Testing Biometric Sensors for Use in Micromobility Safety

Posted on April 2, 2024April 3, 2024

Biometric sensors have long been used in cognitive psychology to measure the stress-level of individuals. These sensors can measure a variety of human behaviors that translate as stress: the movement of eyes, stress-induced sweat, and heart rate variability. Recently, this research strategy has moved beyond psychology and into disciplines like transportation planning, to provide an alternative approach to researching micromobility and stress.

We spoke with Dr. Wenwen Zhang, associate professor at the Edward J. Bloustein School of Planning and Public Policy at Rutgers University, about her experience learning about and using biometrics for a micromobility study. Dr. Zhang’s research, “Rider-Centric Approach to Micromobility Safety” examines the stress levels of micromobility users as they transverse a varied path through an urban space.

Q. How is your research funded?

A. Funding comes from multiple sources. The first source is a seed grant from the Rutgers Research Council which supports an interdisciplinary pilot project. Through this grant, we purchased biometric sensors and hired students to conduct a literature review and develop a research design. We also processed the collected pilot data and paid for participation incentives under this funding. I presented preliminary findings from this study, Rider-Centric Approach to Micromobility Safety, at the 2023 NJDOT Research Showcase. At the time that I presented it, I had 24 samples. The presentation ended up inspiring several people who attended the Research Showcase to volunteer as participants—which increased the sample size to 30.

Our other source of funding came from an external grant from the C2Smart University Transportation Center (UTC) at NYU. We used this resource to support obtaining additional stress sensors, data analysis, cleaning, preprocessing, and modeling, as well as collecting more sample data for the E-scooter and bicycle experiments.

Q. How did you get interested in using biometrics sensors (e.g., eye tracking glasses, galvanic skin sensor, heart rate monitors) to study micromobility safety? How does this research differ from your past work?

A. Before I used biometric sensors, most of my work used passive travel behavior data. For example, to determine the revealed preferences of mode and route choices and risk factors, we used travel trajectory or existing crash big data to develop statistical models. I have found that the entire process is very passive, especially since we only explore risk factors after traffic accidents. It’s surprising that in the research field today we know so little about how human beings actually navigate urban environments while using different travel modes and how it relates to perceived safety. I wanted to explore questions like what is their gaze behavior? How do they feel while they travel using different modes? How do they feel traveling on roads with different design features and how is that going to influence their travel satisfaction or experience overall?

Dr. Robert Noland, Distinguished Professor at the Rutgers Bloustein School, suggested I investigate the use of biometrics in planning studies. As I dug more into the literature, I realized that biometrics in transportation is a very fascinating topic that I wanted to get into. Once I did experiments in the field, I realized that I really enjoyed talking with different people about how they perceive the built environment while they travel. Biometrics provide richer data compared with revealed preference data that I used to work with.

Q. In your research, you noticed that some corridors were more stress-inducing (according to biometric sensors) than expected, despite properly designed safety infrastructure. How do you think this discovery may affect how planners and engineers look at urban road design and micromobility safety?

A. This study collected one-time cross-sectional data. We asked people to walk around an area and tell us whether they feel stressed or not. If they are feeling stress, even in the presence of a safety improvement, it does not necessarily mean that the implemented safety design is not working. For example, in New Brunswick, we observed that a lot of people found it stress-inducing to cross Livingston Avenue, although it has been the subject of a road diet and has several pedestrian safety features incorporated into the new design. While outside our scope of research, one way to understand the impact of the safety infrastructure would be to conduct a “before” and “after” study. This leaves an opportunity for more research, to see how effective the pedestrian-only infrastructure is in reducing stress level. Potentially, it can provide evidence to support pedestrian-only design. Biometric sensors used in a “before and after” study can help us to answer which infrastructure is more preferred.

Q. You are in the process of collecting data for cyclists and e-scooters using the same method, what are your principal objectives in addressing this segment? Do you expect the results to be different?

Dr. Zhang conducted one pilot e-scooter experiment at Asbury Park, NJ in 2022 to test out the devices and examine how to set up research experiments. She equipped the e-scooter rider, Dr. Hannah Younes, post-doc researcher at the Rutgers Bloustein School, with an eye tracking glass, a GSR sensor on the hand, and a 360-degree camera on top of the helmet.

A. Yes, absolutely, different travel modes will likely alter a person’s expectation for a safe travel environment. For example, we noticed a big difference in the enjoyment of pedestrians and e-scooters on the same path through a park. We had thought that the e-scooter users would enjoy the ride as the pedestrians had, however, the pavement was too rough for the small wheels of the e-scooters. Although the park was walking-friendly, it was not friendly for e-scooters. This shows that each of these micromobility modes needs different kinds of support to feel safe and comfortable.

Q. What are the limitations to this study? Do you have plans for future research to address this? How would you like to expand your research in this topic?

A. Each of the biometric sensors has limitations. For example, eye trackers face some difficulty when identifying the pupils of a participant in direct sunlight. As a result, the eye tracker renders a low eye tracking rate. Eye trackers also work better with darker eyes as the eye movements are more readily recognized. The eye trackers, kept on glasses, also restrict individuals who wear glasses from participating. The unfortunate result of this is that it often excludes a lot of senior people from the experiment. This issue may be alleviated as we are obtaining additional funding to obtain prescription lenses for eye trackers.

GSR sensors use low voltage on skin to measure skin conductivity, which may interfere with electric health devices. This limits individuals from participating if they have an electric health device like a pacemaker on or in their body. We purposefully excluded this population from participating to align with IRB (Institutional Review Board) protocol and to mitigate any risks.

Another limitation of the study is that we must collect sample data one by one, which is a time-consuming process. We can only collect a very small sample compared to a traditional statistical model kind of study, which may have access to thousands of records in the sample. From our literature review, biometrics sensor studies typically involve 20 to 30 participants, but for each participant we have a very rich dataset. For each participating volunteer, we end up with over one gigabyte of data. The limited number of participants may make it harder to generalize results to the entire population, and people may question the results applicability. In some ways this data is similar to the results of qualitative studies, where we have richer information but small sample size, rendering some generalizability issues.

*Feelings of safety were measured using the traditional self-report survey* as well as *biometric trackers like Heart Rate Trackers, Eye trackers and GSR (pictured above)*.

Q. What challenges have you found in working with biometrics sensors, or in the interpretation of output measures?

A. The eye tracker and heart rate measures are reliable, but some biometrics have posed challenges. The GSR (galvanic skin response sensor), which tests your sweat level, is very sensitive to humidity and time of the day. The sensor also picks up on sweat resulting from physical exertion, making it difficult to distinguish between stress-induced sweat and physical sweat.

Interpretation of output measures for this metric requires data cleaning and processing to eliminate the effect of sweating from physical exertion. We try to decompose the data to separate the emotional peak from the sweating caused by physical activity using various algorithms. We are still underway testing out different algorithms to clean up the data. So far, we have found that GSR data are very real-time in nature and a good indicator for stress level but are very noisy data and requires some manual processing. This means we spend a lot of time preprocessing the collected data before conducting data analysis.

Q. How do you expect this research to inform transportation agencies in New Jersey and elsewhere?

A. This type of research captures such rich data on travel behavior itself. Most of the literature using biometrics has been focused on driving, so this research expands the perspective. Here we’re focusing on slow mobility, like active travel and micromobility. Individuals who participate in slow mobility are more vulnerable road users, and we want to see how they behave in different travel environments. This can help agencies gain more insights into how to design safety infrastructure. Beyond that I can also envision the technology being used to evaluate whether certain improvements or infrastructure designs help to improve travel satisfaction or improve people’s experience at the same location by doing “before and after” studies. This type of study also allows you to measure and quantify the effect of the improvement.

The use of biometric sensors in the field can also be used to foster meaningful public engagement processes to show the lived experience of different people in a neighborhood or traveling through a different corridor, which can be very powerful.

Q. Do you feel the research methods are at a stage where they are “ripe” for use on other demonstration projects, planning or project development studies?

A. After one year of experimentation, our project team can readily work with biometrics. We have a good understanding of sensor limitations and how to set up the sensors to correctly reduce noise as much as possible. Our experience has also helped determine what kind of metrics can be extracted successfully and reliably through the sensors.

The most useful case for those sensors is to evaluate before and after, so that we can quantify how much people appreciate those implementations in a more accurate way. Beyond that, the sensors can also be effective infrastructure assessment tools. For example, imagine that you ask people to wear biometric sensors and do a bicycle infrastructure evaluation; the agencies can get more realistic and rich data compared with a more traditional survey approach. This rich data can help determine the most effective improvement. It ends up being more inclusive that way.

The tools can be very useful for fostering community engagement with vulnerable populations. For example, if agencies want to improve the accessibility for wheelchair users, they can ask individuals in wheelchairs to wear the sensors and move about an area. Recording and reviewing how they experience a journey is more powerful compared with just asking individuals with needs about their travel patterns. It’s going to be a more straightforward way to show the world how we can make the streets more inclusive for those vulnerable populations.

Q. Do you think local governments and non-governmental organizations could make use of biometrics sensors as a strategy to promote community engagement and outreach to local communities, or to address specific community safety or livability issues? Would it be cost-prohibitive to employ such tools for such community-based planning issues at this time?

A. From my point of view, the most effective way would be for the agencies to identify where there are needs and promising projects and then work with skilled researchers or practitioners who have these sensors already and have begun to climb the learning curve in the use of sensors and interpretation — for example, they could work with us. They would need to pay for the researchers’ time and participation incentives, or if they were to collaborate with a UTC (University Transportation Center) to conduct such research collaboratively.

The sensors are not the most expensive part of the study. The most expensive item is the researcher’s time to collect and analyze the data. The data are very complicated to analyze in the first place because it’s a large amount of data with noises. The researchers need to put in a lot of time to get it to the state where you can extract the relevant variables out and start to interpret them.

Q. How would you characterize the “state-of-training” in using biometrics for students or early career or mid-career professionals in transportation?

A. The biometric sensor itself is not very new, but new to the transportation field, especially for slow modes. It has been widely used in cognitive psychology, where there are classes to interpret those as well. Generally, I don’t think the current transportation and urban planning curriculum for students includes enough classes to cover those sensors. We probably need to teach not only biometric sensors, but urban sensing in general.

In an ideal course, students could get their hands dirty by putting those sensors in the field and then once the data are collected, they can learn how to preprocess and analyze the data. It would have to be a one-year kind of curriculum design to get people involved and ready for it. Of course, instruction on the use of sensors will differ by topic. For example, if you are working in the air quality field, then there are many different air quality sensors and each of them come with different data formats and require different experiment design and analytic skills.

Regarding the mid-career transportation professional, at this moment I believe the research is more in the academic field and focusing on testing and evaluation. I wouldn’t suggest that the research is so ripe that a mid-career transportation or urban planner professional should need to invest their time in learning how to use biosensors unless they have a research project that may benefit substantially from using the sensors.

Resources

To learn more about the use of biometrics in the field of active transportation, see:

Ryerson, M., Long, C., Fichman, M., Davidson, J.H., Scudder, K.N., Kim, M., Katti, R., Poon, G. & Harris, M., (2021). Evaluating Cyclist Biometrics to Develop Urban Transportation Safety Metrics. Accident Analysis & Prevention, Volume 159, 2021. Retrieved from https://www.sciencedirect.com/science/article/pii/S0001457521003183?via%3Dihub

Fitch, D.T., Sharpnack, J. & Handy, S. (2020). Psychological Stress of Bicycling with Traffic: Examining Heart Rate Variability of Bicyclists in Natural Urban Environments. Transportation Research Part F: Traffic Psychology and Behavior, Volume 70, 2020, Pages 81-97. Retrieved from https://www.sciencedirect.com/science/article/pii/S1369847819304073?via%3Dihub.

To read more on Dr. Zhang’s work, see:

Zhang, W. (2023). Rider-centric Approach to Micromobility Safety. 25th Annual NJDOT Research Showcase. Presentation. Retrieved from https://www.njdottechtransfer.net/wp-content/uploads/2023/11/Zhang-Safety-2nd-Presentation.pdf.

Zhang. W. 25th Annual NJDOT Research Showcase. Recording starts at: 59:00. Retrieved from https://youtu.be/D_rQP-Dv8gU

Zhang, W., Buehler, R., Broaddus, A. & Sweeney, T. (2021). What Type of Infrastructures do E-scooter Riders Prefer? A Route Choice Model. Transportation Research Part D: Transport and Environment, Volume 94, 2021. Retrieved from https://www.sciencedirect.com/science/article/pii/S1361920921000651.

For more information about the use of biometrics in the broader transportation field, see NYU’s C2SMART’s research project on Work Zone Safety:

Safety Behavior and Gender Split Differences in Micromobility: A Q&A Interview with Researcher

Posted on February 9, 2024February 12, 2024

The State of New Jersey has advanced policies, programs, and projects such as complete streets that encourage greater transportation modal choice and less dependency on the single-occupancy vehicle to address climate change and promote mobility and access for all users. The NJ Statewide Bicycle and Pedestrian Master Plan and Complete Streets policy seek to increase connectivity and micromobility safety through road diets and other infrastructure changes. New Jersey Department of Transportation (NJDOT) is also sponsoring safety programs that aim to reduce traffic related fatalities. In-depth research that looks at differences across gender, race, socioeconomic status and ability may assist in expanding policies and statewide initiatives that prioritize safety and accessibility for all road users.   

This Q&A article has been prepared following an interview with Dr. Hannah Younes, a post-doctoral researcher at the Alan M. Voorhees Transportation Center at Rutgers University, who focuses on cyclists and e-scooter behavioral patterns. Her most recent study conducted in Asbury Park, NJ observed gender differences of chosen micromobility mode, helmet usage, and location. We interviewed her to better understand the data sources and methods used in the research, its limitations and explore the applicability and implications of her research findings to the field of transportation and future research. The Q&A interview has been edited for clarity.

Q. How was your research funded?   

This work was supported by the National Science Foundation under a grant called “Making Micromobility Smarter and Safer”. The lead on this is Dr. Clint Andrews at Rutgers University and there are several other principal investigators. My study acts as a part of this multi-year research. 

Q.  Can you share a brief overview of your findings? Are the results surprising or unique compared to past research?   

We are one of the only studies comparing the safety behavior of cyclists and e-scooter users across genders. Without considering gender, we found that one-third of cyclists wore a helmet. We also found in our observations that e-scooter users did not wear a helmet. It speaks to how important it is to have safe micromobility infrastructure, especially knowing that people are unlikely to wear a helmet. In the U.S., even if you give everyone a helmet, they’re probably not going to wear it. That’s just how it is. Keeping people safe in other ways is paramount. 

We also found that a greater proportion of women were using e-scooters than bicycles. This is important because cycling has long been a male-dominated mode of transportation, for a variety of reasons. That is true across the world. There are studies that suggest women are less likely to cycle to work because of clothing like wearing a skirt or dress or heels, or fears of sweating. E-scooters remove that hurdle since they are not as prohibitive in terms of clothing and require less physical exertion. So, the vehicle type itself may make a difference. Moreover, women place more importance on bike lane infrastructure than men.  If we are seeing that e-scooters are the preferred mode for females, perhaps e-scooters can help narrow the gender gap in micromobility.

Q.  Can you talk a little bit about the methods used for this study? How are these methods different from past research? Why did you choose to use traffic cameras for your observations?

This work was done using manual observations, a common method in micromobility studies. Previous research had used observations collected in the field. Instead of having observers in the field, we observed traffic camera footage at one intersection. Because we were observing gender and race as well as group behavior, the footage was useful as it allowed us to pause when needed. It was also less resource intensive than having a person stand in the field since no travel expenses were associated with the analysis. 

Q.  What challenges have you found in working with and interpreting traffic camera footage? With the improvement of AI technologies, do you think there will be an opportunity to automate this process in the future?  Are there any limitations you expect from this type of innovation? 

It is very time consuming and tedious to analyze this much camera footage. We analyzed 35 hours of footage. I would love to have analyzed more, but you have to draw the line somewhere depending on the resources available for the research or project study. Most of the time, we fast forwarded until a micromobility user was detected, but it still requires undivided attention. There is a possibility with current technology to incorporate AI technologies: to use computer vision to detect humans, which then can be manually viewed by a human to assess micromobility mode, gender, and helmet use. This would likely reduce the manual labor… It would be interesting to compare the computer vision model to the work I have done… Nonetheless, computer vision does not differentiate properly between pedestrians and e-scooter users, so it is prone to misidentification, which would lengthen the time taken to observe manually.

At this point, computer vision cannot detect gender, helmet use, and group riding properly from traffic camera footage. More high-resolution images would be needed to differentiate gender and helmet use (like unobstructed face images) and group riding requires context clues like making eye contact, waiting for one another, etc. AI has the potential, but it is not there yet.  As time consuming as it is, I am confident that we detected every person, which is why we chose to observe the footage ourselves. 

Q.  What are the limitations of this study? Do you have plans for future research to address these?  How would you like to expand your research on this topic?  

The main limitation is the geographical scope of this research; it’s a lot of work for one city. We only analyzed the behavior of micromobility in one location, Asbury Park. It isn’t clear how much the results will translate from one location to another. Mode of transportation and behavioral use depends on many different factors that vary from location to location. There is evidence that the gender gap is smaller for e-scooter users in Brisbane, Australia, but not to the extent observed in Asbury Park. Same goes with helmet use. A larger scale study would be useful. Other limitations include the types of micromobility modes: we only observed shared e-scooters and privately owned bicycles in Asbury Park. So, we’re comparing two different vehicles and two different share types to one another. When analyzing the data, we must consider both of these factors. For example, are behaviors attributed solely to the vehicle or to the share type? Probably both. When you’re looking at the gender gap, is it because it’s an e-scooter or is it because it’s shared that there is a narrower gender gap? 

 An analysis comparing shared and privately owned e-scooters with shared and privately owned bicycles would be great. Differentiating between e-bikes and bicycles would be great too, although the resolution of traffic camera footage makes it very hard to differentiate between the two. Even with an observer onsite, it would be hard to detect, so you would need a survey, but this could alter behavior. In Asbury Park, a lot of people have privately owned e-scooters now, so we could do another study in 1.5-2 years and get additional insights in the same location.

E-bikes are a growing mode of transportation, but even with traffic camera footage, it is very hard to tell an e-bike apart from a bicycle, so maybe in that case you would need somebody on site actually observing. You’re losing the ability to pause footage, but it might be more useful if you’re looking at e-bikes. Race and age were also very difficult to observe from the footage. It could be easier if someone was in person to observe in addition to the traffic camera footage. Even then, without asking directly the age and race/ethnicity of the user, there will be bias. There are a lot of different things to consider; it really depends on what the question is. 

Q.  How would you like this research to inform transportation agencies and practitioners in New Jersey and elsewhere?   

There are several key points. Users of shared e-scooters and privately owned bicycles are different and behave differently. E-scooter users are more likely to take risks like not wearing a helmet or riding on the road. Planners must ensure that the infrastructure keeps them safe. That is, implementing dedicated protected bike lanes that are connected to a greater network and adding traffic calming measures to slow the speeds of motor-vehicles like raised crosswalks or narrower traffic lanes.

Understanding the reasons behind lane use is important as well, as there are concerns for pedestrian safety. Our research observed that lane use was different; for example, 7 percent of male cyclists rode on the sidewalk, compared to 28 percent of female e-scooter users.

Additionally, having a shared e-scooter system in a city can increase female participation in micromobility use. It is a more gender equitable mode than bicycles. Other agencies might want to implement an e-scooter share program in their town. 

Q.  Your research shows that women were more likely than men to ride on the sidewalk while using an e-scooter or bike. Given that this strategy is illegal in most parts of the country, how can planners, engineers and policymakers use this information to increase feelings of safety for female micromobility users?    

This is really interesting. From my research, there is not a lot that I could say. Implicitly, one of the reasons for someone to ride on the sidewalk instead of the road is that they feel safer on the sidewalk. There is a need to ensure that micromobility users feel just as safe on the road–that is, implement a dedicated and protected bike lane, and provide a clear separation from motor-vehicles.

From our work, we know that there are other more complex factors at play: our research had clear results for road lane use with the implementation of the bike lane, but less clear ones for sidewalk use: sidewalk use was not significantly reduced by the presence of a pop-up bike lane. To encourage safe road use, ensuring a complete network would be a start. The pop-up bike lane was not connected to another bike lane going downtown, for instance. If you’re already coming downtown on the sidewalk, you might be more likely to stay there given the existing curb that would need to be crossed to go from the sidewalk to the pop-up bike lane. 

Q.  NJDOT is sponsoring a program to ensure the implementation of the Statewide Bicycle and Pedestrian Master Plan. In what ways could this master plan or a future one align with the findings in your study? 

The results of this study reinforce that implementing a bike lane provides a layer of safety for micromobility users. Nearly all the increase in bike lane usage came from a reduction in traffic lane usage, not in sidewalk usage. There is so much research out there that shows that bike lanes save lives; in the case of a crash, someone in a bike lane is less likely to be injured. Ensuring that plans accommodate both bicycles and e-vehicles–like e-bikes and e-scooters–is also paramount. 

Q.  The Biden Administration has set a goal to achieve a net zero emissions economy by 2050. How might a shift toward micromobility help the nation reach its climate and carbon emission goals?   

Bicycles are zero emission vehicles. E-bikes and e-scooters produce few emissions, especially privately owned ones since they don’t require rebalancing. Rebalancing shared vehicles requires a car or van and those gasoline emissions are absorbed by those shared e-scooters. Having an e-vehicle do that for rebalancing helps to reduce those emissions. Bicycle-friendly infrastructure, which reduces motor-vehicle infrastructure such as the number of traffic lanes, or parking, can also reduce motor-vehicle use and induce more environmentally friendly travel.  

Q.  How could a focus on reaching these climate goals impact the way that planners and engineers design streets?   

Motor-vehicle travel is the most polluting transportation mode. A shift towards micromobility Infrastructure, particularly in cities, can help induce travel from eco-friendly modes and reduce motor-vehicle use. This should be in conjunction with other measures, such as increasing gasoline prices, increasing parking costs, and providing subsidies for eco-friendly modes.

Resources

Blickstein, S.G., Brown, C.T., & Yang, S. (2019). “E-Scooter Programs Current State of Practice in US Cities.” Retrieved from https://njbikeped.org/e-scooter-programs-current-state-of-practice-in-us-cities-2019/

Marshall, H. (2023). “How do Female Cyclists Perceive Different Cycling Environments? – A Photo-elicitation study in Stockholm, Sweden.” Retrieved from https://gupea.ub.gu.se/handle/2077/78209

NJDOT Technology Transfer. (2020). “Tech Talk! Launching Micromobility in NJ and Beyond.” Retrieved from https://www.njdottechtransfer.net/2020/02/25/launching-micromobility-in-nj-and-beyond/

NJDOT Technology Transfer. (2021). “How Automated Video Analytics Can Make NJ’s Transportation Network Safer and More Efficient.” Retrieved from https://www.njdottechtransfer.net/2021/11/08/automated-video-analytics/

NJDOT Technology Transfer.(2022). “Research Spotlight: Exploring the Use of Artificial Intelligence to Improve Railroad Safety”. Retrieved from https://www.njdottechtransfer.net/2022/08/19/researchspotlightairailroadsafety

Rupi, F., Freo, M., Poliziani, C., & Schweizer, J. (2023). “Analysis of Gender-Specific Bicycle Route Choices Using Revealed Preference Surveys Based on GPS Traces.” Retrieved from https://www.sciencedirect.com/science/article/pii/S0967070X2300001X

Salazar-Miranda, A., Zhang, F., Maoran, S., & Ratti, C. (2023). “Smart Curbs: Measuring Street Activities in Real-Time Using Computer Vision,” Retrieved from https://www.sciencedirect.com/science/article/pii/S0169204623000348?casa_token=XPecGlOM6UQAAAAA:vnISsmV2aoJ3iVJefEeqjM24R5izcs66bvukCQObjuSWGTNokotT4CG_1h8UfLih16wn3FMg_Jo [DA1] [KR2]

Von Hagen, L.A., Meehan, S., Younes, H., et. al. (2022), “Asbury Park Bike Lane Demonstration,” Retrieved from https://storymaps.arcgis.com/stories/c014811ac0c14735bc9c9adc2639e88f.

Younes, H., Noland, R., & Andrews, C. (2023). “Gender Split and Safety Behavior of Cyclists and E-Scooter Users in Asbury Park, NJ,” Retrieved from https://www.sciencedirect.com/science/article/abs/pii/S2213624X2300127X#b0055.

Younes, H., Noland, R., & and Von Hagen, L.A. (2023). “Are E-Scooter Users More Seriously Injured than E-Bike Users and Bicyclists?” Retrieved from https://policylab.rutgers.edu/are-e-scooter-users-more-seriously-injured-than-e-bike-users-and-bicyclists/.