Robotics Research Lab
CRES
USC Computer Science
USC Engineering
USC
/ Research / Projects / The Role of Robot Personality for Post-Stroke Physical Therapy

Overview Details Images
Videos Publications Support Contact Details

Overview

This research describes a hands-off socially assistive therapist robot that monitors, assists, encourages, and socially interacts with post-stroke users engaged in rehabilitation exercises. We investigate the role of the robot's personality in the hands-off therapy process, focusing on the relationship between the level of extroversion-introversion of the robot and the user. We also demonstrate a behavior adaptation system capable of adjusting its social interaction parameters (e.g., interaction distances/proxemics, speed, and vocal content) toward customized post-stroke rehabilitation therapy based on the user's personality traits and task performance.

Project Details

Achieving a psychological "common ground" between the human user and the robot is necessary for a natural, nuanced, and engaging interaction. Therefore, in this work we investigated the role of the robot's personality in the assistive therapy process. We focus on the relationship between the level of extroversion-introversion (as defined in the Eysenck Model of Personality) of the robot and the user, addressing the following research questions:
1) Is there a relationship between the extroversion-introversion personality spectrum (assessed with the Eysenck model) and the challenging vs. nurturing style of patient encouragement?
2) How should the behavior and encouragement of the therapist robot adapt as a function of the user's personality and task performance?

In our interaction design, we chose to use the following traits: 1) sociability and 2) activity. These traits can be most readily emulated in robot behavior. We expressed those traits through three main parameters that define the therapist robot behavior: 1) interaction distance / proxemics, 2) speed, and 3) verbal and para-verbal communication.

The main goal of our robot behavior adaptation system is to enable us to optimize on the fly the three main parameters (interaction distance/proxemics, speed, and vocal content) that define the behavior (and thus personality) of the therapist robot, so as to adapt it to the user's personality and thus improve the user's task performance. Task performance is measured as the number of exercises performed in a given period of time; the learning system changes the robot's personality, expressed through the robot's behavior, in an attempt to maximize the task performance metric. We formulated the problem as policy gradient reinforcement learning (PGRL) and developed a learning algorithm that consists of the following steps: (a) parametrization of the behavior; (b) approximation of the gradient of the reward function in the parameter space; and (c) movement towards a local optimum.

Personality-matching experimental results: The experimental studies validated our two hypotheses. The participants with extroverted personalities had a preference for a robot that challenged them during exercises over the one that focused the interaction on praise. Analogously, users with introverted personalities preferred the robot that focuses on nurturing praise rather than on challenge-based motivation during the training program. Thus, the results show user preference for human-robot personality matching in the socially assistive context.

Robot behavior adaptation experimental results: The experimental results provide first evidence for the effectiveness of robot behavior adaptation to user personality and performance.

Images

Example of Experiment

Nate and the robot
Test Subject Interacting with the Robot
Test-beds
Robot
The non-biomimetic robot

Publications
Support

This work is supported by USC Women in Science and Engineering (WiSE) Postdoctoral Fellowship, the Okawa Foundation, the National Science Foundation grants titled "HRI: Personalized Assistive Human-Robot Interaction: Validation in Socially-Assisted Post-Stroke Rehabilitation" - Award Number 0713697.

Contact

Dr. Adriana Tapus: tapus at usc dot edu or adriana dot tapus at ieee dot org