Natural language interface for programming sensory-enabled scenarios for human-robot interaction

Abstract

Previous research has shown that robot-mediated therapy may be effective in improving different mental or physical conditions, but this effectiveness strongly depends on how well the therapy can be translated to robot training. The goal of this study is to assist the end-users such as occupational and rehabilitation therapists to create without help of technical professional therapy-specific and sensory-enabled scenarios for the robotic assistant for use in an unstructured environment. The Cognitive Dimension of Notations framework was applied to assess the usability of the programming interface and the Cyclomatic complexity method was used to evaluate the complexity of the created robot scenarios. Eleven therapists with a mean age of 39 years working in the care for persons with visual-and-intellectual disabilities participated. The results show good usability of the interface, as measured via the CDN framework and the cyclomatic complexity analysis showed an increased complexity of the created by the occupational and rehabilitation therapist's scenarios. The participants did not request for very specifically defined behaviors for the robot, and therefore descriptions in natural text can be successfully used for robot programming.