This study explores the speed perception of electric scooters (e-scooters) compared to pedestrians in a virtual environment, based on the hypothesis that e-scooters are perceived as slower due to the absence of walking-related movements, such as arm swings. Participants compared a pedestrian moving at 1.3 m/s and an e-scooter traveling at 1.0-1.6 m/s. Results suggest that, despite individual diferences, e-scooters tend to be perceived as slower. Future research will increase the number of participants and trials and simulate real-world scenarios to validate our assumption quantitatively.
debris [5, 9]. The speed diference between e-scooters and pedestrians on sidewalks is also a factor. Since ePersonal Mobility Vehicles (PMVs) are small transporta- scooters travel faster than pedestrians, the risk of collition devices designed for short-distance travel, and they sions increases, especially on narrow or crowded sideare expected to contribute to easing trafic congestion, walks [10, 11]. reducing CO2 emissions, and addressing the "last mile" This study focuses on pedestrians’ perception of eproblem. Among them, e-scooters have rapidly gained scooter speeds and aims to test the hypothesis that pedespopularity as an eficient means of transportation in ur- trians are prone to misjudging the speed of e-scooters. A ban areas [1, 2, 3]. virtual environment will be used to conduct an experi
However, with the increasing prevalence of e-scooters, ment evaluating speed perception errors. In this experconcerns about their safety have also risen [1, 4, 5]. For iment, we will measure how pedestrians perceive the instance, in some cities in Europe and the United States, speed of other pedestrians and e-scooters around them stricter regulations have been implemented due to con- and analyze the accuracy of their perception. cerns over trafic accidents and safety, with an increasing number of cases where e-scooters are prohibited from riding on sidewalks [1]. 2. Related Research
In contrast, in Japan, since July 1, 2023, e-scooters are allowed to travel on sidewalks because their maximum speed does not exceed 6 km/h [6]. However, during the period when sidewalk riding was prohibited, from 2020 to Jan. 2023, 76 accidents involving e-scooters were reported, 8 of which involved pedestrians [7]. As the number of e-scooters on sidewalks increases, there are concerns about a rise in accidents involving pedestrians, and indeed, an increase in such accidents has been reported overseas [8, 5].
The causes of accidents involving e-scooters include factors related to the scooters themselves, such as their small tires, which make them more susceptible to tipping over or colliding with obstacles like bumps and In addition to the studies mentioned above, research on behavioral prediction tasks has also been conducted. A behavioral prediction task involves predicting the time it takes for a moving object to reappear, which temporarily disappears due to occlusion or other factors. Battaglini et al. conducted a study using images of bicycles and motorcycles, demonstrating that when the moving speed is low, the symbolic meaning of the objects (for example, a motorcycle being perceived as faster than a bicycle) significantly influences the prediction of the object’s movement [12].
Furthermore, numerous studies have investigated personal mobility vehicles (PMVs) and e-scooters. Paudel et al. explored collisions between e-scooters and pedesAPMAR’24: The 16th Asia-Pacific Workshop on Mixed and Augmented trians in a virtual environment, revealing that the risk Reality, Nov. 29-30, 2024, Kyoto, Japan of pedestrians sustaining head injuries sharply increases * Corresponding author. when the speed of the e-scooter is between 10 and 15 ($A.eUot2s1ui1m0i4);@ygaumha.zuo-eh@yoegnog.a.uc-.jhpy(oHg.oM.aca.tjspu(uHr.a)Y;aumtsauzmoei@)atr.jp km/h [13]. Additionally, the research by Hishikawa et © 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License al. conducted a questionnaire survey on the appearance Attribution 4.0 International (CC BY 4.0). of PMVs and pedestrian risk perception, reporting that the appearance of a PMV influences how pedestrians perceive risk [14].
In contrast, this study focuses not on subjective risk perception but on speed perception. It aims to evaluate how the perception of speed changes when pedestrians and e-scooters approach within a virtual environment.
The following experiment was conducted using a virtual environment to evaluate how participants perceive the speed of approaching e-scooters and pedestrians and Figure 1: Experimental environment. whether there is a diference in perceived speed.
In the experiment, participants wore a head-mounted display (Meta Quest 3, 72Hz refresh rate, resolution 3616×1952 pixels, field of view: horizontal 110 degrees, vertical 96 degrees) and observed e-scooters and pedestrians approaching in the virtual environment, then evaluated their speed. Figure 1 1 shows the experimental setup in the virtual environment. Participants stood at a designated standing point in the virtual environment and were instructed to focus on a fixation point 8.5 meters ahead and 1.7 meters above the ground throughout the experiment.
Participants observed two scenarios: a pedestrian approaching (left side in Figure 2) and a person on an escooter approaching (right side in Figure 2). They were asked to report which of the two felt faster. To prevent Figure 2: Pedestrian and E-scooter in VR environment. speed judgment based on the time taken to appear and approach, the pedestrian and the e-scooter appeared randomly at distances of 5m, 6m, or 7m from the partici- Participants were allowed to take a break every five trials pant. They disappeared once they approached within if they wished. The experiment included 9 participants less than 1 meter. The second stimulus was presented (eight males and one female, with a mean age of 22.1 three seconds after the first. This constituted one trial. years and a standard deviation of 1.07 years), all of whom The walking speed for the pedestrian was set at the aver- had normal vision. age walking speed of an adult male, 1.3 m/s [15], while the speed of the e-scooter was set at 1.1, 1.2, 1.3, 1.4, and 1.5 m/s. The presentation order of the pedestrian and 4. Results scooter, as well as the speeds, were counterbalanced. The experiment was conducted under two conditions: the Figures 3 and 4 present the experiment results. Each approaching object came from the front (0 degrees) or 30 figure shows the results for the approach angles of 0 and degrees outside the participant’s field of view. 30 degrees, respectively. The horizontal axis represents
The virtual environment was created using Unity the speed of the e-scooter (1.1–1.5 m/s). In contrast, the (2023.2.20f1). For the male model, [16] was used, and vertical axis shows the average proportion (with standard for the e-scooter model, [17] was utilized. Additionally, deviation) of participants who judged that "the e-scooter to simulate the natural movements of a person riding is faster" compared to a pedestrian walking at 1.3 m/s. an e-scooter, the pitch and yaw angles of the head were The curve indicates the logistic curve fitted to the data. randomly varied. The results show that, for both the 0-degree and 30
The experiment followed this process: First, the partic- degree conditions, the proportion of participants judging ipants were explained the experiment, and verbal consent the e-scooter as faster increased as the scooter’s speed to participate was obtained. Then, the participants put increased. This suggests that participants were generon the HMD, and the experiment began. For each an- ally able to perceive the approaching speeds accurately. gle condition, five trials were conducted for each of the Specifically, focusing on the scooter speed of 1.3 m/s, the ifve-speed conditions of the e-scooter, totaling 25 trials. average proportion of responses indicating "the e-scooter is faster" was 0.4 under the 0-degree condition, whereas Table 1 it was 0.533 under the 30-degree condition. This suggests Diference threshold (5 conditions) that the e-scooter might be perceived as slower under the 0-degree condition. 0 degrees 30 degrees
Next, the point of subjective equality (PSE), where Diference threshold 0.137 0.155 the choice ratio reaches 0.5, was calculated based on the logistic curve fitting. The PSE for the 0-degree condition Table 2 was 1.317 m/s, and for the 30-degree condition, it was PSE and Diference thresholds (7 conditions) 1.334 m/s. Both values were slightly higher than 1.3 m/s. These results suggest that the e-scooter’s speed PSE Dif. thresholds tends to be perceived as slower, aligning with the initial participant 0 deg. 30 deg. 0 deg. 30 deg. hypothesis. 1 1.326 1.316 0.052 0.073
Next, based on the results shown in Figures 3 and 4, we 2 1.275 1.240 0.055 0.129 found that none of the conditions yielded response ratios 3 1.231 1.226 0.128 0.167 close to 0 or 1. This suggested that the speed range for 4 1.343 1.475 0.100 0.113 the e-scooter might not have been suficient for accurate 5 1.361 1.315 0.111 0.093 logistic curve fitting. Therefore, the speed conditions 6 1.450 1.354 0.070 0.049 were extended to cover a range of 1.0–1.5 m/s, with incre- 7 1.400 1.584 0.145 0.336 ments of 0.1 m/s, resulting in seven conditions. An addi- 8 1.364 1.230 0.136 0.236 tional experiment was conducted with eight participants, with five trials for each speed condition. Furthermore, due to the large standard deviations and variability in ence thresholds calculated based on the logistic curve participant results, the data for each participant were iftting results for each condition. The results show that, summarized separately. Figures 5–8 present the results although the magnitude of the diferences varies, the for two participants (1 and 2) out of the eight, whose PSEs are higher under the 0-degree conditions, whereas results exhibited contrasting tendencies. the diference thresholds are larger under the 30-degree
In the results for Participant 1 (Figures 5 and 6), when conditions. comparing the pedestrian moving at 1.3 m/s with the e-scooter, the proportion of responses indicating "the 5. Discussions e-scooter is faster" was 0.4 in both trials. The points of subjective equality (PSE) were 1.326 m/s and 1.316 m/s, This study evaluated how participants perceive the aprespectively, indicating that the e-scooter tended to be proaching speeds of e-scooters and pedestrians in a virperceived as slower than the pedestrian. tual environment. When comparing a pedestrian walk
Conversely, for Participant 2 (Figures 7 and 8), the pro- ing at 1.3 m/s with an e-scooter, the proportion of reportion of responses indicating "the e-scooter is faster" sponses indicating that "the e-scooter is faster" was bewas 0.6 in both trials. The PSE values were 1.275 m/s low 0.5 in most cases. Similarly, the points of subjective and 1.240 m/s, showing a tendency for Participant 2 to equality (PSE) were above 1.3 m/s for most participants. perceive the e-scooter as faster than the pedestrian. These results suggest that, despite individual diferences, Finally, Tables 1 and 2 present the PSE and the difer- e-scooters tend to be perceived as slower than they are.
One possible reason for this perceptual diference is the absence of large bodily movements, such as arm swings or leg motions, typically accompanying walking. Al- This study conducted a speed perception experiment in though statistical testing has not yet been conducted, a virtual environment based on the hypothesis that ethe additional experimental results showed higher PSE scooters are perceived as slower than pedestrians due to values under the 0-degree condition (six out of eight the absence of bodily movements associated with walkparticipants). This suggests that central vision (0 de- ing, such as arm swings and leg motions. In the experigrees), which is more sensitive to fine movements, may ment, participants observed a pedestrian approaching at be more influenced by the absence of arm swings and 1.3 m/s and an e-scooter approaching at speeds ranging other walking-related motions. In contrast, Hassan et from 1.0 to 1.6 m/s, and they were asked to report which al., although not investigating approaching objects, con- felt faster. Although the results showed large individducted experiments on speed perception of moving light ual diferences, they suggest that e-scooters tend to be points and reported that speeds tend to be perceived as perceived as slower. slower in peripheral vision [18]. These findings indicate We plan to expand the experiment by increasing the that multiple factors may interact to influence speed per- number of participants and trials to assess the diferences ception, and future experiments will aim to investigate in speed perception further. Additionally, we aim to conthese efects further. duct experiments in more realistic scenarios to simulate
Regarding the relationship between approach angle real conditions better and validate these findings. and diference threshold, Tables 1 and 2 show that the diference threshold was larger under the 30-degree con- Acknowledgments dition than the 0-degree condition. This may be because, under the 30-degree condition, participants relied on pe- This research was supported by JSPS KAKENHI Grant ripheral vision, which likely resulted in lower speed per- Numbers JP23H01387 and JP21K11968. ception accuracy than in the 0-degree condition. [9] W. Li, C. J. van Driel, J. C. de Winter, M. H. Martens,
How do diferent micro-mobility vehicles afect [1] J. A. Bigazzi, D. Wong, Dockless electric scooters: longitudinal control? results from a field experiA review of a growing micromobility mode, In- ment, Journal of Safety Research 83 (2022) 215–226. ternational Journal of Sustainable Transportation doi:10.1016/j.jsr.2022.08.011. 16 (2022) 846–860. doi:10.1080/15568318.2022. [10] J.-Y. Lee, H.-H. Bai, K. Choi, Users’ attitudes 2044097. on electric scooter riding speed on shared foot[2] H. Yu, Y. Chai, C. Xie, Shared e-scooter micromobil- path: A virtual reality study, International Journal ity: review of use patterns, perceptions and environ- of Sustainable Transportation 15 (2021) 647–654. mental impacts, Transport Reviews 43 (2023) 383– doi:10.1080/15568318.2020.1718252. 404. doi:10.1080/01441647.2023.2171500. [11] M. Wu, K.-C. Tung, Z. Cao, S. Chen, Pedestrians’ [3] R. L. Sanders, M. Branion-Calles, T. A. Nelson, To safety using projected time-to-collision to electric scoot or not to scoot: Findings from a recent survey scooters, Nature Communications 14 (2023) 50049. about the benefits and barriers of using e-scooters doi:10.1038/s41467-023-50049-x. for riders and non-riders, Transportation Research [12] L. Battaglini, G. Mioni, The efect of symbolic meanPart A: Policy and Practice 139 (2020) 217–227. ing of speed on time to contact, Acta psychologica doi:10.1016/j.tra.2020.07.009. 199 (2019) 102921. URL: https://www.sciencedirect. [4] S. Hock, M. Branion-Calles, J. Gilliland, Elec- com/science/article/pii/S0001691818305444. tric scooter safety: An integrative review of evi- doi:https://doi.org/10.1016/j.actpsy. dence from transport and medical research domains, 2019.102921.
Journal of Transport & Health 26 (2022) 101328. [13] M. Paudel, F. F. Yap, T. B. M. Rosli, K. H. Tan, H. Xu, doi:10.1016/j.jth.2022.101328. A computational investigation of the dynamic fac[5] N. Salamon, K. Thrun, M. Razaghi, R. M. Pendyala, tors governing severity of head injury to pedestriS. Khoeini, S. Joshi, K. Ismail, Characteristics and ans involved in e-scooter collisions, Transportation risk factors for electric scooter-related crashes and Research Interdisciplinary Perspectives 22 (2023) injury crashes among scooter riders: A two-phase 100972. survey study, International Journal of Environ- [14] T. HISHIKAWA, M. IRYO, Y. HASEGAWA, Impact mental Research and Public Health 19 (2022) 10129. of personal mobility vehicle types on perception of doi:10.3390/ijerph191610129. risks by pedestrians (in japanese), Journal of Japan [6] Tokyo Metropolitan Police Department, Regard- Society of Civil Engineers, Ser. D3 (Infrastructure ing trafic rules for specified small motorized bi- Planning and Management) 75 (2019) I_595–I_605. cycles (in japanese), https://www.keishicho.metro. [15] R. W. Bohannon, Normal walking speed: a descriptokyo.lg.jp/kotsu/doro/tokuteikogata.html, 2023. tive meta-analysis, Physiotherapy 83 (1997) 77–84.
Accessed: October 18th, 2024. [16] AtelierBrown, Free 3D model A-Pose [7] National Police Agency, Japan, Status of traf- 068_Syun, https://ddd.pink/product/ ifc violations and accidents related to electric free-3d-model-a-pose-068_syun, n.d. Accessed: kickboards, https://www.npa.go.jp/bureau/trafic/ October 18th, 2024. council/newmobility0503.pdf, 2023. Accessed: Oc- [17] ERLHN, Turbosquid electric scooter, tober 18th, 2024. https://www.turbosquid.com/3d-models/ [8] L. J. Mayhew, C. Bergin, T. Knight, B. Mitra, In- free-electric-scooter-3d-model-2022443, n.d. jury patterns and circumstances associated with Accessed: October 18th, 2024. electric scooter collisions: a scoping review, In- [18] B. Roulston, Z. Jiao, M. Barnett-Cowan, Perceived jury Prevention 27 (2021) 490–499. doi:10.1136/ speed in peripheral vision can go up or down, Jourinjuryprev-2020-043988. nal of Vision 17 (2017) 3. doi:10.1167/17.12.3.