Analysis and applications of reachability and capability maps for robotic manipulators

University essay from German Aerospace Center, Lulea Tekniska Universitet, Julius Maximillians Universitat Wurzburg

Abstract: Robotic arms are the most popular robots on the market. Technology behind the arm manipulators and their
sensors is getting more accessible, which results in an increased interest from both commercial and research
communities. Autonomous applications of robotic arms needs a combination of good sensory input and previous
knowledge to speed up application development. Path planning for the robotic end effector, required to design
a trajectory to move from an initial to a goal position, requires both knowledge of the kinematic structure of
the robot as well as sensory information coming from the environment, that helps to identify key elements like
the objects to manipulate, surfaces of support or possible obstacles. The kinematic structure that results from
a design stage is fixed, and knowledge of the workspace of the robot and its dexterity in this space can be
preprocessed to speed up on-line applications.
This thesis proposes a novel way to address the problem of computing the workspace of a robotic arm and
its dexterity within this space. Our proposed off-line analysis of reachability is designed to answer questions
about workspace shape and quality. We show how to use the precomputed structure for on-line grasp selection,
operational workspace selection, collision free path planning, path validation or robot pose selection.
Our approach builds on top of the concept of reachability and capability maps. Traditional methods to
generate such maps use forward or inverse kinematics, and we investigate the advantages and limitations of
both. We later propose a hybrid method which combines their advantages while maintaining low generation
time. Quality of the results is evaluated by a prediction accuracy test. The structure of the map is designed
such that it is also suitable for on-line applications. Real-time visual information is incorporated into the map’s
data structure to improve real-world interactions such as grasp selection or collision-free path planning. To con-
clude, several examples of real applications that illustrate the usefulness of the map are presented and discussed.