systematic behavioural analysis of pedestrian and ... · systematic behavioural analysis of...

Systematic Behavioural Analysis of Pedestrianand Autonomous Vehicle Interactions using

Virtual Immersive Reality Environment

Arash Kalatian, Bilal FarooqRyerson University

Presented at IATBR 2018, Santa Barbara

July 17, 2018

Kalatian and Farooq IATBR 2018 July 17, 2018 1 / 30

Table of Contents

1 Motivation and background2 Methodology3 Data Collection4 Results and Analysis5 Summary and Future Work

Kalatian and Farooq IATBR 2018 July 17, 2018 2 / 30

Table of Contents

1 Motivation and background2 Methodology3 Data Collection4 Results and Analysis5 Summary and Future Work

Kalatian and Farooq IATBR 2018 July 17, 2018 3 / 30

Problem StatementAVs are expected to be on the streets in near future

I Potential to fundamentally altering urban travel patterns andbehaviours

I Urban areas need to be prepared to benefit from these changes

Bloomberg Philanthropies Survey:

“We asked a subset of 30 cities about the barri-ers to moving forward on AV pilots, policy, andplans. These cities told us they are struggling tofind the human and financial capacity to delivermore projects, and the right actions are not yetclear or urgent enough. ”

Kalatian and Farooq IATBR 2018 July 17, 2018 4 / 30

Solution

To rethink conventional urban policies, designs and plans

Some possible changes (NACTO)

Reducing streets’ lane width

Increasing public spaces

Sidewalk and bike lane width

Restrictions on roads based on real-time demand

Managing traffic gaps

Setting lower speed limits

Kalatian and Farooq IATBR 2018 July 17, 2018 5 / 30

Project Overview

Objective: To explore the features of future streets and AV trafficcharacteristics that influence pedestrian movement

Approach: Quantifying the effects of specific traffic parameters andstreets’ geometric designs on pedestrians’ walking behaviour indifferent conditions

Method: Design an experiment in Virtual Immersive RealityEnvironment

Kalatian and Farooq IATBR 2018 July 17, 2018 6 / 30

Table of Contents

1 Motivation and background2 Methodology3 Data Collection4 Results and Analysis5 Summary and Future Work

Kalatian and Farooq IATBR 2018 July 17, 2018 7 / 30

Virtual Immersive Reality Environment

VIRE (Farooq, Cherchi, Sobhani, 2018)

3D scenarios are created based on theoretical experiment design

Traffic movement is represented using an agent based simulation

Positions and interactions of dynamic objects are projected onto HeadMounted Display

Immersive and responsive environment

Responses from users are recorded

Kalatian and Farooq IATBR 2018 July 17, 2018 8 / 30

Participants using VIRE

Kalatian and Farooq IATBR 2018 July 17, 2018 9 / 30

Street views in VIRE

Kalatian and Farooq IATBR 2018 July 17, 2018 10 / 30

Scenarios

Each scenario is defined as a combination of designed multi-levelfactors (independent variables)

Changes in speed limit, lane width, no. of braking levels andMinimum Gap, are limited to scenarios with 100% AVs

*Current standards for residential urban streets in Canada

Kalatian and Farooq IATBR 2018 July 17, 2018 11 / 30

Dependant Variable

Automatically Measured

Pedestrian comfort: required waiting time before crossingCox Proportional Hazard Model

Stress level: Galvanic Skin Response (GSR)Binary Logit Model

Pedestrian safety: Post Encroachment Time (PET)Binary Logit Model

Kalatian and Farooq IATBR 2018 July 17, 2018 12 / 30

Experiment Design

Testing all combinations of factor levels is impossible

A subset of all combinations should be selected

D-Optimal Design: attribute level combinations are sought to be assuch to minimize parameter covariances in the model

Prior information on model parameters required: demo design

D-Optimal design for:I Cox Proportional Hazard Model with three covariates Schmidt

and Schwabe (2017)I 9 covariates are involved in our design

Kalatian and Farooq IATBR 2018 July 17, 2018 13 / 30

Table of Contents

1 Motivation and background2 Methodology3 Data Collection4 Results and Analysis5 Summary and Future Work

Kalatian and Farooq IATBR 2018 July 17, 2018 14 / 30

Experiment Setup

D-Optimal Designs are developed for each dependent variable

100 scenarios are selected based on their overlap in different designsand weights.

Participants are asked to fill a questionnaire on their:I sociodemographic informationI walking habitsI previous VR experienceI health condition

Number of trials for each participant:

Adults 15 scenarios, 2 times each

Adolescents 15 scenarios, no repetitions

Kalatian and Farooq IATBR 2018 July 17, 2018 15 / 30

A Sample of Scenarios

A sample of experimented scenarios

Kalatian and Farooq IATBR 2018 July 17, 2018 16 / 30

Stress Level sensorsGSR Sensors

I Skin resistance between two reusable electrodes attached to two fingersof one hand.

I In response to internal and external stimuli, sweat glands become moreactive, increasing skin conductance and thus decreasing skin resistance.

PPG to HRI capture a PPG signal and convert it to estimate heart rate (HR),I Using the PPG ear clip

Shimmer GSR+ Sensor

Kalatian and Farooq IATBR 2018 July 17, 2018 17 / 30

Table of Contents

1 Motivation and background2 Methodology3 Data Collection4 Results and Analysis5 Summary and Future Work

Kalatian and Farooq IATBR 2018 July 17, 2018 18 / 30

Sociodemographic Characteristics

Campaign: June 15 – August 30

102 participants so farI Adults: 66I Adolescents: 36I 57% used virtual reality before

Kalatian and Farooq IATBR 2018 July 17, 2018 19 / 30

Adults Sociodemographic

Kalatian and Farooq IATBR 2018 July 17, 2018 20 / 30

Transporation Information

Kalatian and Farooq IATBR 2018 July 17, 2018 21 / 30

Compare Means-Sociodemographic

Kalatian and Farooq IATBR 2018 July 17, 2018 22 / 30

Compare Means-Transportation Information

Kalatian and Farooq IATBR 2018 July 17, 2018 23 / 30

Compare Means (Experiment Factors)

Kalatian and Farooq IATBR 2018 July 17, 2018 24 / 30

Compare Means (Experiment Factors) (continued)

Kalatian and Farooq IATBR 2018 July 17, 2018 25 / 30

Preliminary Hazard Model for Waiting time

Negative Coefficient means longer waiting times

Kalatian and Farooq IATBR 2018 July 17, 2018 26 / 30

Table of Contents

1 Motivation and background2 Methodology3 Data Collection4 Results and Analysis5 Summary and Future Work

Kalatian and Farooq IATBR 2018 July 17, 2018 27 / 30

Summary

Virtual Immersive Reality Environment is used to capture pedestrians’reactions to different traffic, road geometric design and environmentalconditions

Modeling results for Cox Proportional Hazard model showed thatfactors such as arrival rate, gender, level of education, walking towork and main mode of transportation have significant impact onwaiting time

Kalatian and Farooq IATBR 2018 July 17, 2018 28 / 30

Future Work

Model safety measure (PET) and stress level (GSR)

Expanding the data set to cover different areas in the city

Expanding the data set to elderly and disabled participants

Kalatian and Farooq IATBR 2018 July 17, 2018 29 / 30

Acknowledgment

We would like to thankRafael Vasquez, Grace Yip, Abenaya Selvakumaran

The Canada Research Chairs programRyerson University

City of TorontoCity of Markham

Angus Glenn Library

Kalatian and Farooq IATBR 2018 July 17, 2018 30 / 30