IEEE Robotics & Automation Magazine - September 2018 - 86

Object
Database
Supervisor

HRI Interface
High-Level
Actions

MoveIT! and
Libraries
Executors

Low-Level
Actions

Dynamic
Manipulation
Task

Perception Data

Dynamic
Manipulation Tasks
DyT 1
DyT 2
...
DyT n

Perception
Planned
Trajectory
Base Velocity
Commands

Joint Trajectory
Controller
Base Controller

Set Joint
State
Get Joint
State

Real World

Set Joint
Position
Set Wheels
Position

Simulation
Environment

Raw Sensors Data
Figure 2. The RoDyMan high-level control architecture. DyT: dynamic manipulation task.

developed [12]. Future opportunities for the platform
include increasing the level of collaboration between the
robot and human operators, allowing for shared-task planning and execution.

(a)

(b)

Figure 3. (a) The volumetric tetrahedral mesh (elements in blue
colors), and (b) the modeling of fractures.

composed by the system to achieve the desired goal [15]. In
this context, if a nonprehensile manipulation action is
required, the related low-level controller is invoked from the
dynamic manipulation task list while the supervisor awaits its
termination. Otherwise, the executor module is responsible
for the action by implementing both the path and the motionplanning functionalities to find an obstacle-free route for the
end effectors of the robot and its base. This module relies on
MoveIT! [16], a framework that integrates a universal robot
description file, an open motion-planning library, and other
tool kits. The generated trajectories for the joints and the base
of the robot are streamed to the robot actuators from the controller modules. Information about the robot's environment is
then extracted from the perception module via image-elaboration algorithms.
The proposed high-level control architecture perfectly
matches the requirements of the RoDyMan robotic platform, statically allocating the best low-level controller to
accomplish desired actions. This improves the current literature because very few methods of decomposing a highlevel task for nonprehensile manipulation have been
86

IEEE ROBOTICS & AUTOMATION MAGAZINE

september 2018

Perception of Deformable Objects
In this section, we present a method to cope with real-time
(35 ft/s) tracking of a deformable object using the point cloud
data provided by the RGB-D sensor. We propose several contributions, such as handling various large elastic deformations
while ensuring physical consistency and coping with fractures, rigid motions, and occlusions [17]. An example is
stretching and tossing pizza dough.
Because the considered system attempts to deal with large
deformations and elastic volumetric strains, a realistic mechanical model is employed that is based on continuum mechanics
and on a volumetric tetrahedral finite element method (FEM).
This model can be used for real-time applications through the
Simulation Open Framework Architecture. Explicit physical
modeling would enable a reliable prediction of internal forces
undergone by the object.
To model elastic deformations, the infinitesimal strain theory and Hooke's law are taken into account, providing a linear
relation between the displacement of the tetrahedral elements
of the mesh and the internal forces exerted on their nodes.
The corotational approach is used as a compromise between
the ability to model large deformations of the elements and
the computational efficiency.
Based on the FEM corotational model, fractures in the
mesh are detected by decomposing the internal forces on the
nodes into tensile and compressive forces to measure pure
tensile forces acting on each node; this is done through something called a separation tensor. The fracture is propagated by
simply removing attached elements intersected by the fracture
plane. The model is illustrated in Figure 3.

IEEE Robotics & Automation Magazine - September 2018

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2018