KURE: A Two-Way Adaptive System for Intuitive Robot Control

This paper presents KURE as a two-way adaptive system that facilitates intuitive robot control through mutual adaptation between user and system. Unlike traditional unidirectional control interfaces where users must adapt to specific paradigms and devices, KURE implements a bidirectional learning mechanism that allows both the human operator and the robotic system to adapt to each other over time, resulting in increasingly natural interactions. The system utilises principles from self-organisation and machine learning to create a control interface that evolves with use, becoming personalised to individual operator preferences and control patterns. Starting from a tabula rasa basis, the architecture identifies control patterns (behaviours) for the given robotic morphology and successfully merges them with user control signals, regardless of the input device used. This approach enables a transparent connection between user and robot, allowing users to shape control motifs according to their preferences without requiring operation manuals or specific device training. The two-way adaptive nature of KURE represents a paradigm shift in human-robot interaction, introducing a new level in the taxonomy of human-in-the-loop systems.

IEEE RO-MAN Workshop on Behavior Adaptation, Interaction and Learning for Assistive Robotics