Konstantinos Tsiakas     Ph.D. Candidate     HERACLEIA Human-Centered Computing Laboratory     Computer Science and Engineering Department     University of Texas at Arlington     Intitute of Informatics and Telecommunications, NCSR Demokritos     Office: ERB 306, CSE@UTA     Office Hours: TuTh 12:30pm - 2:00pm     Phone: 817-805-9043     Email: konstantinos.tsiakas@mavs.uta.edu     LinkedIn Google Scholar ResearchGate |
During the task, the robot announces a sequence of a letters: "A", "B" and "C", of a give length L. The difficulty of each round task is proportional to the sequence length L = [3, 5, 7, 9]. The user has to press the corresponding buttons in the correct order, as fast as possible. The system stores information about the task difficulty, user performance, robot feedback, etc., in order to examine patterns in interaction data, towards the definition of a personalized SAR system for sequence learning task, which will adjust the task parameters (e.g., difficulty) and the robot behavior (positive, negative or no feedback) to maximize user engagement and thus, training effects.
As a first step, we provide a dataset, as an outcome of the data collection, along with a set of data analysis, including machine learning, data mining and statistical analysis in order to get insight towards the definition of an adaptive SAR system using Interactive Reinforcement Learning methods for real-time robot adaptation, applying our proposed Interactive Learning and Adaptation framework.. More details, in the published Social Robotics paper (ICSR '16) and in the accepted Human Robot Interaction paper (HRI '17). |