Road user interactions in a shared space setting: Priority and communication in a UK car park

Highlights

• Behaviours in road user interactions recorded using novel observation protocol.

• Looking behaviour and use of signals associated with response of drivers.

• Communication between road users can lead to greater certainty during interactions.

• Results have implications for communication by automated vehicles.

Abstract

Appropriate communication between road users can lead to safe and efficient interactions in mixed traffic. Understanding how road users communicate can support the development of effective communication methods for automated vehicles. We carried out observations of 66 pedestrian-driver and 124 driver-driver interactions in a shared space setting. Specific actions and reactions of the road users involved were recorded using a novel observation protocol. Overall, results showed that pedestrians’ failure to look towards a driver created the greatest uncertainty in the interaction, with the driver slowing down, but not completely stopping, in response to pedestrians. Looking towards the driver also influenced which road user took priority in driver-driver interactions. Groups of pedestrians were more likely to be given priority than an individual pedestrian, and the use of vehicle-based signals were also associated with taking priority during an interaction. Our observations show the importance of non-verbal communication during road user interactions, highlighting it as an essential area of research in the development of automated vehicles, to allow their safe, cooperative, interactions with other road users. Observations were made on a limited number of interactions to inform challenges facing future automated vehicles. Further work should therefore be done to corroborate and extend our findings, to examine interactions between human road users and automated vehicles in shared space settings.

Introduction

Road traffic injuries are the eighth leading cause of death globally, and the first for people aged between 5 and 29 years (World Health Organization, 2018). Approximately 1.35 million people are killed on the world’s roads each year. More than half of these deaths are amongst vulnerable road users, including pedestrians (World Health Organization, 2018). Reducing road traffic casualties is a target for many national and international agencies (e.g. World Health Organization, 2015) and is particularly important if goals to encourage greater active travel are to be realised (Department for Transport, 2017, European Cycling Federation, 2017).

Communication between road users plays a significant role in road safety. For pedestrians, for example, information about a driver’s intentions, actions, and planned behaviour is important if the pedestrian is to assess the driver’s actions safely and accurately (e.g. Hamilton, Waterson, & Snell, 2014). Research suggests that failure to understand a driver’s behaviour may increase a pedestrian’s risk of being involved in a collision. For example, Otte, Jänsch, and Haasper (2012) found that more than half of the 475 pedestrian collisions they analysed were caused by a lack of safety-critical information being signalled to, or recognised, by the pedestrian. An analysis of the underlying causes related to pedestrian collisions showed “faulty diagnosis” was an important factor in pedestrian collisions. This meant that pedestrians had incorrectly assumed that they had been noticed by drivers (Bjorkman, 2008). Other research has also confirmed the significance of misinterpretation as a cause of collisions and incidents involving pedestrians. Habibovic and Davidsson (2012) examined the causes of collisions involving vulnerable road users (VRUs) and found that in 70% of cases, the VRU had seen the conflict vehicle but had misunderstood the traffic situation or made an inappropriate plan of action. Habibovic and Davidsson (2012) concluded that helping VRUs correctly understand traffic situations is essential for improving road safety.

In a mixed road setting, successful communication is essential, not only for good traffic flow and safe pedestrian-driver interactions, but also for driver-driver interactions. The importance of nonverbal communication for safe interactions between drivers has been known for decades, with Shor (1964), suggesting the nonverbal communication of driver intentions emerges out of shared social expectations in different situations, with problems arising when drivers do not have shared expectations. For instance, communication signals between drivers can be ambiguous, leading to potential conflicts, and reducing safety if they are misinterpreted (Risser, 1985). Vehicle-based signals are important between drivers, to show intent and planned behaviour, for example when using the left or right indicator signal to change lanes (Kauffmann, Winkler, Naujoks, & Vollrath, 2018). However, vehicle-based signals and nonverbal communication between drivers can be interpreted in different ways, which can impact on the potential safety of interactions. This is illustrated by cross-cultural variations in the interpretation of gestures. For example, the honk of a horn may be considered as an expression of anger or irritation in some countries, but in China it is frequently used as a friendly greeting, and in Southern Europe the same signal may be used alongside a rapid acceleration when a driver is merging into a gap in a lane of traffic (Färber, 2016). Receiving feedback about driving behaviour from other drivers is shown to reduce driving violations (Wang, Terken, Yu, & Hu, 2015), and successful communication between drivers is thought to lead to a more positive social driving climate and safer interactions (Zaidel, 1992).

Understanding how pedestrians and drivers communicate in an urban environment can help us understand how to improve the safety of pedestrian-driver and driver-driver interactions. The successful integration of automated vehicles (AVs) into the road transport system also requires a good understanding of how pedestrians and drivers communicate, and subsequently interact, with these new forms of transport. The Society of Automotive Engineers (SAE) currently provides a six-stage taxonomy of driving automation (SAE levels 0 to 5), where from SAE level 3 onwards, and based on the particular operational design domain, the vehicle (rather than driver) undertakes all aspects of driving control, including object and event detection and response (SAE International, 2018). Therefore, it can be argued that occupants of higher level automated vehicles are not necessarily required to attend to the events of the road, and may, therefore, not engage in any communication with other road users during conflict situations, for example when they are sharing the same road space. It can also be envisaged that SAE Level 4 and 5 vehicles may travel without any occupants at all. Therefore, understanding how interactions between road users currently unfold will provide insights into the interaction strategies and communication requirements of AVs in the future.

In recent years, a growing body of research has focussed on the use of external Human-Machine Interfaces (eHMI) for communicating the intentions or behaviour of AVs (e.g. Habibovic et al., 2018, Clamann et al., 2015, Deb et al., 2018, Hensch et al., 2019, Rettenmaier et al., 2020). However, findings within this research area are mixed. For example, Clamann et al. (2015) examined the effect of three different eHMI signals used to communicate with a pedestrian about to cross a road. These signals were intended to provide information to support the pedestrian’s decision making, regarding whether or not to cross the road. No effect on response times to cross were found for the three different signals, and vehicle behaviour (speed and braking profile) was found to be more important. In contrast, however, Mahadevan, Somanath, and Sharlin (2018) reported that pedestrians preferred to receive explicit information about a vehicle’s intentions via an eHMI, rather than deducing the vehicle’s intentions from its motion cues. Therefore, a first step in helping resolve some of the discordant findings within research into automated vehicle communication is to understand how communication and interactions currently take place between human road users in a real world setting, to ensure these new forms of technology provide the right, and most clearest, information to all road users (Schieben et al., 2019).

Previous work has shown that the interaction between pedestrians and drivers is influenced by the behaviour of other road users (Rosenbloom, 2009), the speed and stopping distance of vehicles (Sun, Zhuang, Wu, Zhao, & Zhang, 2015), as well as pedestrian/driver demographics (e.g. Tom & Granié, 2011). Studies have also highlighted the use of non-verbal communication cues during pedestrian-driver and driver-driver interactions. For example, Sucha, Dostal, and Risser (2017) found that signals provided by the driver, such as eye contact, hand waving, or flashing the vehicle lights were important factors in determining whether a pedestrian decided to cross at a marked crossing. Rasouli, Kotseruba, and Tsotsos (2017) found that, before crossing a non-signalised crosswalk, pedestrians looked at an approaching car in more than 90% of cases, and provided some form of explicit communication in 15% of cases, such as nodding or using a hand gesture. However, whether the pedestrian actually chose to cross also depended on other factors, including the driver’s response. For example, pedestrians were more likely to cross if the driver acknowledged their intention to cross, by slowing down, or stopping the vehicle. This highlights the importance of reciprocal communication between road users during an interaction, in determining how that interaction unfolds – to understand such reciprocal communication, we need to observe the actions and reactions of different road users.

Most previous studies of pedestrian-driver and driver-driver interactions (e.g. Sucha et al., 2017, Rosenbloom, 2009, Salamati et al., 2013) have been conducted on relatively regulated road sections, with well-understood road rules and formal, universally accepted guidelines (for example, as designated by the Highway Code in the United Kingdom – Department for Transport, 2018). The value of these rules is that they potentially reduce the likelihood of uncertainties in interactions. For example, drivers are expected to give way to pedestrians waiting to cross at designated locations, such as zebra crossings. Likewise, a driver approaching the main road from a side road is required to wait for an appropriate gap in traffic, since the right of way is with the drivers on the main road. However, interactions between road users are likely to be more uncertain and ambiguous in un-signalised road sections, where there are no clear rules of the road or behavioural norms (Kaparias, Bell, Miri, Chan, & Mount, 2012). It can be argued that future AVs may benefit from some type of eHMI in such scenarios, which may communicate the planned actions of the vehicle to other road users, and reduce uncertainty during an interaction. Research is therefore required to further understand pedestrian-driver and driver-driver interactions in such environments, which are not formally governed by rules and standards.

Shared spaces are a good example of traffic environments that do not function based on formal rules and standards. The shared space concept is an urban design approach where different types of road users move and interact with each other on the basis of informal social protocols and negotiation (Hamilton-Baillie, 2008). Interactions in shared spaces normally take place between low speed vehicles and other road users, in potentially ambiguous situations, where the intended actions of either driver or pedestrian are unclear, and it is not obvious who has priority. Shared spaces are commonly seen as specifically designed and engineered to promote safe interaction between different types of road users. They may use specific design principles, such as limited demarcation between roads and footpaths, restriction of vehicle speeds through street design, and clearly marked access points to the shared space (Karndacharuk, Wilson, & Dunn, 2014). Despite a growth in popularity of the concept of specifically-designed shared spaces (as evidenced by the introduction of specific transport planning guidance related to shared spaces, e.g. Department for Transport, 2011), they remain relatively uncommon in urban contexts. Car parks (also known as parking lots in North America), however, represent a common example of a shared space, due to the use of social protocols and negotiation during interactions between drivers (searching for, or leaving, a parking space) and pedestrians (travelling to, or from, their vehicle). Car parks therefore provide an important, but under-researched, context for understanding interactions between different road users.

Shared spaces (including car parks) are associated with lower vehicle speeds, and these lower speeds have been shown to lead to more conflicts between road users (Salamati et al., 2013), where a conflict is defined as “an observational situation in which two or more road users approach each other in space and time to such an extent that a collision is imminent if their movements remain unchanged” (Svensson, 1998). However, it can also be argued that shared spaces, such as car parks, may lead to a reduction in conflicts. This can be due to increased vigilance and better cooperation between road users, to manage the higher number of likely interactions (e.g. Kaparias et al., 2013). Such shared spaces, therefore, provide a valuable context in which to study pedestrian-driver and driver-driver interactions, to understand how potential conflicts are resolved between the two types of road user.

With regards to designing more successful communication strategies for future AVs, this type of observation can be useful for understanding how priority is determined during conflicts between different road users. This knowledge may, for example, help avoid deadlock situations, where a lack of communication prevents either actor from moving forward (Imbsweiler, Ruesch, Weinreuter, León, & Deml, 2018). Shared space settings have been highlighted as an important scenario to be understood when considering the introduction AVs into mixed traffic, and the behaviour of AVs in a shared space is an important research question (Parkin, Clark, Clayton, Ricci, & Parkhurst, 2018). Therefore, a first step towards addressing this research question is to investigate the interactions between existing road users in such a setting.

To further investigate the factors which determine how priority is established between different road users in a shared space, this study used a bespoke observation protocol (Dietrich, Bengler, & Portouli, 2018) to investigate road-user behaviour in a railway station car park in Leeds, UK. One key aim of the study was to establish how behaviours in the initial phase of an interaction were associated with the final outcome of the interaction. The study was part of the wider EU-funded project ‘interACT’ (grant number 723395), the overall goal of which was to understand current road user interactions, and apply this knowledge to AV strategies for communication and interaction. Specifically, our two main research questions were: 1) How are behaviours in the initial phases of an interaction, such as looking and hand signals, related to the latter phases of that interaction; and 2) What are the factors that determine which road user takes priority in an interaction?