered machines, and 2) deep learning algorithms have massively improved. When you’re solving something like what biped is doing, you have to capture massive amount of data over three different cameras, process it in real-time, identify all the obstacles, their types (cars, pedestrians, electric scooters, etc.), their trajectories, and predict if there will be a risk of collision, before sending the audio feedback. All of these, in just a few milliseconds. Pretty much the same software as a self-driving car. It’s a big challenge when you don’t have a large GPU sitting in a car or in the cloud. But combining all the recent evolutions on the hardware and software level in AI now make these use-cases possible. Challenging, but possible.

So when it comes to developing all these algorithms by ourselves, we sometimes encounter some hurdles. Lack of data, and algorithm still being too slow. That’s the reason why we partnered with Honda Research Institute. They basically provide a state-of-the-art trajectory prediction algorithm that worked better, and faster than what we had. It was the perfect tradeoff.

How do you ensure your mobility aid aligns with the latest best practices, do you have any partnerships with organizations or healthcare specialists?

We’ve worked since the very beginning with leading organizations in Switzerland and abroad – The Swiss Federation for the Blind, SZ Blind or the ophthalmic hospital in Lausanne. We’re now working directly with our end-users and their low vision orientation and mobility trainers to help match and improve the best practices in this field! The main idea was always to develop a tool that could become the dream tool of blind and visually impaired themselves, but also of low vision professionals who look for devices that can be customized, without replacing white canes or guide dogs.

What were some of the key hurdles you faced while developing biped, and how did you overcome them?

The technical challenge was of course immense. Dealing with hardware and software is a pain. But we found great people, aligned with our mission and vision, with deep knowledge in robotics and AI to help us solve this. Operating on a niche and highly regulated market is probably the number one challenge we faced. biped is a medical device, so that means certifications, finding key opinion leaders, aiming for the reimbursement of the solution, and then generating traction in a market where traditional marketing won’t work. We have not solved everything there of course, but it helps us gain lots of maturity on the execution process, and a great ability to operate on other markets that will feel a lot easier.

The needs of people with disabilities can vary greatly. How does biped customize its features to meet the specific requirements of each user, ensuring that it truly serves as an all-in-one mobility aid?

We designed biped to be all-in-one by nature. That meant that it should fit a large variety of end-users, and become their single mobility device on top of white canes or guide dogs.

The reason why biped used three cameras was to replicate the field of view of a sighted person. This allows us to customize what biped understands of the environment based on the specific needs of each user. If the person has residual vision on the right, we can focus on detecting elements on the left. If the person is sitting in a wheelchair and would like to discard high obstacles, we can adapt to the user’s point of view. If the person has a hearing impairment, we can connect to hearing aids or generate vibrations. All of that can be customized via a smartphone app, in a few clicks, to truly adapt to end-users needs.

Additionally, if we wanted biped to be an all-in-one device, we needed to embed two main features: 1) telling people where obstacles are and 2) telling them where to go. This second feature is currently under development, an advanced turn-by-turn GPS navigation system, that can help explore new places and reach destinations independently.

How does biped ensure compliance with relevant guidelines to reach global markets, or how does it adapt to regional differences? 

Using AI can be a challenge but can also have some advantages. Adapting to regional differences for a device like ours mainly means two things: making sure that the algorithms work properly in the surroundings of a person, and adapt all of the language feedback to make sure that warnings and navigation instructions are understood. Luckily, computer vision algorithms can “simply” be trained on new data that biped collects. That means that the more people use biped, the better it gets overall, and the more versatile the software gets. That’s the main reason why, beyond regulatory constraints, self-driving cars are opening up their services city by city. San Francisco, Austin, LA…

Looking ahead, what are the future goals and aspirations for biped? Are there any plans for expanding the range of assistive features or exploring new markets to make mobility aid more accessible worldwide?

Our ambition is quite simple: making the world collision-free. That will mean helping not only blind but also visually impaired people, elderly people, wheelchair users, and looking further, even construction workers or cyclists. We’re starting to design our v2 these days, and we’ll share some exciting news on this expansion in the next few months.

Do you have any learnings that you could share: what would you do differently if you were to start all over again?

I like to keep a list of learnings of the past few years on my side. Here are my top 5 learnings I’ve written down:

• Pick co-founders who can really stay for the long run
• Find competent people especially when it comes to hardware
• Beta-testers want a project to succeed and might provide feedback that is too gentle in front of founders
• Fundraising always take twice the time you planned
• Everything ends up being late, so plan for the worst and add 20% more.

How can the Swiss Startup Association bring value to the startup founders?

It’s hard not to appreciate how much support Swiss startups get in general. The Swiss Startup Association brings really inspiring interviews and some of the coolest events in the Swiss startup scene. Entrepreneurship can be a lonely journey, so knowing that you can meet other entrepreneurs and that the SSA advocates for Swiss startups is definitely a big added value.