Open Access

Matthias Stueben

• Assistive robots have the potential to improve human lives significantly in many different forms. The possible applications are widespread, ranging from workplace assistance to healthcare, therapy, and household applications. Mobile manipulators, combining mobility and manipulation abilities, are an especially promising type of robot for these kinds of applications. Robots operating in unstructured, human-centered environments, as well as interacting with humans, have to be able to safely and robustly deal with a dynamically changing environment and sometimes unexpected motions by humans. Creating robot motions that respect and appropriately react to any possible change in the environment is a challenging research question. It begins with the specification of the requirements for robot motions for the respective task. Both planning and reactive control are required for robust robot motions. These two approaches are not always simple to combine: following a plan often conflicts withAssistive robots have the potential to improve human lives significantly in many different forms. The possible applications are widespread, ranging from workplace assistance to healthcare, therapy, and household applications. Mobile manipulators, combining mobility and manipulation abilities, are an especially promising type of robot for these kinds of applications. Robots operating in unstructured, human-centered environments, as well as interacting with humans, have to be able to safely and robustly deal with a dynamically changing environment and sometimes unexpected motions by humans. Creating robot motions that respect and appropriately react to any possible change in the environment is a challenging research question. It begins with the specification of the requirements for robot motions for the respective task. Both planning and reactive control are required for robust robot motions. These two approaches are not always simple to combine: following a plan often conflicts with reacting to changes and a method of reconciling them is required. This dissertation presents a method of specifying and generating robot motions for mobile manipulators based on geometric constraints. Constraints express whether the current robot position satisfies its requirements, and if not, how far off it is. Constraint rules are introduced, and are used to adapt the parameters of the constraints to changes in the environment. Constraint controllers are applied to generate reactive robot motions based on the current value of the constraints. Different types of constraints allow for different types of reactions. The concept of an action brings together multiple constraint specifications, in which priorities, tolerances, and weights are used to express the relative importance of the requirements. Besides reactive control, a method to create and safely execute motion plans for action specifications is presented. Furthermore, a method for automatically detecting when planning is required is described. In order to demonstrate the capabilities of the presented approach, it is evaluated on two different case studies. The first consists of an object handover between robot and human, the second considers a robot shining a flashlight to provide light to a human. Both case studies are evaluated in a successful case and different failure scenarios to assess the robustness of the approach.…