Widespread deployment of robots in offices, hospitals, and homes is a highly anticipated breakthrough in robotics. In such environments, the robots are expected to fulfil new tasks as they arrive in contrast to repetitive tasks. Such environments are unstructured and may impose various constraints on a robot's motion. Therefore, robots should be able to solve new instances of complex problems where kinematic and dynamical constraints must be respected. In this thesis we investigate how to autonomously incorporate these constraints in a robotic problem specification and how to develop motion planning and control techniques that solve such a problem.

The combination of environment-imposed constraints including obstacles, with a robot's own limitations, e.g., actuation bounds, results in many scenarios where a robotic task becomes a non-convex problem. This inhibits the commonly assumed independence between motion planning and control, making various classical control approaches practically infeasible. In the first part of this thesis, we introduce our contribution toward autonomous incorporation of kinematic and dynamical limitations at planning level by representing the robot action space as a set of feedback motion primitives. We show how the dual path planning and path non-existence problems can be solved for a mobile robot even in presence of external disturbance using feedback motion primitives. We further extend the applicability of motion primitives to long-horizon planning problems and address complex tasks specified as linear temporal logic (LTL) formulas by introducing a guided search scheme. Ultimately in this part, we lay a theoretical foundation for automated construction of such feedback motion primitives for a large class of systems by decomposing their state-space into smaller regions where locally linear controllers can be synthesized.

In the second part of the thesis, we investigate the motion planning and control problem in the presence of dynamical uncertainty. In absence of accurate models (e.g., when a manipulator operates while it is in contact with the environment) we should be able to collect data from the interaction to make informed decisions toward task satisfaction. Data-driven approaches are becoming increasingly popular in robotic control under uncertainty and have enabled tackling a wide range of new tasks. However, methods developed to verify safety constraints for a controller are inherently model based. This motivated us to adopt a model-based approach to data-driven control that enables explorative data collection while ensuring system safety. Furthermore, we propose data collection policy alternatives to reduce the well-known distribution shift effect in model learning.

A crucial step towards robot autonomy-in environments other than the strictly regulated industrial ones-is to create controllers capable of adapting to diverse conditions. Human-centric environments are filled with a plethora of objects with very distinct properties that can still be manipulated without the need to painstakingly model the interaction dynamics. Furthermore, we do not need an explicit model to safely complete our tasks; rather, we rely on our intuition about the evolution of the interaction that is built upon multiple repetitions of the same task.Accurately translating this ability in how we control our robots in contact-rich tasks is almost infeasible if we rely on controllers that operate based on analytical models of the contacts. Instead, it is advantageous to utilize data-driven techniques that approximate the models based on interactions, much like humans do, and encompass the varying dynamics with a single model. However, for this to be a feasible alternative, we need to consider the safety aspects that occur when we move away from rigorous mathematical models and replace them with approximate data-driven ones.

This thesis identifies three safety aspects of data-driven control in contact-rich manipulation: good predictive performance, increased interpretability for the models, and explicit consideration of safe inputs in the face of modelling errors or uninterpretable predictions. The first point is addressed through a model-training scheme that improves the long-term predictions in a food cutting task. In the experiments it is shown that models trained this way are able to adapt to different dynamics efficiently and their prediction error scales better with longer horizons. The second point is addressed by introducing a framework that allows the evaluation of data-driven classification models based on interpretability techniques. The interpretation of the model decisions helps to anticipate failure cases before the model is deployed on the robot, as well as to understand what the models have learned. Finally, the third point is addressed by learning sets of safe states through data. These safe sets are then used to avoid dangerous control inputs in a control scheme that is flexible and adapts to dynamic variations while effectively encouraging the safety of the system.