Paper Image

Predicting person's intention to interact with a service robot

Published on:

2 April 2024

Primary Category:


Paper Authors:

Simone Arreghini,

Gabriele Abbate,

Alessandro Giusti,

Antonio Paolillo


Key Details

Presents method for robot to predict person's intention to interact early

Classifier uses sequence of person's pose, gaze, facial landmarks

Adding gaze cues boosts accuracy and detection distance considerably

Approach adapts to new environments in self-supervised manner

Shows applications for waiter robot in real qualitative experiments

AI generated summary

Predicting person's intention to interact with a service robot

This paper presents a method for a service robot to perceive early when a nearby person intends to interact, so it can proactively enact friendly behaviors for better user experience. A sequence-to-sequence classifier predicts interaction intention based on the person's pose, gaze, and facial landmarks. Experiments on a novel dataset show including gaze cues significantly improves performance: accurate classification is achieved from 3.2 meters away, up from 2.4 meters without gaze. Additional experiments quantify the approach's ability to adapt to new environments without supervision, and demonstrate applications with a waiter robot.

Answers from this paper


No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up