Enabling Customer-Driven Learning and Customisation Processes for ML-Based Domestic Robots

Abstract

Smart domestic robots are poised to revolutionise the way household chores and everyday tasks are
carried out in the home of the future. At the heart of the “intelligence” and behaviour of these robots
will be complex machine learning (ML) systems that, in addition to extensive training at the manufacturing stage, will most likely require further on-site adjustments to adapt to customers and their
environments. Drawing from the robotics literature on Learning from Demonstration and Human Robot Interaction, we review relevant techniques which we hypothesise customers could realistically
use to perform these adaptation and customisation steps as smoothly and effortlessly as possible.