New version page

Benchmarking Grasping and Manipulation

Upgrade to remove ads

This preview shows page 1-2-3 out of 8 pages.

Save
View Full Document
Premium Document
Do you want full access? Go Premium and unlock all 8 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 8 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 8 pages.
Access to all documents
Download any document
Ad free experience

Upgrade to remove ads
Unformatted text preview:

 Abstract—This paper presents a number of concepts related to benchmarking and evaluation of grasping and manipulation. A set of “Objects of Daily Living” based on a review of common domestic objects for manipulation as identified from sources in the literature is put forward, along with the physical properties of sample objects in those categories. Next, an experimental evaluation of the coefficient of static friction between these objects and a number of common household surfaces is performed. A key failure mode in unstructured object grasping occurs when the manipulator applies large contact forces that move the object out of grasp range. These results therefore give insight into the likelihood of a target object remaining in place to be successfully grasped in the presence of contact forces from the robot arm. This paper also presents a new classification of the Activities of Daily Living (ADLs), putting forth a standard categorization for the application of robotics in human environments. These topics and results have a number of uses related to benchmarking and performance evaluation in robotic manipulation, assistive technology, and prosthetics. I. INTRODUCTION S robotic grasping and manipulation moves closer to practical implementation in human environments, it has become clear that quantitative metrics for evaluating performance in the presence of uncertainty must be developed. However, the large variability in the types and specifics of the grasping and manipulation tasks that can be performed by robots in domestic or workplace settings, as well as separating hardware performance from software-related factors (e.g. planning and control), makes creating absolute and translatable measures difficult. This paper is the first in a series of planned papers related to benchmarking for grasping and manipulation and contributes to the topic in a number of ways. First, we create a new sub-classification of the Activities of Daily Living [1, 2] for the application of robotics in human environments, putting forth a standard categorization that allows robotic tasks to be discussed in terms of the analogous human tasks and their hierarchal classifications. We then put forth an extensive list of “Objects of Daily Living”, collected from key publications in the literature of the fields of robotics, prosthetics, and occupational therapy. K. Matheus is with the School of Engineering and Applied Science, Yale University, New Haven, CT 06511 USA (e-mail: [email protected]). A.M. Dollar is with the School of Engineering and Applied Science, Yale University, New Haven, CT 06511 USA (phone: 203-436-9122, fax: 203-432-6775, e-mail: [email protected]). These objects are categorized according to the most relevant Activities of Daily Living subcategory and the mass and dimensions of representative examples are given. This collection of common objects for grasping and manipulation in human environments can be referred to by researchers seeking to select a standard set of objects for testing and evaluation or a means of relating objects to a standard Activity of Daily Living category, and vice-versa. Finally, we experimentally determine the amount of contact force required to displace those objects on a variety of common household surfaces. In unstructured human environments, the uncertainty inherent with imprecise sensing of unknown objects typically leads to a poor model of target object geometry and position/orientation. This poor object model leads to positioning errors of the robot manipulator arm and/or finger placements, which, in turn, can lead to large forces being inadvertently applied to the target object, potentially displacing it such that it is out of grasp range. One measure of grasping performance in unstructured environments, therefore, is the magnitude of force applied to a target object during acquisition [3, 4]. If the horizontal component of this force exceeds the frictional force between the object and the surface it rests on, the object will be moved from its resting position, often causing the grasp attempt to fail. A table of experimentally-determined frictional properties for a large number of common objects and surfaces allows the researcher to predict whether target objects will move under certain grasping conditions, as well as a means to evaluate hardware, sensing, and/or algorithm performance. This information will also prove useful for simulation environments (e.g. [5-7]), as well as in developing grasp and planning databases [8], allowing for hardware designs and planning algorithms to be evaluated against a large number of target objects. A more precise estimate of the coefficient of static friction between an object and surface will add fidelity to simulation results, improving translation to real-world applications. We begin this paper with a discussion of the Activities of Daily Living and the proposed sub-categorization. We then introduce the concept of “Objects of Daily Living”, a collection of common household items associated with the Activities of Daily Living. We review the associated literature and put forth an extensive list of objects identified as common and/or important in domestic environments, including the physical proportions of common embodiments of the objects. Lastly, we present an experimental study in Benchmarking Grasping and Manipulation: Properties of the Objects of Daily Living Kayla Matheus and Aaron M. Dollar, Member, IEEE AThe 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems October 18-22, 2010, Taipei, Taiwan978-1-4244-6676-4/10/$25.00 ©2010 IEEE5020which we determine the frictional properties of these objects on a number of common household surfaces in order to lend insight into the likelihood of the objects being successfully grasped in the presence of the uncertainty inherent with manipulation in unstructured environments. II. ACTIVITIES AND OBJECTS OF DAILY LIVING A. Activities of Daily Living Many fields related to occupational therapy, rehabilitation, and gerontology use the term “Activities of Daily Living” (ADLs) in evaluating the ability of a patient to perform self-maintenance and other daily tasks crucial for unassisted living [1, 2, 9-13]. The term is generally used broadly and qualitatively. Many different sub-categories of the ADLs have been proposed to classify an individual’s level of independence, including Physical Self-Maintenance (PSM) [9],


Download Benchmarking Grasping and Manipulation
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Benchmarking Grasping and Manipulation and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Benchmarking Grasping and Manipulation 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?