This paper describes the functioning and development of HeartPy: a heart rate analysis toolkit designed for photoplethysmogram (PPG) data. Most openly available algorithms focus on electrocardiogram (ECG) data, which has very different signal properties and morphology, creating a problem with analysis. ECG-based algorithms generally don’t function well on PPG data, especially noisy PPG data collected in experimental studies. To counter this, we developed HeartPy to be a noise-resistant algorithm that handles PPG data well. It has been implemented in Python and C. Arduino IDE sketches for popular boards (Arduino, Teensy) are available to enable data collection as well. This provides both pc-based and wearable implementations of the software, which allows rapid reuse by researchers looking for a validated heart rate analysis toolkit for use in human factors studies.@en

Traffic congestion is a major societal challenge. By advising drivers on the optimal lane to drive, traffic flow can be improved, and congestion reduced. In this paper we describe the development of a lane-specific advice Human Machine Interface (HMI). Persuading drivers to follow an advice that is beneficial to the traffic situation, but may not be immediately beneficial to the drivers themselves, is challenging. In this paper we define persuasive elements to encourage drivers to follow the lane-specific advices. We then describe the interface design process, followed by its evaluation using a driving simulator study. In the simulator study, the effect of two types of persuasion are evaluated: a competitive variant where drivers could earn points and compete with others, and a cooperative variant where real-time information on the number of compliant drivers was available. Participants drove in the simulator on two days. Between days, the treatment groups viewed a Web-portal showing their performance and encouragement from an avatar. Those in the competitive and cooperative groups followed significantly more advices (117 and 111) than those in the control group (89). No significant differences were visible between competitive and cooperative groups. The differences between groups only emerged on the second day.@en

Cyclists may have incorrect expectations of the behaviour of automated vehicles in interactions with them, which could bring extra risks in traffic. This study investigated whether expectations and behavioural intentions of cyclists when interacting with automated cars differed from those with manually driven cars. A photo experiment was conducted with 35 participants who judged bicycle–car interactions from the perspective of the cyclist. Thirty photos were presented. An experimental design was used with between-subjects factor instruction (two levels: positive, neutral), and two within-subjects factors: car type (three levels: roof name plate, sticker–these two external features indicated automated cars; and traditional car), and series (two levels: first, second). Participants were asked how sure they were to be noticed by the car shown in the photos, whether the car would stop, and how they would behave themselves. A subset of nine participants was equipped with an eye-tracker. Findings generally point to cautious dispositions towards automated cars: participants were not more confident to be noticed when interacting with both types of automated cars than with manually driven cars. Participants were more confident that automated cars would stop for them during the second series and looked significantly longer at automated cars during the first.@en

Listening to music while cycling impairs cyclists’ auditory perception and may decrease their awareness of approaching vehicles. If the impaired auditory perception is not compensated by the cyclist himself or other road users involved, crashes may occur. The first aim of this study was to investigate in real traffic whether teenage cyclists (aged 16–18) compensate for listening to music by increasing their visual performance. Research in real traffic may pose a risk for participants. Although no standard ethical codes exist for road safety research, we took a number of ethical considerations into account to protect participants. Our second aim was to present this study as a case study demonstrating ethical dilemmas related to performing research in real traffic. The third aim was to examine to what extent the applied experimental set-up is suitable to examine bicyclists’ visual behaviour in situations crucial for their safety. Semi-naturalistic data was gathered. Participants’ eye movements were recorded by a head-mounted eye-tracker during two of their regular trips in urban environments. During one of the trips, cyclists were listening to music (music condition); during the other trip they were ‘just’ cycling (the baseline condition). As for cyclists’ visual behaviour, overall results show that it was not affected by listening to music. Descriptive statistics showed that 21–36% of participants increased their visual performance in the music condition, while 43–64% decreased their visual performance while listening to music. Due to ethical considerations, the study was therefore terminated after fourteen cyclists had participated. Potential implications of these results for cycling safety and cycling safety research are discussed. The methodology used in this study did not allow us to investigate cyclists’ behaviour in demanding traffic environment. However, for now, no other research method seems suitable to address this research gap.@en

Human physiology is a window to our physical, mental, and emotional states; our well-being. Today, a new wave of objective data derived from consumer grade body sensors-like those equipped by smartwatches-paves the way toward a new approach in how well-being is being measured, continuously and unobtrusively. Here, we developed a framework for collecting and analyzing physiological data using smartwatches in-the-wild, and demonstrated its robustness in data obtained away from controlled laboratory settings. We found that changes in people's heart rate and heart rate variability are predictive not of momentary well-being (a scientific idea that continues to live on in the absence of in-the-wild evidence, aka, zombie theory) but of daily well-being.@en

Human physiology is a window to our physical, mental, and emotional states; our well-being. Today, a new wave of objective data derived from consumer grade body sensors-like those equipped by smartwatches-paves the way toward a new approach in how well-being is being measured, continuously and unobtrusively. Here, we developed a framework for collecting and analyzing physiological data using smartwatches in-the-wild, and demonstrated its robustness in data obtained away from controlled laboratory settings. We found that changes in people's heart rate and heart rate variability are predictive not of momentary well-being (a scientific idea that continues to live on in the absence of in-the-wild evidence, aka, zombie theory) but of daily well-being.@en

Persuasive in-vehicle systems aim to intuitively influence the attitudes and/or behaviour of a driver (i.e. without forcing them). However, the challenge in using these systems in a driving setting, is to maximise the persuasive effect without infringing upon the driver's safety. This paper proposes a conceptual model for driver persuasion at the tactical level (i.e., driver manoeuvring level, such as lane-changing and car-following). The main focus of the conceptual model is to describe how to safely persuade a driver to change his or her behaviour, and how persuasive systems may affect driver behaviour. First, existing conceptual and theoretical models that describe behaviour are discussed, along with their applicability to the driving task. Next, we investigate the persuasive methods used with a focus on the traffic domain. Based on this we develop a conceptual model that incorporates behavioural theories and persuasive methods, and which describes how effective and safe driver persuasion functions. Finally, we apply the model to a case study of a lane-specific advice system that aims to reduce travel time delay and traffic congestion by advising some drivers to change lanes in order to achieve a better distribution of traffic over the motorway lanes. @en

Lower levels of automation are designed to work in specific conditions referred to as the Operational Design Domain (ODD). Beyond these conditions, the human driver is expected to take control. A mismatch between a driver's understanding and expectations of the automated vehicle capabilities and its actual capabilities as prescribed in the Original Equipment Manufacturers (OEMs) manual, could affect their safety and trust in automation. The main aim of this study is to develop a method for assessing the ODD of lane keeping system equipped vehicles. The analysis method is composed of an objective driving risk measure based on the Probabilistic Driving Risk Field (PDRF), and a subjective risk measure based on driver behavior, trust and situation awareness. We demonstrate the method applicability using the Automated Lane Keeping system of the Tesla Model S. A field test was conducted with 19 participants on public roads in the Netherlands including situations within and outside the defined ODD by the OEM. Across all test situations, a mismatch was observed between the ODD specified by the OEM and by the driver. Situations outside the ODD (i.e. no-lane markings and on/off-ramp) were often regarded as within the ODD by the participants. Situations inside the ODD (i.e. tunnel and curve) were mostly correctly classified by the participants. This analysis method has the potential to aid OEMs and road operators in defining more clearly the ODD while taking into account the driver's safety and awareness of the system capabilities.@en

Heart rate data are often collected in human factors studies, including those into vehicle automation. Advances in open hardware platforms and off-the-shelf photoplethysmogram (PPG) sensors allow the non-intrusive collection of heart rate data at very low cost. However, the signal is not trivial to analyse, since the morphology of PPG waveforms differs from electrocardiogram (ECG) waveforms and shows different noise patterns. Few validated open source available algorithms exist that handle PPG data well, as most of these algorithms are specifically designed for ECG data. In this paper we present the validation of a novel algorithm named HeartPy, useful for the analysis of heart rate data collected in noisy settings, such as when driving a car or when in a simulator. We benchmark the performance on two types of datasets and show that the developed algorithm performs well. Further research steps are discussed.@en

Heart rate data are collected often in human factors studies. Advances in open hardware platforms and offtheshelf photoplethysmogram (PPG) sensors allow the nonintrusive collection of heart rate data at very low cost. However, the signal is not trivial to analyse, since the morphology of PPG waveforms differs from electrocardiogram (ECG) waveforms and shows different noise patterns. PPG is often preferable because it can be collected less intrusively. However, few validated open source available algorithms exist that handle PPG data well, as most of these algorithms are specifically designed for ECG data. We have developed a novel algorithm specifically for PPG data collected in noisy fieldor simulatorbased settings. The main aim of this paper is to present the validation of a novel algorithm on a PPG dataset collected in a recent driving simulator experiment. The dataset was manually annotated, and performance of the algorithm compared to two other popular open source available algorithms. We show that the algorithm performs well and displays superior performance on the PPG dataset. Implications and further steps are discussed. @en

The present study aims to add to the literature on driver workload prediction using machine learning methods. The main aim is to develop workload prediction on a multi-class basis, rather than a binary high/low distinction as often found in litearature. The presented approach relies on measures that can be obtained unobtrusively in the driving environment with off-the-shelf sensors, and on machine learning methods that can be implemented on low-power embedded systems. Two simulator studies were performed, one inducing workload using realistic driving conditions, and one inducing workload with a relatively demanding lane-keeping task. Individual and group-based machine learning models were trained on both datasets and evaluated. For the group-based models the generalising capability, that is the performance when predicting data from previously unseen individuals, was also assessed. Results show that multi-class workload prediction on the individual and group level works well, achieving high correct rates and accuracy scores. Generalising between individuals proved difficult using realistic driving conditions, but worked very well in the high demanding lane-keeping task. Reasons for this discrepancy are discussed as well as future research directions.@en

The aim of this research is to work towards building an open-source, platform-independent algorithm capable of predicting driver workload in real-time and in a non-intrusive way. To work towards a system that can also be implemented in on-road settings, we aimed at using off-the-shelf, non-intrusive sensors that could be implemented into the steering wheel and dashboard of current and future generations of cars, making them non-intrusive. In order to build the initial predictive model, a driving simulator experiment was performed. Nineteen participants were required to drive a virtual replication of the Dutch A67 C-ITS corridor between Eindhoven and Venlo. We attempted to induce driver workload by varying weather, traffic composition, traffic density and by asking participants to perform various manoeuvres such as lane changing, merging and exiting. We measured heart rate, skin response, blink and performance measures. Results show that within individuals and within the experimental group, workload was predictable with a high correct rate in both individual models as well as group models. We also evaluated how well the models would generalise when used outside of the experimental setting. Preliminary results for this generalisation are poor. We discuss possible reasons for this and next steps we are planning to take to increase this performance.@en

To assess the potential impact of the higher speeds of pedal-assisted bicycles on safety, this study compared conventional bicycles, pedelecs and speed pedelecs (hereafter called s-pedelecs) on mean speeds, speed variability, harsh braking events (decelerations > 2 m/s), and mean speeds above the speed limit (MSAL) in rural and urban areas in the Netherlands Data were collected in daily traffic, while the legal maximum speed for speed-pedelecs was 25 km/h, and pedelecs and s-pedelecs shared the infrastructure with conventional bicycles. Data were collected, using two-wheelers equipped with accelerometers and GPS. Personality factors – sensation seeking and risk taking - were measured with surveys. Regular commuters used one of the three bicycle types for two weeks. Participant bias was intentionally included by allowing participants to select a bicycle type of their preference, resulting in 12 conventional bicycle riders (71 % women), 14 pedelec riders (67 % women) and 20 s-pedelec riders (25 % women). S-pedelecs were much faster than conventional bicycles, amounting to a speed difference with conventional bicycles of 10.4 km/h in urban areas (M =28.2 km/h vs. 17.8 km/h) and of 13.2 km/h in rural areas (M = 31.4 km/h vs. 18.2 km/h). The speed differences between pedelecs and conventional bicycles were much smaller: 2.3 km/h in urban areas (20.1 km/h vs 17.8 km/h) and 4 km/h in rural areas (22.2 km/h vs. 18.2 km/h). Compared to conventional bicycles, s-pedelecs varied their speed to a greater extent and also braked harshly more frequently, showing a greater need for speed adjustment. These adjustments were larger at higher speeds. In contrast, pedelecs did not differ from conventional bicycles on speed variation. MSAL for s-pedelec riders differed by gender. For men the MSAL was 87 % on urban sections and 91 % on rural sections. For women, the MSAL was lower, respectively 23 and 69 %. None of the personality factors were associated with speed variability, harsh braking or MSAL. However, sensation seeking was associated with higher mean speeds on all three bicycle types. To conclude, pedelecs and conventional bicycles are similar in speed patterns, whereas the speed patterns of s-pedelecs differ significantly from the former two. The safety implications are discussed.@en

The use of in-car technology has become more prevalent, both as driver assistance systems as well as connectivity or entertainment systems. Driver assistance systems can be built-in, after-market or run on a smartphone. The challenge however, is to increase drivers’ compliance with these systems. Stimulating the driver to adopt certain behaviours over others is a growing area within driver assistance systems. These approaches have the potential to be very effective, but only when they attract or persuade road users to use it. This study has two main aims. The first, is to provide a review of the different approaches in the literature to influence driver behaviour using in-car technology. The second aim is to develop a conceptual model to guide the research efforts on influencing driver behaviour at the tactical level. A structured review was conducted exploring the literature available in the automotive, behavioural and traffic safety domains. First, we explored the available psychological models that describe behaviour and their applicability to the driving task. Following this, we investigated the methodological ways used to influence behaviour with a focus on the traffic domain. Finally, a conceptual model is presented which encompasses the behavioural basis, methods and techniques for influencing the behaviour using in-car technology and the strategic planning of behavioural change technology. The role of driver workload within the model is discussed. Possible applications using the composite model are discussed.@en

Reducing congestion through persuasive in-car advice.@en

The authors regret that a mistake was made in Table 3 of the article. This table should read: [Table presented] The figures in bold, italic and red differ from the figures in the published article. As a result of these different values, the sixth sentence of the abstract should read: “S-pedelecs were much faster than conventional bicycles, amounting to a speed difference with conventional bicycles of 9.6 km/h in urban areas (M = 26.9 km/h vs. 17.3 km/h) and of 13.1 km/h in rural areas (M = 31.4 km/h vs. 18.3 km/h).” and the seventh sentence of the abstract should read: “The speed differences between pedelecs and conventional bicycles were much smaller: 2.8 km/h in urban areas (20.1 km/h vs 17.3 km/h) and 3.9 km/h in rural areas (22.2 km/h vs. 18.3 km/h). In the Discussion, Section 4.1. Comparison with previous studies on cycling speed characteristics, the second sentence of the second section should read: “Our study found similar patterns among Dutch riders. S-pedelecs were much faster than conventional bicycles, amounting to a speed difference in mean speeds in urban areas of 9.6 km/h (M = 26.9 km/h vs. 17.3 km/h) and in rural areas of 13.1 km/h (M = 31.4 km/h vs. 18.3 km/h). Values different from those in the published text are in bold, italic, and red. The authors would like to apologise for any inconvenience caused.@en

The present study aims to add to the literature on driver workload prediction using machine learning methods. The main aim is to develop workload prediction on a multi-level basis, rather than a binary high/low distinction as often found in literature. The presented approach relies on measures that can be obtained unobtrusively in the driving environment with off-the-shelf sensors, and on machine learning methods that can be implemented in low-power embedded systems. Two simulator studies were performed, one inducing workload using realistic driving conditions, and one inducing workload with a relatively demanding lane-keeping task. Individual and group-based machine learning models were trained on both datasets and evaluated. For the group-based models the generalizing capability, that is the performance when predicting data from previously unseen individuals, was also assessed. Results show that multi-level workload prediction on the individual and group level works well, achieving high correct rates and accuracy scores. Generalizing between individuals proved difficult using realistic driving conditions but worked well in the highly demanding lane-keeping task. Reasons for this discrepancy are discussed as well as future research directions.@en