| Algorithms & Applcations Group | Home | Research | Publications | People | Resources | News
Algorithms & Applications Group
Deformable Objects

Project Personnel:Nancy Amato

Probabilistic Roadmap Motion Planning for Deformable Objects
In this work, we find a path which requires the robot to deform in order to follow it. The path may contain collisions for the rigid (undeformed) version of the robot. After finding such path, we employ bounding box deformation or geometric deformation to deform the robot to avoid collisions. Our approach deforms the robot only in necessary conditions (if there is a collision). The proposed method generates perceptually convincing motion efficiently.

Planning Motion in Completely Deformable Environments
We present a framework for planning paths in completely deformable, elastic environments. In particular we apply a deformable model to the robot and obstacles in the environment and we present a kinodynamic planning algorithm suited for this type of deformable motion planning. To the best of our knowledge, this is the first work that plans paths in totally deformable environments.

Robot automation and motion planning has been inseparable since very first robot. There has been lots of interest in motion planning, especially in methods that utilize the probabilistic roadmap methods. It is a common practice to represent robots as rigid objects or a series of links each of which is a rigid object. Although this is a realistic situation, there may be cases where a more flexible representation, such as the one proposed in this work, of the robot would be preferred.

In our approach, we first find a path which requires the robot to deform in order to follow it. The path may contain collisions for the rigid (undeformed) version of the robot. Note that there is a direct relation between collision volume and the energy that we need to deform the robot to a collision-free shape. We associate this volume with the deformation energy. Since we don't know the energy priory and collision volume is hard to get, we used the parameters returned from our feasibility metrics as the approximate energy. These parameters are the size of the shrunk robot with respect to the original robot and the value of penetration for each configuration found. Following figure shows roadmap generated by these methods.

Left : roadmap generated by shrunk robot. Right : roadmap generated by enabling penetration.

After finding such path, we employ two different methods, bounding box deformation or geometric deformation to deform the robot to avoid collisions. Our approach deforms the robot only in necessary conditions (if there is a collision). It is our observation that the robot is deformed only if it has intersection with obstacles. Deformation can be divided into deformer and deformable object. Deformer pushes part of deformable object to collision free state and deformable object then changes shape according to external forces. We can see from following pictures that obstacles are made as deformer which pushes deformable object into collision-free configuration.

Left : Bounding Box Deformation. Using modified ChainMail3D and FFD. Right : Geometric Deformation.

We study the following examples:

 Sliding of DSMFT letters. Using Bounding-Box Deformation. (avi 17.8MB) (mpeg 4.2MB) Using Geometric Deformation. (JPG ) Sliding of a teapot and duck. Using Bounding-Box Deformation. (mpeg 1.6MB) Narrow Passage. (note: for visibility, the bounding ball is removed during rendering.) Using Bounding-Box Deformation. (avi 3.6MB) (mpeg 2.5MB) Undeformed v.s. deformed (mpeg 2.0MB) Using Geometric Deformation. (JPG ) Stamping. Using Bounding-Box Deformation. (avi 1.9MB) (mpeg 1.5MB) Using Geometric Deformation. (JPG )

Though motion planning has been studied extensively for rigid and articulated robots, motion planning for deformable objects is an area that has received far less attention. In this paper we present a framework for planning paths in completely deformable environments.

In particular we apply a deformable model to the robot and obstacles in the environment and we present a kinodynamic planning algorithm suited for this type of deformable motion planning. The planning algorithm is based on the Rapidly-Exploring Random Tree (RRT) path planning algorithm. To the best of our knowledge, this is the first work that plans paths in totally deformable environments.
 Forces are applied to both robots and obstacles including volume and distance preservation, collision response, manipulation, and gravity forces. States are expanded from using an RRT during planning. A state S(t) is brought toward a randomly sampled point x by applying forces F(t).

 Manipulation forces are used to drag a robot through the environment and resulting states are stored in the tree. Manipulation forces are selected from Body, Control Point and Interpolation forces.

Simulation Results:

 Plates Environment. Low Resolution. (avi 2.8MB) (mov 12MB) High Resolution. (avi 2.8MB) (mov 35MB) Windows Environment. Low Resolution. (avi 2.3MB) (mov 23MB) High Resolution. (avi 2.3MB) (mov 74MB) Falling Objects Environment. Low Resolution. (avi 2.0MB) (mov 13MB) High Resolution. (avi 2.0MB) (mov 37MB)

Planning Motions for Shape-Memory Alloy Sheets, Mukulika Ghosh, Daniel Tomkins, Jory Denny, Sam Rodriguez, Marco Morales Aguirre, Nancy M. Amato, Origami 6, 6(6):501-511, Dec 2015.
Journal(pdf)

Approximate Convex Decomposition and Its Applications, Jyh-Ming Lien, Ph.D. Thesis, Department of Computer Science and Engineering, Texas A&M University, Dec 2006.
Ph.D. Thesis(pdf, abstract)

Planning Motion in Completely Deformable Environments, Samuel Rodriguez, Jyh-Ming Lien, N. M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. 2466-2471, Orlando, FL, May 2006.
Proceedings(ps, pdf, abstract)

Probabilistic Roadmap Motion Planning for Deformable Objects, O. Burchan Bayazit, Jyh-Ming Lien, Nancy M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. 2126-2133, Washingon, D.C., May 2002. Also, Technical Report, TR01-003, Parasol Laboratory, Department of Computer Science, Texas A&M University, Oct 2001.
Proceedings(ps, pdf, abstract) Technical Report(ps, pdf, abstract)

Supported by NSF, Dept. of Education, Texas Higher Education Coordinating Board

Project Alumni:O. Burchan Bayazit,Jyh-Ming Lien,Marco Morales

 Parasol Home | Research | People | General info | Seminars | Resources   Parasol Laboratory, 425 Harvey R. Bright Bldg, 3112 TAMU, College Station, TX 77843-3112  parasol-admin@cse.tamu.edu      Phone 979.458.0722     Fax 979.458.0718  Department of Computer Science and Engineering | Dwight Look College of Engineering | Texas A&M University     Privacy statement: Computer Science and Engineering Engineering TAMU Web Accessibility Policy and Law - Web Accessibility and Usability Standards  -   Contact Webmaster