The DESIRE Service Robotics Initiative

In 2006 a consortium of twelve German partners -including some leading providers of industrial robotic technology components (RTCs)teamed up with some research institutes and universities to work on a joint publicly funded four year project called DESIRE. It is focused on service robots, which shall offer their services under unconstrained conditions in immediate proximity of users in everyday situations. The dedicated goal was to address the extraordinary high demands on functional competence (“capability”), reliability (“dependability”) and an acceptable price (“affordability”). In this article we focus on the open issues of defining a suitable software architectural framework on one hand in combination to the use of an advanced planer to cope with the high complexity demands of service robotics on the other. We report on the current status of the DESIRE work, especially the architecture and the planer component. For the architecture side it is a widely accepted fact that in the large a major revision of the methods and technologies currently used for the implementation of robotic software is called for. Like in many other areas the construction of robot control software should focus more on the deployment of common off-the-shelf software and components in combination with a conceptually clean integration of them. This topic has drawn major attention of a large number of researchers [1], conference sessions and workshops and also some major organizations initiated and funded respective activities [7]. The issues are much under debate and far from being settled. The integration of a task planner into an autonomous robot increases the robot’s level of intelligence and flexibility by altering the way the robot is controlled, moving away from predefined sequences of detailed user instructions to a more sophisticated target oriented approach. It is not longer required to provide the robot with a fully worked out description of its task (e.g., “Go to the big table, take the plate, come back to me an give the plate to me!”) but rather to state some declarative targets (e.g., “Give me the plate!”) and leave it to the robot to find a suitable way to achieve them on its own. Accordingly, planning can be understood as reasoning on a human level of abstraction. Task planning itself is a thoroughly investigated subfield in artificial intelligence [3]. However, in a robotics context, one has to deal with aspects complicating the application of task planning, some of which are: Imperfect knowledge of the surroundings, non deterministic changes, and user interaction. One of the main goals of this project, as far as the planning part is concerned, is to make task planning more suitable for everyday use in a robotics context.