Autonomous Driving: Framework for Pedestrian Intention Estimationin a Real World Scenario

TL;DR

This paper presents a framework for autonomous vehicles to estimate pedestrian crossing intentions in real-world scenarios, enhancing interaction safety without eye contact, validated through field tests.

Contribution

It introduces a novel pedestrian intention estimation framework specifically designed for autonomous vehicles in shared spaces, tested in real vehicle deployments.

Findings

Feasibility demonstrated through real-world vehicle tests

Effective communication messages improve pedestrian interaction

Framework reduces uncertainty in pedestrian crossing behavior

Abstract

Rapid advancements in driver-assistance technology will lead to the integration of fully autonomous vehicles on our roads that will interact with other road users. To address the problem that driverless vehicles make interaction through eye contact impossible, we describe a framework for estimating the crossing intentions of pedestrians in order to reduce the uncertainty that the lack of eye contact between road users creates. The framework was deployed in a real vehicle and tested with three experimental cases that showed a variety of communication messages to pedestrians in a shared space scenario. Results from the performed field tests showed the feasibility of the presented approach.