Source-linked AI summary
Occlusion-induced risk and interventions in pedestrian-autonomous truck interactions on multi-lane roads: A virtual reality study
Yun Ye, Yuan Che, S. C. Wong, Stergios-Aristoteles Mitoulis, Haoyang Liang
TL;DR
Autonomous trucks may create pedestrian-safety risks through truck characteristics, constrained braking, limited communication cues, and occlusion of surrounding traffic. Using a controlled virtual-reality experiment, this study tested risk formation in unsignalized multi-lane crossings and compared three occlusion-targeted interventions. ATs increased perceived risk and cautious behavior, but rain weakened this compensation; projected eHMI performed best overall by improving safety margins, risk awareness, and behavioral adjustment.
Problem
The study addresses limited understanding of how autonomous trucks simultaneously create direct pedestrian-interaction risk and occlusion-induced risk involving surrounding traffic.
Method
A controlled virtual-reality experiment with 54 participants examined vehicle type, weather, far-side yielding strategy, and three targeted interventions in an unsignalized multi-lane crossing.
Results
Projected eHMI showed the most favorable overall performance, improving DST, significantly increasing ITADV, and supporting behavioral adjustment consistent with risk localization.
Takeaways & Limitations
AT pedestrian-safety interventions should address both intention communication and localization of hidden risk sources.
Takeaways & Limitations
The study used virtual reality, mainly young university participants, and a specific unsignalized two-lane crossing scenario.
Abstract
from arXiv · showhide
Autonomous trucks (ATs) may introduce distinct pedestrian-safety risks because of their large physical dimensions, constrained braking capability, limited driver-based communication cues, and potential to occlude surrounding traffic. This study employed a controlled virtual reality experiment with 54 participants to investigate pedestrian-AT interaction risk in an unsignalized multi-lane crossing scenario and to evaluate occlusion-targeted risk mitigation strategies. The experiment examined the effects of near-side vehicle type, weather condition, and far-side vehicle yielding strategy on pedestrian behavior, perceived risk, and objective safety. Based on a representative high-risk scenario, three targeted interventions were designed and tested: an environment-aware external human-machine interface (eHMI), a projected eHMI, and an auditory warning. The results showed that ATs increased perceived risk and encouraged more cautious crossing behavior, suggesting a risk-compensation effect. However, this compensation was weakened under rainy conditions, where braking-related safety margins were reduced. AT-induced occlusion further increased far-side interaction risk by limiting pedestrians' recognition of hidden vehicles. Among the three interventions, the projected eHMI showed the best overall performance, improving objective safety margins, enhancing risk awareness, and supporting behavioral adjustment. These findings highlight the need for AT-specific interface and warning strategies that address both intention communication and risk localization.
7 table in the Appendix Occlusion-induced risk and interventions in pedestrian-autonomous
The supplied passages identify the paper’s authors, institutional affiliations, acknowledgements, and keywords.
- The paper lists Yun Ye, Yuan Che, S.C. Wong, Stergios-Aristotelis Mitoulis, and Haoyang among its authors.
- The listed affiliations include research and transportation institutions in Ningbo, Hong Kong, Shanghai, and Tongji University.
- The keywords cover autonomous trucks, external human–machine interfaces, pedestrian safety, virtual reality, and occlusion-induced risk.
1. Introduction
The introduction frames autonomous trucks as a distinct pedestrian-safety problem involving direct truck interaction and truck-induced occlusion. It motivates a virtual-reality study of multi-lane crossings and targeted interventions that communicate hidden risks.
- Autonomous trucks raise pedestrian-safety concerns because freight environments include uncontrolled crossings where decisions depend on surrounding-vehicle interactions.
- ATs combine large dimensions, higher potential crash severity, constrained braking, and fewer driver-based communication cues, shaping risk perception and crossing decisions.
- Truck-induced occlusion can conceal far-side vehicles or conflict areas, creating secondary risks beyond direct pedestrian–truck interaction.
- Conventional eHMIs mainly communicate ego-vehicle awareness or yielding intention, which may be insufficient when the critical hazard is hidden beyond the truck.
- Rain may reduce braking performance and safety margins, while far-side yielding strategy can influence risk recognition after pedestrians pass the occluding vehicle.
- The study uses controlled virtual reality to examine multi-lane interaction risk and evaluate environment-aware eHMI, projected eHMI, and auditory-warning interventions.
2. Literature Review
The literature review distinguishes direct truck-interaction risk from occlusion-induced secondary risk and surveys behavioral, surrogate-safety, communication, and warning measures. It identifies unresolved needs for indicators and interventions that localize hidden risks.
- AT risk attributes: AT-related pedestrian risk follows two interconnected pathways: direct interaction with the truck and occlusion-induced secondary risk involving surrounding traffic.
- AT risk attributes: Truck size, mass, braking constraints, and missing driver cues can increase collision severity, reduce stopping capability, and increase uncertainty about yielding intention.
- Occlusion-induced risk: Large truck geometry can restrict visibility and create hidden-risk conditions in which the AT is both the interaction counterpart and an occluding object.
- Risk measurement: Existing behavioral indicators describe crossing decisions and movement execution but do not directly quantify whether sufficient safety margin remains.
- Risk measurement: TTC, PET, and ITADV capture different temporal conflict dimensions, while pedestrian-AT studies require measures that also reflect braking dynamics and occlusion.
- Interventions: Conventional eHMIs communicate vehicle status or yielding intention, whereas auditory warnings may have limited spatial directivity for identifying the precise risk source.
- Interventions: The review therefore leaves open whether visual and auditory measures can support occlusion awareness, hidden-risk recognition, spatial localization, and objective safety improvement.
3. Methods
The study used a controlled two-stage VR design to examine pedestrian–autonomous truck risk and test occlusion-targeted interventions in an unsignalized multi-lane crossing scenario. It manipulated vehicle type, weather, and far-side interaction strategy while measuring behavioral, perceived-risk, and objective-safety outcomes.
- Study design: The two-stage design examined risk formation under varied vehicle, weather, and far-side strategy conditions, then evaluated interventions in a representative high-risk scenario.The high-risk scenario combined strong truck-induced occlusion, adverse weather-related braking constraints, and a non-yielding far-side vehicle.
- Platform and apparatus: The VR platform combined head-mounted display, interactive movement control, dynamic vehicle simulation, and synchronous recording of pedestrian and vehicle behavior.Participants completed the task from a first-person perspective in a consistent virtual scene across conditions.
- Experimental scenario: The crossing scenario used a one-way, two-lane road with near-side vehicles that directly interacted with pedestrians and maintained visual occlusion during first-lane crossing.Participants began at the curb and crossed toward the opposite side; the leading near-side vehicle started 40 m away and the following vehicle was 5 m behind.
- Parametric design: Near-side vehicle type compared large autonomous trucks with passenger AVs, while weather varied between sunny and rainy conditions affecting perception and braking capability.The ATs measured 9.5 m × 4 m × 4.1 m, whereas passenger AVs measured 5 m × 2 m × 1.5 m.
- Interventions: Three interventions addressed occlusion by linking warnings to the occluding truck, the far-side conflict area, or a non-visual warning channel.The interventions were designed to support occlusion awareness, hidden-risk recognition, and spatial risk localization.
- Measures: DST measured the ratio between the minimum deceleration required to avoid collision and the vehicle’s actual average deceleration, capturing direct-interaction safety margins.DST is dimensionless and incorporates vehicle dynamics and weather-related braking-efficiency changes.
4. Results
The experiments found that ATs heightened perceived risk while producing mixed objective-safety effects across lanes and conditions. Rain weakened the apparent safety benefit of cautious crossing, and AT-related occlusion increased far-side interaction risk.
- Lane-specific risk: Second-lane perceived risk was higher than first-lane perceived risk (7.51 versus 6.51), while second-lane safety showed greater instability across conditions.The descriptive results characterized the scenario as phase-dependent, with second-lane interactions involving unstable time advantage and higher subjective danger.
- Intervention experiment: The intervention experiment targeted a constrained high-risk scenario in which DST was near the critical boundary and ITADV was relatively low.Its purpose was to test whether targeted interventions could improve behavior and safety under a more critical interaction context.
- Risk perception: ATs significantly increased first-lane perceived risk relative to passenger AVs (β = 3.046, p < 0.001).Participants were sensitive to truck-related attributes including larger body size and stronger visual pressure.
- Risk perception: Rain significantly increased first-lane perceived risk (β = 1.000, p < 0.001), showing that pedestrians incorporated adverse weather into their judgments.
- Objective risk: Compared with passenger AVs, ATs were associated with lower DST, indicating a larger first-lane temporal safety margin consistent with possible risk compensation.The authors attribute this pattern to more conservative crossing timing in response to ATs’ salient threat cues.
- Objective risk: The AT–weather interaction was significant for DST (β = 0.383, p < 0.05), indicating that rainy conditions weakened this apparent compensation by constraining braking efficiency.
- Lane-specific risk: ATs significantly reduced second-lane ITADV, with a total effect of −1.494 (95% bias-corrected bootstrap CI [−1.829, −1.159]), increasing objective interaction risk.The mediation analysis reported a 27.10% mediated proportion and a 72.90% direct-effect proportion.
5. Discussion
The discussion interprets AT risk as both direct and occlusion-mediated, with pedestrians recognizing salient danger cues but not fully capturing combined dynamic effects. Among the tested interventions, projected eHMI best aligned warnings with the hidden conflict area.
- Risk structure: ATs can increase perceived risk while also eliciting conservative crossing timing, but this compensation becomes less effective when rain reduces braking efficiency.
- Risk structure: Pedestrians may recognize ATs and rain as separate danger cues without fully internalizing their combined effect on actual safety margins.The combined AT–rain effect appeared in objective risk but not in the perceived-risk model.
- Occlusion-induced risk: AT occlusion increased subsequent far-side risk by restricting information acquisition and compressing pedestrians’ time advantage.Mediation separated a timing-related pathway from a remaining direct effect, supporting an occlusion-related mechanism.
- Occlusion-induced risk: The findings extend conventional eHMI concerns from communicating ego-vehicle intention to supporting hidden-risk recognition and spatial risk localization.Large vehicle bodies can affect pedestrians’ later interactions with vehicles beyond the truck itself.
- Intervention effectiveness: Projected eHMI significantly reduced DST, increased ITADV, and produced more favorable behavioral adjustment by placing warnings near the relevant conflict space.Its advantage was linking warning information directly to the potential occluded conflict area.
- Intervention effectiveness: Environment-aware eHMI changed perceived risk, waiting, and observation behavior but did not significantly improve first-lane DST or second-lane ITADV.Vehicle-mounted information may focus attention on the truck rather than the visually blocked far-side hazard.
- Intervention effectiveness: Auditory warning increased perceived risk, but unstable objective-safety effects reflected its weak spatial directivity in occlusion scenarios.It may be more suitable as a supplementary risk-arousal method than as a standalone spatial-risk aid.
6. Conclusions
The study found that autonomous trucks create both direct interaction and occlusion-induced risks, while projected eHMI provided the most favorable overall intervention performance. The findings support interventions that help pedestrians both detect danger and locate its source, within the tested scenario.
- Study design: The two-stage VR study examined pedestrian-AT risk formation and three interventions in an unsignalized multi-lane crossing scenario.The first stage varied weather, near-side vehicle type, and far-side interaction strategy; the second evaluated environment-aware eHMI, projected eHMI, and auditory warning.
- Risk formation: ATs affected interaction risk through both direct pedestrian-truck interaction and occlusion-induced risk.In the second lane, reduced ITADV was associated with both second-lane entry timing and occlusion caused by the large truck body.
- Intervention effects: All three interventions increased pedestrians’ risk awareness, but their effects on behavior and objective safety differed.The interventions were designed around the need to identify hidden risks, recognize conflict areas, and provide warnings under visually constrained conditions.
- Intervention effects: Projected eHMI showed the most favorable overall performance by improving DST, significantly increasing ITADV, and supporting behavioral adjustment consistent with risk localization.Environment-aware eHMI changed perceived risk and behavior without consistently favorable objective safety outcomes, while auditory warning showed limited behavioral or objective safety improvement.
- Implications: Pedestrian-oriented AT interventions should not only alert pedestrians to danger but also help them locate the risk source.This conclusion follows the observed occlusion-induced risk and the projected eHMI’s behavioral adjustment consistent with risk localization.
- Limitations: The findings are limited by VR’s incomplete realism, a mainly young university-student sample, and one unsignalized two-lane crossing scenario.Future validation should examine more diverse environments, participant groups, vehicle configurations, speeds, and combined interventions.
Authorship contribution statement
The authors’ contributions covered conceptualization, methodology, analysis, investigation, validation, funding, supervision, visualization, data curation, and writing.
- The contribution statement assigns research design, data, analysis, investigation, validation, administration, funding, supervision, visualization, and writing roles across the author team.Specific roles vary by author and include both original drafting and review/editing.
Ethical statement
The study protocol received ethical approval from the Science and Technology Ethics Committee of Tongji University.
- Ethical approval was granted before the study began by Tongji University’s Science and Technology Ethics Committee.The approval number was tjdxsr2024041.
Appendix A1. Virtual reality sickness questionnaire (VRSQ)
The VRSQ appendix lists symptoms used to assess virtual reality sickness, including oculomotor symptoms, general discomfort, fatigue, eyestrain, headache, dizziness, and vertigo.
- The VRSQ lists oculomotor and general-discomfort symptoms such as fatigue, eyestrain, difficulty focusing, headache, blurred vision, dizziness, and vertigo.These symptoms are presented as questionnaire items in Appendix A1.
Appendix A2. Condition-wise descriptive statistics for Stage 1 risk formation component
The appendix organizes Stage 1 risk-formation descriptions by weather, near-side vehicle type, and far-side yielding strategy.
- Weather is listed as a Stage 1 condition-wise factor.
- Near-side vehicle type is listed as a Stage 1 condition-wise factor.
- Far-side strategy is listed alongside DST, ITADV, and Lane 1 risk perception.
Appendix A3. Condition-wise descriptive statistics for Stage 2 intervention-effect component
The appendix reports model coefficients for autonomous-truck, weather, strategy, and entry-time predictors, alongside intercepts and model-fit statistics.
- Autonomous truck had coefficient −1.494 for risk perception in one reported model.The coefficient had SE 0.157, z-value −9.51, p-value <0.001, and 95% CI [−1.802, −1.186].
- The reported models include intercept estimates of 6.465 and 6.604.The intercepts had p-values 0.010 and 0.001, respectively.
- Autonomous truck had coefficient 1.004 in the mediator model for entry T.The coefficient had SE 0.134, z-value 7.51, p-value <0.001, and 95% CI [0.742, 1.265].
- Rainy weather had coefficient 0.272 in the mediator model, with p-value 0.042.Its 95% CI was [0.010, 0.534].
- Autonomous truck had coefficient −1.089, while second-lane entry time had coefficient −0.403, in another reported model.Both coefficients had p-values <0.001; the corresponding 95% CIs were [−1.387, −0.792] and [−0.484, −0.323].
- The appendix reports total-effect, mediator, and direct-effect model-fit statistics.The reported ITADV values were 0.475 for the total effect and 0.183 for the direct effect; the mediator entry T value was 2.229.