Improving auto safety with Humanising Autonomy
Humans are far from ideal drivers, as anyone who’s gotten behind the wheel can tell you. But it’s pedestrians who often — and increasingly — pay the toll. Nevertheless, the future of transportation will be driver-less. As autonomous vehicles gain ascendance, the question quickly arises: “Can self-driving cars do better than their human counterparts?”
For Maya Pindeus, founder and CEO of London-based Humanising Autonomy, the answer is yes. Her company is developing AI-powered computer vision that can help autonomous vehicles identify pedestrians and predict their behavior — which can help to minimize the dangers of self-driving cars. Maya is among a burgeoning group of entrepreneurs and transportation companies helping to usher in the future of autonomous transportation — one that promises to improve our lives while keeping both passengers and drivers safe.
On the latest episode of the What’s NEXT podcast, I spoke with Maya about Humanising Autonomy, the future of autonomous transportation, and how her company is taking a human-centered approach to driverless vehicles.
Putting humans in the driver-less seat
Autonomous transportation — self-driving cars, buses, and trucks — will be the future. It’s an exciting prospect for pedestrian safety. This is particularly important in light of an increasing number of pedestrian and cyclist fatalities on U.S. roads, which is at a nearly 30-year high.
Driver-less cars, equipped with sophisticated AI and machine vision systems, hold the potential to remove human error from the equation and to dramatically reduce the danger of getting on the road.
But we’re not quite there yet. As with many technologies, the initial hype related to self-driving cars has outpaced the technical capability. While autonomous cars are extremely good at navigating flat, orderly, and sunny streets, they can’t always handle the traffic, people, and weather-filled roads typical of urban settings.
In particular, autonomous cars still have difficulty with unpredictable obstacles, and nothing is more unpredictable than people. That challenge — teaching cars to safely interact with highly distracted and ever-present pedestrians — is precisely why Maya founded Humanising Autonomy three-years-ago.
“In 2016 and 2017, during peak hype, everyone thought autonomous vehicles would be everywhere today,” Maya says. “But no one was thinking of pedestrians and cyclists, who are key elements of any urban environment, or any environment, really.”
Now Maya believes the first step toward a driverless future is for self-driving vehicles to be able to understand and predict human behaviors.
To reach that end the company has developed a series of AI-based sensors and computer vision technologies that allow autonomous vehicles to identify pedestrians on the road. Not only that, but the technology assesses if those pedestrians are distracted — eating a sandwich, talking to friends, staring at their phone, among many diversions — and tells the vehicle in seconds whether it needs to take action to avoid them.
“It predicts all the things that might happen, looking a couple of seconds in the future, that really matter for an autonomous car,” Maya says.
Unlike many traditional autonomous sensors, Humanising Autonomy technology can work with just about any kind of input data — including any type of video feed.
“It can be from a dashcam, from a CCTV on a bus, from the sensors of an autonomous car,” says Maya. “Anything that allows us to collect data from lot of different environments and instances that might not be possible if you work purely with autonomous sensors. You get the diversity of behavior right off the street.”
But humans are complicated. To capture the full range of unpredictable human behaviors, Humanising Autonomy is working with a behavioral data science team — whose members have heavy psychology backgrounds. They are busy creating human intent-based machine learning models and AI algorithms, which together can analyze massive amounts of sensor and video data.
The company is focusing, in particular, on the most vulnerable pedestrians on city streets: people with disabilities, families, the elderly, people with walking-frames and wheelchairs, and the like.
Humanising Autonomy also wants the technology to be adaptable to any autonomous car, in any environment — whether it’s driving in snow or rain, in New York or New Delhi. In fact, it’s partnering directly with cities — from London to Detroit, to cities in Germany and Japan — to capture the myriad varieties and settings of vehicle-human interaction around the globe.
Making autonomous transportation a reality
Until autonomous vehicles are fully in control, Maya believes the self-driving evolution will happen in five waves, or levels, of autonomy. Level one is already here: basic AI sensors that provide today’s drivers with lane assistance, limited self-driving, and automatic braking features. Level five means no driver and no steering wheel — in other words, full AI control.
We’re still a few years away from that kind of technology. But level four vehicles — fully autonomous, with a driver backup — are already being tested today. Humanising Autonomy is currently working with transportation partners, such as Daimler and Airbus, to help make such mostly autonomous vehicles both safe and road-worthy.
In the meantime, Maya sees the opportunity to retrofit traditional vehicles, like buses and taxis, with the company’s cutting-edge technology. For instance, the company is currently working with the London transport authority, Transport for London, to help it achieve its Vision Zero goals: the elimination all deaths and serious injuries from London’s transport network by 2041.
The eventual rollout of entirely autonomous transportation, Maya believes, will also require companies and countries to come together on ethical standards that put people first — and not just around safety.
For instance, Humanising Autonomy has insisted, even prior to the EU’s GDPR privacy regulations, that any data it receives is completely anonymized and pseudonymized so that there’s no risk to personal data. The company urges other industry actors to adopt a similar stance.
More than anything, Maya believes that humans will need to have full confidence in autonomous transportation for them to relinquish the steering wheel and control of their vehicles. “The most important thing, of course, is to have no further accidents while testing and deploying autonomous vehicles,” Maya says. “It’s about building a successful service that people actually trust, and that actually caters to people.”
To learn more about the future of autonomous transportation and how Humanising Autonomy is working to minimize the dangers of self-driving cars, you can listen to the full podcast episode in the embedded player above or subscribe to What’s NEXT on Apple Podcasts, Google Play, Spotify, or via RSS on your podcast app of choice.