Hands-Off Therapist Robot Behavior Adaptation to User Personality for Post-Stroke Rehabilitation Therapy Adriana T ¸˘ apus ¸, Member, IEEE, Cristian T ¸˘ apus ¸, Member, IEEE, and Maja J Matari´ c, Senior Member, IEEE Abstract— This paper describes a hands-off therapist robot that monitors, assists, encourages, and socially interacts with post-stroke users in the process of rehabilitation exercises. We developed a behavior adaptation system that takes advantage of the users introversion-extroversion personality trait and the number of exercises performed in order to adjust its social interaction parameters (e.g., interaction distances/proxemics, speed, and vocal content) toward a customized post-stroke rehabilitation therapy. The experimental results demonstrate the robot’s autonomous behavior adaptation to the user’s personality and the resulting user improvements of the exercise task performance. Index Terms— Rehabilitation Robotics, Socially Assistive Robotics, Social Human-Robot Interaction, Learning and Adaptive Systems I. I NTRODUCTION T HE recent trend toward developing a new generation of robots that are capable of moving and acting in human-centered environments, interacting with people, and participating and helping us in our daily lives has introduced the need for building robotic systems able to learn how to use their bodies to communicate and react to their users in a social and engaging way. Social robots that interact with humans have thus become an important focus of robotics research. Nevertheless, Human-Robot Interaction (HRI) for socially assistive applications is still in its infancy. Socially assistive robotics, which focuses on the social interaction, rather than the physical interaction between the robot and the human user has the potential to enhance the quality of life for large populations of users, such as the elderly [30], people with physical impairments and in rehabilitation therapy (e.g., post- stroke patients) [7], [28], people with cognitive disabilities and social and developmental disorders (e.g., children with autism, children with attention deficit/hyperactivity disorder (AD/HD)) [23], [24], [25]. In our work, the target user population is post-stroke patients. Stroke is the leading cause of serious, long-term Manuscript received September 15, 2006; This work was supported by USC Women in Science and Engineering (WiSE) Program and the Okawa Foundation. Dr. Adriana T ¸˘ apus ¸ is with the Robotics Research Lab/Interaction Lab, Department of Computer Science, University of Southern California, Los Angeles, USA (phone: +1 (213) 740 6245: fax: +1 (213) 821 5696; e-mail: tapus@robotics.usc.edu) Dr. Cristian T ¸˘ apus ¸ is with the Mojave Research Lab, Department of Computer Science, California Institute of Technology (Caltech), Pasadena, USA (e-mail: crt@caltech.edu) Prof. Maja J. Matari´ c is with the Robotics Research Lab/Interaction Lab, Department of Computer Science, University of Southern California, Los Angeles, USA (e-mail: mataric@usc.edu) disability among American adults, with over 750,000 people suffering a new stroke each year [19]. Stroke patients are unable to perform movements with the affected limb, even though the limb is not completely paralyzed. This loss of function, termed learned disuse, can improve with rehabil- itation therapy during the critical post-stroke period. The best strategy of any post-stroke rehabilitation program is the repetitive practice of exercises, which can be passive and active. In the passive exercises (also knows as hands- on rehabilitation), the patient is helped by the human (or robot) therapist to move the affected limb, while in the active exercises, the patient performs the exercises with no physical hands-on assistance. The vast majority of existing work into rehabilitation robotics focuses on hands-on robotic systems (e.g., [4], [5], [22]). However, recent results from physical therapy research show that such therapy may not be the most effective means of recovery from stroke, and is certainly not the only necessary type of much-needed treatment [7]. Our work focuses on hands-off therapist robots that as- sist, encourage, and socially interact with patients during their active exercises. We previously demonstrated [7], [11], [12], [27], [28], through real-world experiments with stroke patients, that the physical embodiment (including shared physical context and physical movement of the robot), encouragement, and monitoring play key roles in patient compliance with rehabilitation exercises. Recently, we also investigated the role of the robot’s personality in the hands- off therapy process, by focusing on the relationship between the level of extroversion/introversion (as defined in Eysenck Model of personality [10]) of the robot and the user [28]. Building robotic systems capable of adapting their behav- ior to user personality, user preferences, and user profile so as to provide an engaging and motivating customized protocol is a very difficult task, especially when working with vulnerable users. Different learning systems for human-robot interaction have been proposed in the literature [3], [20], but none of them includes the human profile, preferences, and/or personality in the model. To the best of our knowledge, no work has yet tackled the issue of robot behavior adaptation as a function of user personality in the assistive human-robot interaction context. In our work, we address this issue and propose a behavior adaptation system based on reinforcement learning. The robot incrementally adapts its behavior as a function of the users introversion-extroversion level and of the amount of exercises he/she has performed, aiming toward a more individualized and appropriately challeng- ing/nurturing therapy style that will help to improve user task performance. Our robot behavior adaptation system monitors 2007 IEEE International Conference on Robotics and Automation Roma, Italy, 10-14 April 2007 WeE8.4 1-4244-0602-1/07/$20.00 ©2007 IEEE. 1547