Public defence in Automation, Systems and Control Engineering, M.Sc.(Tech.) Tran Nguyen Le

The title of the thesis: Harnessing the physical properties of objects for robotic grasping and manipulation

Doctoral student: Tran Nguyen Le
Opponent: Prof. Roberto Calandra, Technical University of Dresden, Germany
Custos: Prof. Ville Kyrki, Aalto University School of Electrical Engineering, Department of Electrical Engineering and Automation

To deal with the global challenges of the modern world, such as the accelerating aging of the population as well as the emergence of unforeseeable epidemics, it is expected that robots will soon seamlessly integrate into daily life. Let us envision a future where robots assist humans in daily tasks such as washing dishes, folding clothes, or taking care of patients in healthcare settings. The key to this transformative vision lies in their ability to robustly and autonomously grasp and manipulate diverse objects, ranging from cutlery and tools to delicate food items. While substantial progress has been made towards this goal in the last decade, the current methods often rely on assumptions about an object's physical properties, such as uniform rigidity and friction across the surface. These assumptions limit the effectiveness of such methods in handling more complex artefacts, such as multi-material or deformable objects, which are commonly encountered in domains such as healthcare and household tasks. 

To address this problem, this dissertation seeks to explore the potential of explicitly estimating and harnessing two crucial physical properties of objects, i.e., surface friction and deformability, to develop resilient grasp and manipulation planners beyond these assumptions. Specifically, the dissertation first analyzes the importance of these physical properties in robotic grasping and manipulation. It then introduces an interactive perception method utilizing machine learning techniques to learn the physical properties of objects through both observation and interaction. Finally, it explores different ways to harness the learned physical properties in robotic grasping and manipulation, ranging from grasp analysis (assessing the quality of a grasp), to grasp planning (planning where to grasp to yield the best performance), and manipulation planning where the method is demonstrated through a practical dish-cleaning task. 

Together, all the results presented in this dissertation indicate that acquiring and harnessing knowledge of an object's physical properties beyond its shape increases the robustness and performance of grasp planning and manipulation planning methods. This dissertation hopefully will motivate the robotics community to move beyond current assumptions and explore deeper object understanding when developing new grasping and manipulation approaches.

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact:

Doctoral theses in the School of Electrical Engineering: https://aaltodoc.aalto.fi/handle/123456789/53