AR Accident Simulation 

Accident Simulator is an AR product designing at Phoenix On the Ground(POG) to prevent traffic accidents by simulating frequently happening accidents between cars and pedestrians. The application works based on deep learning data that perceive lanes.

I am working as an AR programmer and product designer to build a safe city together with a deep learning engineer and a project manager. We are continuously developing our content through user testing and researching the effectiveness of the app in decreasing future traffic accidents.

PROJECT Timeline

2019 – Current

Category

Augmented Reality
User Research

Responsibilities

AR Programming
Interface Design
Qualitative Research
3D Character Modeling

Tools

Unity – Vuforia
Fusion 360
Blender

Research Glimps

Design Background

23.5 Bilion Dollars

Social costs spent from traffic accidents is
23.8 billion dollars in 2018, South Korea.
It is the equivalent amount of operating profits of South Korea’s four largest companies. 

Countermeasures Focused On Death Casualties

56.5 percent of traffic accidents are due to negligence in keeping eyes forward.
However, the prevention of accidents is more focusing on the cause of death casualties.

Fail to Convince Likelihood of Accidents to Each Individual

Current education on accident prevention
is not enough to convince that
it can happen to anyone.

Design Approach

“Planting Memory of Traffic Accident”

An AR experience
that brings other’s traffic accident
to an individual level

An AR traffic accident simulating content will allow the individual to situate oneself at the accident-prone location and observe accidents with their own eyes.

Data-driven accident simulation
that allows individuals to remind
similar accident at similar location

Ten cases of frequently happening accidents analyzed from traffic accidents will remind pedestrians
to keep their eyes forward at any location.

User Scenario

A. The user finds the POG marker on the street and scans the QR code to start the AR experience that fits the corresponding location

B. The user faces the mobile camera toward the lanes in order to start the deep learning process which sorts out lanes from other sources

C. The user learns about the accident frequently happening on the location through AR simulation

D. The user checks out the information on the injury and a written description of the cause of the accident

Pog Bot

I created a 3D character named Pog Bot – a robot substituting the human getting hit by the car and lessens the shock that may come from the accident simulation. The bone of the Pog Bot is rigged using Blender and the character is animated in Unity animator.

3D character designed in Fusion 360

Nodding and looking around motion applied in Unity

3D character designed in Fusion 360

Nodding and looking around motion applied in Unity

Applying Deep Learning

The product aims to provide a seamless user experience where the simulation contents are augmented at where it should be, especially vehicles. I worked with a deep learning engineer who trained a deep learning model to differentiate between lanes and not-lanes and converted the data into coordination usable at Unity. I was responsible for using those data and augment the simulation at the right location on the screen.

Scene to train

Trained result showing lane in white and non-lane in black

Change of car position after applying the trained result, moving from the leftmost to the rightmost corner of the lane

User Test

To observe how experience affects audiences, a usability test with 36 participants was conducted.