A Domain-Specific Language for Prototyping the Behavior of a Humanoid Robot that Allows the Inclusion of Sensor Data.

We have been developing software for training tasks for patients after stroke within the project E-BRAiN (Evidence-Based Robotic Assistance in Neurorehabilitation). It is the idea of the project that a humanoid robot is instructing and observing the performance of patients. There are four different therapies that are supported by our software. These are the Mirror Therapy, Arm Basis Training, Arm Ability Training and the Neglect Therapy. The paper characterizes the different therapies and discusses opportunities for support by human robot. It discusses the challenges of observing training tasks and providing hints to improve the performance. A Domain-Specific Language was designed to model the behavior of participants in a task-based way. It contains predefined specific robot tasks (commands) like say text, show image, play video, raise arm, turn around etc. There exists an interpreter of the DSL. It sends messages to a robot if a command is interpreted. Additionally, the language allows the specification of preconditions of training tasks. This language construct can be used for expressing activities that are related to delivered values of sensors.