Quantcast

SocRob@Home

Research paper by Pedro U. Lima, Carlos Azevedo, Emilia Brzozowska, João Cartucho, Tiago J. Dias, João Gonçalves, Mithun Kinarullathil, Guilherme Lawless, Oscar Lima, Rute Luz, Pedro Miraldo, Enrico Piazza, Miguel Silva, Tiago Veiga, Rodrigo Ventura

Indexed on: 26 Nov '20Published on: 16 Sep '19Published in: KI - Künstliche Intelligenz



Abstract

This paper describes the SocRob@Home robot system, consisting of a mobile robot (MBOT) equipped with several sensors and actuators, including a manipulator arm, and several software modules that provide the skills and capability to perform domestic tasks while interacting with humans in a domestic environment. We describe the whole system holistically, explaining how it integrates the contributing modules, and then we focus on the most relevant sub-systems, pointing out the original contributions of our research and development on the system in the last 5 years. The robot system includes metric and semantic mapping, several navigation modes (way-point navigation, person following and multi-sensor obstacle detection and avoidance), vision-based object detection, recognition, servoing and grasping, speech understanding, task planning and task execution. The robot system is mostly activated by speech commands from a human, and these commands, after being interpreted, are executed by the robot sub-systems, coordinated by a task executor. Lessons learned during the development and use of this system, which are useful as guidelines for the development of similar robot systems, are provided. MBOT’s performance is assessed using the task benchmarks scoring system of the European Robotics League competitions on Consumer Service robots.