Please use this identifier to cite or link to this item:
Title: Reactive Motion Planning for Robotic Manipulators and Interaction Control with Deformable Environment: Application to Physical Human-Assistive Tasks
Authors: Δομέτιος, Αθανάσιος, Δρ.
Τζαφέστας Κωνσταντίνος
Keywords: Ρομποτική; Ρομποτική υποβοήθησης; Φυσική αλληλεπίδραση ανθρώπου - ρομπότ; Σχεδιασμός ρομποτικής κίνησης; Έλεγχος αλληλεπίδρασης
Issue Date: 2-Jul-2021
Abstract: Natural human-robot physical interaction has a key role in the acceptance of robotic applications in everyday life. Especially in the case of applications for the care and assistance of people with mobility or other impairments, the functional requirement of physical contact between humans and robots is one of the key factors that determine the safety of the robotic system. Such human-robot interaction applications require an interactive motion planning system based on a continuous monitoring of the human condition, in order to achieve a completely safe and continuously adaptable robotic motion and task planning. Towards this end, this dissertation deals with reactive adaptation and motion generation of human demonstrated interactive tasks with deformable surfaces, such as those of the human body parts. Particularly, in the first part of this thesis, and in the context of a broader research effort aimed at building a flexible robotic bath mechanism, a motion planning algorithm was developed, which uses the visual feedback from a depth camera and the corresponding scene perception information, in order to adapt predefined, time scalable trajectories on curved and deformable surfaces, such as the human body parts, with simultaneous avoidance of obstacle areas, such as injuries. The adaptation is achieved with the establishment of bijective transformations, which reformulate the tracking problem to a 2D Canonical Space. Accurate trajectory tracking is then realized with a Navigation Function (NF) controller with proven globally uniformly asymptotic convergence. The proposed algorithm was tested both in lab conditions and in a real clinical environment with elderly users in both dry and humid conditions. A clinical validation study was conducted, which focused on the acceptance and operation aspects of such a complex system by elderly users .In the next phase of the thesis, an integrated system based on Dynamic Motion Primitives (DMP) approach is proposed, which can learn and encode demonstrated washing actions by professional nursing experts, imitating their actions. The washing actions were recorded with the use of optical motion tracker systems, analyzed and decomposed into primitive actions appropriate for robotic execution. The learned motions can then be adapted to the user’s body parts, compensating their motion or deformation, as well as on-line modified with respect to their execution parameters, in order to meet the user’s requirements during of the robotic assistance process. This system was experimentally validated with the use of a humanoid robot, which executed a wiping scenario, demonstrating the applicability of this method in real world scenarios of assistive robotics. Alternatively to the DMP approach, a NF method is proposed in order to learn and reproduce the way an expert clinical carer executes the bathing activities, by means of constructing artificial repulsive potential fields generated by virtual obstacles, which in essence represent the demonstrated motions. In the final stage of this dissertation, an efficient interactive motion planning framework is proposed, to effectively and accurately control a robotic manipulator executing interactive tasks on the surface of a deformable object. The proposed interactive motion planning framework is based on a mesh representation of the object, integrating three efficient preprocessing algorithmic steps, including visual object segmentation, FEM deformation tracking and local mesh parameterization. The use of barycentric coordinates, defined on the mesh triangles, enables the establishment of bijective transformations between the deformable part of an object surface and its planar (static and dynamic) parameterized mapping. By merging these spatial transformations with the preprocessing steps, in combination with an active stiffness scheme for robot manipulator control, we are able to achieve accurate and reactive motion planning of interactive trajectories, even under large and persistent visual occlusions. An extensive experimental evaluation study is presented, involving a robotic manipulator in interaction with a hemispherical model of controllable periodic active deformation, which permits precise ground truth derivation. Motion planning accuracy is evaluated in comparison with the previously described direct vision-based approach, showing clearly superior performance of the mesh-based approach under all experimental conditions. The performance of the proposed framework is also further highlighted in tasks involving physical point tracking, interactive programming by human demonstration, as well as contact force regulation.
Appears in Collections:Διδακτορικές Διατριβές - Ph.D. Theses

Files in This Item:
File Description SizeFormat 
athdom_phd_thesis_artemis.pdf32.77 MBAdobe PDFView/Open

Items in Artemis are protected by copyright, with all rights reserved, unless otherwise indicated.