Christian Martens

The Rehabilitation Robots FRIEND-I & II: Daily Life Independency through Semi-Autonomous Task-Execution

The rehabilitation robotic systems FRIEND1-I (Martens et al., 2001) and its successor FRIEND-II have been developing at the Institute of Automation (IAT), University of Bremen, Germany, since 1997 and 2003 respectively. The systems belong to the category “intelligent wheelchair mounted manipulators”. They focus on users with high spinal cord injury, or with similar handicaps, who are unable to control the manipulator by means of a keyboard or joystick. The systems offer support during daily life activities and at professional life. The strategic objective of the FRIEND as well as the succeeding AMaRob2 project, which focuses on the usage of FRIEND-II within the context of an intelligent environment, is to research into new methods to control the robotic system in such a way that their users become independent for at least 1.5 hours without support by nursing staff. Beside the aspect that this is one of the main requirements expressed by potential users, the fulfillment of this objective would have a strong impact on the commercialization of the rehabilitation robotic system itself. This article gives an overview of the FRIEND project and the robotic systems there from evolved. It is divided into a practical part, which presents the systems from a user oriented perspective, and into a theoretical part, which satisfies the system-engineer’s point of view. The user oriented part outlines the different development steps, functional improvements, hardware setups and lessons learned since 1997. Here, the facilities of the FRIEND-I system as well as a description of the innovations of the FRIEND-II system, currently under development, are described with specific emphasis of the AMaRob project. The reader becomes aware of the functionalities and services offered by the FRIEND rehabilitation robot and of the challenging technical complexity with which the development has to deal. The theoretical part is focused on the concept of semi-autonomous task-execution as a means of reasonable complexity reduction. Due to the consequent application of this concept a technically manageable robotic system emerges, which is able to execute tasks on a high level of abstraction in a reliable and robust manner. Within this context semi-autonomous task-

System-controlled user interaction within the service robotic control architecture massive

This paper presents an approach to reduce the technical complexity of a service robotic system by means of systematic and well-balanced user-involvement. By taking advantage of the users cognitive capabilities during task execution, a technically manageable robotic system, which is able to execute tasks on a high level of abstraction reliably and robustly, emerges. For the realisation of this approach, the control architecture MASSiVE has been implemented, which is used for the control of the rehabilitation robot FRIEND II. It supports task execution on the basis of a priori defined and formally verified task-knowledge. This task-knowledge contains all possible sequences of operations as well as the symbolic representation of objects required for the execution of a specific task. The seamless integration of user interactions into this task-knowledge, in combination with MASSiVEs user-adapted human–machine interface layer, enables the system to deliberately interact with the user during run-time.

Simulation tool for kinematic configuration control technology for dexterous robots

A simulation tool for designing and testing control devices for advanced robots with an arbitrary number of redundant joints is presented. The novel real-time suitable control technology, called kinematic configuration control, bases on a new approach of resolving redundancy. For redundant manipulators with practical relevant regular kinematic chains, the approach allows to obtain a closed-form solution of the inverse kinematics, as for common nonredundant industrial robots. These offline evaluated symbolic inverse functions need low online computing power. Therefore also controllers for dexterous robots with high degree of redundancy can be implemented on low-cost consumer computer. The described PC-based simulator supports the design of this kind of controller. During a simulation process, the algorithm-specified parameters of inverse functions are optimized. The control algorithm is formed as a separate controller module. After verification it can be effortlessly transferred into a real robot control device.

Programming of Intelligent Service Robots with the Process Model “FRIEND::Process” and Configurable Task-Knowledge

In Alex Proyas’s science fiction movie “I, Robot” (2004) a detective suspects a robot as murderer. This robot is a representative of a new generation of personal assistants that help and entertain people during daily life activities. In opposition to the public opinion the detective proclaimed that the robot is able to follow his own will and is not forced to Isaac Asimov’s three main rules of robotics (Asimov, 1991). In the end this assumption turned out to be the truth. Even though the technological part of this story is still far beyond realization, the idea of a personal robotic assistant is still requested. Experts predicted robotic solutions to be ready to break through in domestic and other non-industrial domains (Engelberger, 1989) within the next years. But up to now, only rather simple robotic assistants like lawn mowers and vacuum cleaners are available on the market. As stated in (Grafe & Bischoff, 2003), all these systems have in common that they only show traces of intelligence and are specialists, designed for mostly a particular task. Robots being able to solve more complex tasks have not yet left the prototypical status. This is due to the large number of scientific and technical challenges that have to be coped with in the domain of robots acting and interacting in human environments (Kemp et al., 2007). The focus of this paper is to describe a tool based process model, called the “FRIEND::Process”1, which supports the development of intelligent robots in the domain of personal assistants. The paper concentrates on the interaction and close relation between the FRIEND::Process and configurable task-knowledge, the so called process-structures. Process-structures are embedded in different layers of abstraction within the layered control architecture MASSiVE2 (Martens et al., 2007). Even though the usage of layered control architectures for service robots is not a novel idea and has been proposed earlier (Schlegel &

Rahmenwerk zur teilautonomen Aufgabenbearbeitung mit Rehabilitationsrobotern (Framework for Semi-autonomous Task Execution with Rehabilitation Robots)

Abstract Rehabilitationsroboter mit einem realen Nutzen für ihre Anwender müssen Aufgaben auf hohem Abstraktionsniveau bearbeiten können. Dies macht die situationsbezogene Planung und Ausführung von Handlungen durch das Robotersystem erforderlich. Beim jetzigen Stand der Forschung und Entwicklung ist die Umsetzung dieser Forderung nur durch die Einbeziehung der kognitiven Fähigkeiten des Anwenders in den Prozess der Aufgabenbearbeitung umsetzbar. Im vorliegenden Beitrag wird ein softwaretechnisches und algorithmisches Rahmenwerk vorgestellt, das eine Erweiterung von Rehabilitationsrobotern um die Fähigkeit zur teilautonomen Aufgabenbearbeitung ermöglicht.

Soft computing-based robust contact/non-contact detection during serving a drink task

This paper presents a novel application of fuzzy logic in the field of rehabilitation robots. This soft computing-based approach is utilized for the task of serving a drink to a person, to determine whether contact between the persons mouth and the drinking glass has occurred. The decision is based on force sensor data, which are pre-processed before being applied for decision-making. The design or decisionmaking is accomplished by means of force information as well as the first and second differential of the force. The results demonstrate that the detection algorithm is of high reliability and robustness under various environmental conditions.