Helping automated vehicles communicate with vulnerable road users

By Heather Montague

As automated vehicles become more prevalent, there are concerns about the lack of communication between the vehicles and other vulnerable road users (VRUs) like pedestrians and cyclists. Dr. Pavlo Bazilinskyy, Assistant Professor at TU Eindhoven, researches how automated vehicles interact with these vulnerable road users in order to improve safety and potentially save lives.

An advocate of open science, Bazilinskyy used a grant from the 4TU.ResearchData FAIR Data Fund (Spring 2021) to refine his dataset from a coupled simulator capable of running immersive simulations with dozens of people, and recording of a real traffic situation from a portable sensor.

Human factors

With a human behind the wheel, it is common for drivers to communicate with pedestrians and other road users via eye contact or hand gestures. “In the future there will be no driver in the car and as a result the paradigm will shift,” said Bazilinskyy. “If you’re in a self-driving vehicle, maybe you’re reading a book, working, looking down or even sleeping. A pedestrian might not know if they should cross the street.”

His research focusses on how to communicate with people outside of the car by means of artificial things like displays or sounds. One method of doing so is external human-machine interfaces (eHMIs) which are currently being developed to communicate intention or provide advice to pedestrians on when it is safe to cross. For example, if an autonomous car approaches an intersection it could show by means of a light what it intends to do, explained Bazilinskyy.

Virtual simulation

To test design concepts like this, some of Bazilinkskyy’s research involves the use of a coupled virtual-reality simulator. Although it’s still a work in progress he said they are developing a tool that can be used for research with multiple humans in the same environment. “In my domain, experiments are mostly done with one person or one car and the rest is simulated,” he said. “But in real life if you imagine yourself at an intersection in Manhattan there could be 200 people around you, maybe even 2000.”

“In my domain, experiments are mostly done with one person or one car and the rest is simulated, but in real life if you imagine yourself at an intersection in Manhattan there could be 200 people around you, maybe even 2000.”

-Pavlo Bazilinskyy, Assistant Professor, TU Eindhoven

For Bazilinskyy, it’s important to use a large and representative sample population in order to produce robust, reliable and reproducible research. To achieve this, he uses crowdsourcing which enables respondents from all over the world to participate in experiments.

Clean and shiny data

A long-time advocate of open science, Bazilinskyy says that making data FAIR is intuitive for him. In fact, he estimates that the data from his project was already approximately 90% FAIR. But with a lot of code and data already in an open domain, he applied for the grant to try and make it a bit more organised. “Through this funding, which I’m really grateful to have received, I’m more focussed on presentability of my data,” he said. “Now I have made it a bit nicer, cleaner and shinier.”

Part of Bazilinskyy’s work also involved making a game tailored for academic research, which is configured to save data in an efficient, robust and synchronised way. So, he used his funding to hire a freelance programmer to help develop the data a bit further and to structure his GitHub repository. “Before FAIR data I focussed more on the numbers inside of my repositories and the actual data, I made sure that the data is easy to access and open,” he said. “But now I’m trying to step it up a bit and make it more presentable.”

Keep it simple

Using the FAIR data fund, Bazilinskyy also worked on a host panel for his simulator. That basically means that if someone downloads data through the GitHub link then it will be much easier to configure and will also make it a bit more open. “I’m trying to make the simulator in such a way that you don’t need to be a scientist or a programmer to use it,” he said. “The public can play around with it, even a kid could run the simulator. In fact, sometimes I try to tailor my research towards education in schools.”

FAIR from the start

For Bazilinskyy, making his research open and FAIR has always been a priority. “I never even thought about any other way to approach things,” he said. “We live in the 21^st century and it’s quite easy to make your things open if you spend a bit of time. And if you think about it from the first day of a project, it can be done without much work at all.”

“I think science needs to be open by default because we do science for the benefit of the whole society, not just for our own academic career.”

– Pavolo Bazilinskyy, Assistant Professor, TU Eindhoven

There’s also the matter of credibility for Bazilinskyy. “95% of the time when I look at work by someone else, if I don’t see data and code my level of trust in that work decreases. I think science needs to be open by default because we do science for the benefit of the whole society, not just for our own academic career.” He also noted that having done a PhD in human factors, he knows that humans are not perfect. “If I see data, I trust that way more than humans these days.”

Future goals

So far, Bazilinskyy says they have published one paper that resulted from the couple simulator. “It was a massive project I did with my students, but it shows the power of this tool,” he said. “My goal is by the end of this year to make a major new release of the tool and that will be the main deliverable.” During his PhD and postdoc at TU Delft, Bazilinskyy also volunteered as a data champion, a group of researchers at the forefront of innovation in topics such as research data, code management and open science. And he hopes to continue this work by helping start up a data champions programme at TU Eindhoven.