A Study in Zucker: Insights on Interactions Between Humans and Small Service Robots

TL;DR

This paper introduces a new dataset and analysis of human interactions with small indoor service robots, revealing insights into safety, behavior, and the applicability of existing models in HRI contexts.

Contribution

The study provides a novel dataset on human-robot interactions with small differential drive robots and evaluates the effectiveness of current trajectory prediction models in indoor HRI settings.

Findings

Anticipatory and non-anticipatory controllers impose similar safety constraints.

State-of-the-art trajectory prediction models extend well to indoor HRI.

Humans experience less social discomfort interacting with small robots than with other humans.

Abstract

Despite recent advancements in human-robot interaction (HRI), there is still limited knowledge about how humans interact and behave in the presence of small service indoor robots and, subsequently, about the human-centered behavior of such robots. This also raises concerns about the applicability of current trajectory prediction methods to indoor HRI settings as well as the accuracy of existing crowd simulation models in shared environments. To address these issues, we introduce a new HRI dataset focusing on interactions between humans and small differential drive robots running different types of controllers. Our analysis shows that anticipatory and non-anticipatory robot controllers impose similar constraints to humans' safety and efficiency. Additionally, we found that current state-of-the-art models for human trajectory prediction can adequately extend to indoor HRI settings. Finally, we show that humans respond differently to small differential drives than to other humans when collisions are imminent, since interacting with small robots can only cause a finite level of social discomfort as compared to human-human interactions. Our dataset and related code and models are available at: https://motion-lab.github.io/ZuckerDataset.