submitted at CASE22

Gabriele Abbate, Alessandro Giusti, Antonio Paolillo, Luca Maria Gambardella, Andrea Emilio Rizzoli and Jérôme Guzzi

Dalle Molle Institute for Artificial Intelligence, USI-SUPSI, Lugano (Switzerland)

This work is supported by the European Commission through the Horizon 2020 project 1-SWARM, grant ID 871743.

We introduce an intuitive pointing-based interface to select objects moving on a system of conveyor belts. The interface has minimal sensing requirements, as the operator only needs to wear an Inertial Measurement Unit on the wrist (e.g., a smartwatch). LED strips provide the required visual feedback to precisely point to the objects and select them. We experimentally compare the proposed approach with a baseline mouse-based graphical user interface in which the user can click on packages with a mouse. Quantitative results show that our interface compares favorably to the baseline, especially in difficult scenarios involving many packages moving fast.

This repository is organized as follows:

• submission: contains the video submission, supplementary video materials, experimental datasets and code to analyze them.

The video shows the user performing the relative localization and the pointing phase, highlighting the overlap between the VR and the real world scenarios.

The video shows a couple of runs where the user is interacting with the conveyor belt system. Once the user notices a red (anomalous) package, he triggers the relative localization and then points at that package. Selected packages are diverted towards the first unloading bay, unselected packages towards the third one.