Robot Skills: Design of a Constraint-Based Methodology and Software Support (Robot vaardigheden: Ontwerp van een beperkingsgebaseerde methodologie en software ondersteuning)

The days when robots were only used to move their endpoint from point Ato B, passing through point C, safely hidden behind a fence, are over. Nowadays robots have to grasp objects, move them around, manipulate them using two arms, while keeping controlled distances from humans and fragile objects or even during physical interaction with humans. Furthermore, all these tasks have to be executed as optimally as possible. Six degrees-of-freedom industrial arms are being replaced by mobile platforms with two redundant arms, both equipped with force sensors, three-fingered grippers with touch sensors, a head with cameras, laser scanners, inertial sensors, etc. In the coming years the habitat of robotic systems will evolve from the industrial work cells to domestic, cluttered and populated environments. To cope with the increasing complexity, not only a highly modular methodology for motion specificationis needed. Also the online motion coordination becomes very important since it has to cope with the dynamically changing environment in which the robotic system works.This thesis presents a highly modular control hierarchy for the coordination of constraint-based motion specification that can be used with a high variety of sensors, robotic systems and tasks. To create a highly modular coordination system, a modular motion specification methodology needs to be used as the underlying framework. iTaSC, or instantaneous Task Specification using Constraints, is such a modular framework that allows us to specify motion by imposing those constraints on the interaction between the robot and the environment that are important for the task at hand. This thesisextends the iTaSC methodology to not only allow the specificationof constraints in feature spaces, which are built up from exactly six independent geometric coordinates, but to allow the specification of constraints in any feature space for which an interaction model can befound. Calculating the robot motion from the specified constraints can be very computationally expensive in systems that include a lot of actuated joints, such as humanoid robots. Therefore the quest for efficient algorithms is still ongoing. This thesis discusses one of the most ground-breaking algorithms for acceleration-based constrained hybrid (combination of forward and inverse) dynamics, and explains how it can be extended for tree-like kinematic robot structures and how it can be used to solve an iTaSC motion specification.Building software su