ITS Arizona 29^th Annual Conference … Technical Session 2B: Multi-Modal ITS Applications … 10.20.2022

Evaluating Sidewalk Autonomous Delivery Robot Interactions with

Pedestrians and Bicyclists on Shared-use Transportation Facilities

Steven R. Gehrke, Ph.D.

Assistant Professor, Northern Arizona University

• Sidewalk autonomous delivery robots (SADRs) were first deployed at NAU in March 2019.
• Zero-emission devices offer contactless meal deliveries across campus to students, staff, and faculty.
• Popularity spurred by information and communication technology (ICT) advancements as well as public health safety concerns brought on by the Covid-19 pandemic.
• Larger fleet sizes and more frequent meal deliveries have helped to meet growing demand but have also produced more opportunities for conflict with pedestrians and bicyclists who share off-street facilities.
• At present, there is a lack of city or state codes to regulate the safety feature requirements (e.g., braking systems, lights, size and weight limits) and operation (e.g., pedestrian yielding) of SADRs in public spaces.
• As sidewalk standards and curb management strategies evolve, the management of transportation facilities once designed for pedestrians and bicyclists must now account for possibility that SADRs will vie for room in public spaces.

​

• Study aims to offer novel insights into SADR operation in real-world setting.
• Generate evidence on the traffic safety of active travelers sharing paths with SADRs.
• Collect field-recordings of SADRs operating in mixed traffic settings and identify SADR interactions.
• Inform mitigation and facility management strategies that guide safe SADR operation.
• Spatial and statistical analysis of SADR interaction severity as a function of conflict and site attributes.

Research context

Introduction … Methodology … Results … Conclusions

2

• Passive video collection conducted at ten sites across NAU’s main campus in Sept. 2021.
• Cameras with external batteries were mounted to telescoping poles attached to stationary objects.
• Cameras placed at team-identified hot spots of high activity for SADRs, pedestrians, and bicyclists.
• For analysis, 179.5 hours of video were recorded during mealtimes (9-10:30a, 11a-2p, and 4:30-6p).

Data collection

Introduction … Methodology … Results … Conclusions

3

• Site exposure and conflict metrics were calculated via video review from two researchers.
• Adapted post-encroachment time (PET) metric identified SADR incidents with pedestrians and bicyclists.
• SADR-involved interactions categorized into three severity levels (dangerous, moderate, and no conflict).

Surrogate safety measurement

Introduction … Methodology … Results … Conclusions

4

SADR-involved interactions with pedestrians and bicyclists

Introduction … Methodology … Results … Conclusions

5

Example site and conflict characteristics

Introduction … Methodology … Results … Conclusions

6

• More severe conflicts when SADR reached conflict point first and took no evasive action.
• Higher share of severe interactions when SADR reached conflict point before a pedestrian.

Evasive maneuvers by pedestrians

Introduction … Methodology … Results … Conclusions

7

Statistical model results

Introduction … Methodology … Results … Conclusions

8

• SADR-involved interaction severity tended to increase if robot was first pathway user to reach conflict point.
• Severity of SADR interaction decreased if robot and human pathway user were traveling in opposite direction.
• Evasive action of first pathway user to reach conflict point was more likely to be a swerve in travel direction.

• Study findings and implications for practice.
• Study identified most common types and patterns of SADR interactions with pedestrians and bicyclists.
• To reduce prevalence of crossing conflicts, program SADR routes to prioritize parallel travel along pedestrian corridors.
• SADR routing decisions must factor in sidewalk width to ensure adequate space available for SADRs to overtake pedestrians.
• SADR-bicyclist interactions common where bike network gaps exist and bike parking is located.
• SADR routes should place device on side of path that produces fewer turning movements when bike lanes are not available.
• Position SADR delivery points away from main building entrances, especially when near bike racks or bike parking facilities.
• Future changes to the design of SADR devices and introduction of warning signs should be explored.
• Audible cues when traveling through high-activity zones and alterations to vehicle profiles and lighting to improve visibility.
• Warning signs can be introduced at high-activity zones, but care must be given to placing further hardships on active travelers.

​

• Study limitations and directions for future research.
• Landscape and travelers on a college campus do not represent urban contexts or general population.
• Sample had fewer dangerous conflicts than other two interaction categories, limiting statistical power.
• Regarding PET metrics, some pedestrians and bicyclists may have greater comfort traveling near SADRs.

Discussion and conclusions

Introduction … Methodology … Results … Conclusions

9

• College students more likely to adopt sustainable travel modes and introduction of SADRs to shared-use paths designed for pedestrians and bicyclists creates new safety concerns.

​

• Second study aims to offer subjective insights into SADR operation in real-world setting.
• Describe experiences with and future intentions to use autonomous food delivery services of a younger, university population who is more likely to be early adopters of new last-mile delivery technologies.
• Identify personal attributes and SADR-related experiences or future adoption intentions associated with self-reported comfort in sharing pathways with SADRS from the perspective of a pedestrian or bicyclist.

Study 2. Reported Pedestrian and Bicyclist Comfort in Sharing Pathways with Sidewalk Autonomous Delivery Robots

Research context

Introduction … Methodology … Results … Conclusions

10

• Perceptions of Autonomous Robots for Deliveries survey administered in April 2022.
• Tablet-based survey instrument designed to identify perceptions of and experiences with SADRs.
• Section 1. Survey participant information: Socioeconomic background and general travel behaviors.
• Section 2. Autonomous delivery vehicles: Experiences and perceptions of SADR interactions and current/future adoption.
• Section 3. Stated choice experiment: Reported comfort after viewing four five-second field-recorded videos of SADR conflicts.

​

​

​

​

​

​

​

​

• In all, 522 survey responses collected over two-week administration period at two NAU student unions.
• Three quarters of survey respondents resided on campus, with nearly one half (48%) of respondents of freshman standing.
• Three quarters of survey respondents had adopted SADR delivery services in past, with one half doing so monthly or more.

Study 2. Reported Pedestrian and Bicyclist Comfort in Sharing Pathways with Sidewalk Autonomous Delivery Robots

Data collection

Introduction … Methodology … Results … Conclusions

11

Study 2. Reported Pedestrian and Bicyclist Comfort in Sharing Pathways with Sidewalk Autonomous Delivery Robots

Reported comfort by pedestrians and bicyclists

Introduction … Methodology … Results … Conclusions

12

• Additional study findings and implications for practice.
• Pedestrians tended to be more comfortable in sharing pathways with SADRs than bicyclists.
• However, model results revealed respondents who stated being uncomfortable in sharing paths with SADRs as a pedestrian or having altered their intended path in the past due to an SADR were less comfortable take evasive actions to avoid a conflict.
• Respondents who more frequently used autonomous food delivery services in the past tended to be more comfortable in taking evasive actions to avoid potentially more-severe interactions with SADRs as either a pedestrian or bicyclist.
• Respondents who bicycle around campus were less comfortable having to swerve to avoid a potential SADR-involved collision than respondents who drive around NAU’s campus, highlighting importance of mode-specific experience in SADR perceptions.
• Greater attention needed to inform policies and interventions regarding SADR expansion to new settings.
• Sidewalk management and SADR route programs must ensure new technology does not impede safe, efficient active travel.
• Demonstration projects rather than untested full-scale deployment of SADRs should be pursued to identify safety concerns.

Study 2. Reported Pedestrian and Bicyclist Comfort in Sharing Pathways with Sidewalk Autonomous Delivery Robots

Discussion and conclusions

Introduction … Methodology … Results … Conclusions

13

Evaluation of Sidewalk Delivery Robot Interactions with Pedestrians and Bicyclists

https://www.metrans.org/research/evaluation-of-sidewalk-delivery-robot-interactions-with-pedestrians-and-bicyclists

Steven R. Gehrke, Ph.D. (steven.gehrke@nau.edu)