1 of 17

Modeling Social Situation Awareness in Driving Interactions

AutoUI '24

September 24th

Navit Klein*

Hauke Sandhaus^†

Dr. David Goedicke^†

Prof. Avi Parush*

Prof. Wendy Ju^†

Team

-

*

^†

2 of 17

Driving Encounters as �a Social Situation

Two or more drivers must be involved
These drivers must be influencing one another
There are potential space-time conflicts between drivers

“A situation in which the behavior of at least two road users can be interpreted as being influenced by the likelihood that they will occupy the same region of space at the same time in the near future" (Markkula et al., 2020)

1/12 Motivation

Social activity that involves

coordination and communication

between multiple drivers

3 of 17

2/12 Motivation

Non-linguistic�Implicit/Explicit Communication

Speed change
Eye contact
Car heading direction
Tailing a car

Implicit Communication

Hand gestures
Signaling lights
Using horn

Explicit Communication

4 of 17

Participants

3/12 Method

https://osf.io/bz8vq/

5 of 17

Multi-participant VR driving interaction
Enables quantitative and qualitative analysis

4/12 Method

Apparatus�StrangeLand Driving Simulator

Goedicke, D., Zolkov, C., Friedman, N., Wise, T., Parush, A., & Ju, W. (2022). Strangers in a Strange Land: New Experimental System for Understanding Driving Culture Using VR

Goedicke, D., Haraldsson, H., Klein, N. A., Zhou, L., Parush, A., & Ju, W. (2022). Rerun: Enabling Multi-Perspective Analysis of Driving Interaction in VR

6 of 17

13/30 Study

7 of 17

Ambiguous right-of-way

Driving�Scenarios

5/12 Method

Opposite direction approach

Side approach

Intersection scenarios with a potential for conflict

7

8 of 17

Developing�Social SA Questions

6/12 Method

1 Task definition - Cross the intersection

2 Identifying the critical situational elements

"One of the key contributions of our study is the Social Situational Awareness questionnaire that we developed.

The model of situational awareness, we used to understand the social driving encounters is by Mica Endsley.

This questionnaire is based on critical situational elements during driving interactions. We grouped these elements into the three levels of situational awareness:

Perception: This is the first level, where drivers perceive key elements of their surroundings—like the speed of another vehicle or whether it’s signaling. It's about what drivers notice at the moment of the encounter.
Comprehension: At the second level, drivers interpret the meaning of what they perceive. For example, understanding what the right of way is based on loation, or where another driver is headed.
Projection: The final level involves predicting what will happen next. Based on what drivers perceive and comprehend, they project the other vehicle's likely future behavior—like whether it will cross the intersection first.

Through this method, we could assess how well participants understood these interactions, helping us model their social situational awareness during driving encounters."

9 of 17

SA questions appear after each scenario
An image of the encounter scene relevant to the question displayed to contextualize the question

Social Situation Awareness (SA)�Questions

7/12 Method

While the other vehicle was approaching the intersection what was their speed?
As you approached the intersection, how did you interpret the other vehicle's speed change?
Based on the other vehicle's speed while approaching the intersection, what did you expect the other driver to do?
Who was the first one to enter the intersection?

10 of 17

Social Situation Awareness (SA)

Models

APPROACH MODEL

SIGNALING MODEL

HEADING MODEL

SPEED & SPEED CHANGE MODELS

8/12 Results

11 of 17

Social Situation Awareness (SA)

Models - Structure

9/12 Results

SA Score

39% correct

58% Incorrect

3% undefined

SAQ-1 - While the other vehicle was approaching the intersection what was their speed?

12 of 17

Social Situation Awareness (SA)

Models - Structure

9/12 Results

13 of 17

Social Situation Awareness (SA)

Models - Structure

9/12 Results

14 of 17

SPEED CHANGE is a strong, agreed-upon implicit cue for both participants

Speed & Speed Change

Model

10/12 Results

15 of 17

Social Situation Awareness (SA) Models

SIGNALING MODEL

11/12 Results

APPROACH MODEL

HEADING MODEL

16 of 17

Mutual Driver Awareness helps set expectations at un-signaled intersections.
Effectiveness of Explicit Communication Cues (Signals)

Effective for indicating turns, but less effective for signaling going straight.

Implicit Cues (Heading & Speed)

Suggest continuing straight more than turning, not highly effective alone for conveying intentions.

Role of Implicit Communication is crucial when explicit cues are unclear or absent.
Design Considerations: Autonomous Vehicles should integrate both explicit and implicit cues with an emphasis on clear heading and speed adjustments to convey intentions.

Findings From the Models and Implications

12/12 Implications

To summarize some of the key insights from our social situational awareness analysis: "They highlight the importance of mutual awareness and expectation setting at intersections.

Explicit cues, such as signals, were effective for turns, but less so for indicating going straight. Implicit cues like heading and speed change were critical for resolving ambiguous encounters, but were not effective alone at conveying intentions."

"As we move towards more autonomous vehicles, we need to incorporate both explicit and implicit communication cues into AV design. Clear implicit cues, like vehicle heading and speed adjustments, as well as explicit cues such as intent displays are essential to ensure effective interaction between human drivers and autonomous vehicles."

**Implications**

- **Mutual Awareness and Expectation Setting**

- Drivers are aware of other drivers at un-signaled intersections, which helps them formulate expectations about who will enter the intersection first.

- **Effectiveness of Communicative Cues**

- **Explicit Cues (Signals):**

- Effective for conveying intentions to turn right or left.

- Less effective for indicating the intention to continue straight.

- **Implicit Cues (Vehicle Heading and Speed):**

- Heading direction is more effective for suggesting a continuation straight ahead than for turning.

- Both signaling and heading are not highly effective alone for conveying directional intentions.

- **Role of Implicit Communication**

- Critical for negotiating and resolving ambiguous or vague encounters, especially when explicit communication is unclear or lacking.

- Drivers actively look for and interpret implicit cues under these circumstances.

- **Design Considerations for Autonomous Vehicles (AVs)**

- Integration of both explicit and implicit communicative cues is essential for effective interaction with human drivers.

- Special emphasis should be placed on designing clear implicit cues—such as vehicle heading and speed adjustments—to convey AVs' intentions, particularly in situations where explicit signals may be insufficient.

Drivers accurately pick up on the approach and direction of the other driver at the intersection.

This perception helped drivers form accurate predictions about who would cross the intersection first, largely based on right-of-way understanding.

Signaling was found quite efficient at communicating plans to turn either way.

Lack of signaling was understood as continuing straight however there were cases where participants signaled although they were instructed to continue straight.

We believe that this happened because of a mimic behavior between participants.

This is why we believe that this explicit communication is less effective.

Heading direction was an additional implicit cue that helped to understand the other driver's intention to continue straight.

17 of 17

APPROACH MODEL

SIGNALING MODEL

HEADING MODEL

SPEED & SPEED CHANGE MODELS

Summary

1. VR study at two locations

2. SSA Questionnaire matched with driving ground truth

3. Four Social Situational Awareness Models

Modeling Social Situation Awareness in Driving Interactions. Klein, Sandhaus, Goedicke, Ju, Parush. AutoUI ‘24