|
||||||||||||||
Plenary Lecture A Framework for Physical Interaction and its Execution based on Multiple Sensors
Professor
Angel P. del Pobil
Abstract:
Autonomous robot manipulation is one of the most
important challenges in robotics. It involves three
challenges: versatility, defined as the capability to
adapt to different situations, instead of being limited
to a particular task; autonomy, that concerns the level
of independence in the robot operation, and
dependability, that refers to the capability of
successfully completing an action even under important
modeling errors or inaccurate sensor information. A
complete manipulation task involves two sequential
actions: that of achieving a suitable grasp or contact
configuration, and the subsequent motion required by the
task. We propose a unified framework with the
introduction of task-related aspects into the classical
knowledge-based grasp concept, leading to task-oriented
grasps. In a similar manner, grasp-related issues are
also considered during the execution of a task, leading
to grasp-oriented tasks. We call this unified
representation physical interaction. In the talk I will
first present a theoretical framework for the integrated
specification of physical interaction tasks, supporting
a great variety of actions. Next, the problem of
autonomous planning of physical interaction tasks will
be addressed. I will then focus on the dependable
execution of these tasks, and adopt a sensor-based
approach with three different types of sensor feedback:
force, vision and tactile. The methods proposed provide
important advances with respect to the state-of-the-art
versatility, autonomy and dependability of robotic
manipulation, allowing to address a wide range of tasks.
All these contributions are validated with several
experiments using different real robots placed on
household environments.
Brief Biography of the Speaker: |
|
|||||||||||||
![]() |
![]() |
![]() |
||||||||||||
copyright - designed by WSEAS |