Distributed Manipulation

Kluwer Academic Publishing 2000

Editors:
Karl F. Böhringer, Dept. of Electrical Engineering, University of Washington, Seattle
Howie Choset, Dept. of Mechanical Engineering, Carnegie-Mellon University

Introduction

Distributed manipulation effects motion on objects through a large number of points of contact. The primary benefit of distributed manipulators is that many small inexpensive mechanisms can move and transport large heavy objects. In fact, each individual component is simple, but their combined effect is quite powerful. Furthermore, distributed manipulators are fault-tolerant because if one component breaks, the other components can compensate for the failure and the whole system can still achieve its objective. Finally, distributed manipulators can execute a variety of tasks in parallel.

Distributed manipulation can be performed by many types of mechanisms at different scales. Due to the recent advances of MEMS (micro electro mechanical system) technology, it has become feasible to manufacture distributed micro manipulators at low cost. One such system is an actuator array where hundreds of micro-scaled "cilia" transport and manipulate small objects that rest on them. Macroscopic versions of the actuator array have also been developed and analyzed. Another form of distributed manipulation is derived from a vibrating plate, and teams of mobile robots have been used to herd large objects into desired locations.

There are many fundamental issues involved in distributed manipulation. Since a distributed manipulator has many actuators, distributed control strategies must be considered to effectively manipulate objects. A basic understanding of contact analysis between the actuators and object must also be considered. When each actuator in the array has a sensor, distributed sensing presents some basic research challenges. But even without sensors, distributed manipulators are capable of a wide range of open-loop manipulation tasks. Distributed computation and communication are key issues to enable the successful deployment of distributed manipulators into use. Finally, the trade-off in centralized and de-centralized approaches in all of these algorithms must be investigated.
This book brings together for the first time several key research activities in distributed manipulation from a variety of fields ranging from MEMS to traditional robotics. The second goal of this book is to introduce this new and excited research to the general robotics community and hopefully spark new interests with our colleagues.

The editors are grateful to the authors whose work has made this book possible. We would like to express our appreciation to the many colleagues and "distributed manipulators" for their contributions and discussions of this exciting topic. In particular, Bruce Randall Donald, Noel C. MacDonald, and Ken Goldberg have been an invaluable and continuous source of support and inspiration to Karl Böhringer. Likewise, William Messner, John Luntz, and Eric Rollins have provided Howie Choset with the unique opportunity to work in this field by including him on their research team and then by helping him get up to speed. Also, Joel Burdick, Alfred Rizzi and Dick Wanky have been a continuous source of support and inspiration to Howie Choset.

We would like to thank Howard Moraff, formerly of the National Science Foundation, for all of his suggestions that guided our thinking in this new field. Work on this project by Karl Böhringer was supported by a National Science Foundation Career Award No. ECS-9875367. Howie Choset was also supported by the National Science Foundation Award No. IIS-9872255.

A Distributed, Universal Device for Planar Parts Feeding: Unique Part Orientation in Programmable Force Fields

Karl F. Böhringer, University of Washington
Bruce R. Donald, Dartmouth College
Lydia E. Kavraki, Florent Lamiraux, Rice University

Abstract:

Programmable vector fields are an abstraction to represent a new class of devices for distributed, non-prehensile manipulation for applications in parts feeding, sorting, positioning, and assembly. Unlike robot grippers, conveyor belts, or vibratory bowl feeders, these devices generate force vector fields in which the parts move until they may reach a stable equilibrium pose.
Recent research in the theory of programmable vector fields has yielded open-loop strategies to uniquely position, orient, and sort parts. These strategies typically consist of several fields that have to be employed in sequence to achieve a desired final pose. The length of the sequence depends on the complexity of the part.
In this paper, we show that unique part poses can be achieved with just one field. First, we exhibit a single field that positions and orients any laminar part (with the exception of certain symmetric parts) into two stable equilibrium poses. Then we show that for any laminar part there exists a field in which the part reaches a unique stable equilibrium pose (again with the exception of symmetric parts). Besides giving an optimal upper bound for unique parts positioning and orientation, our work gives further evidence that programmable vector fields are a powerful tool for parts manipulation.
Our second result also leads to the design of "universal parts feeders," proving an earlier conjecture about their existence. We argue that universal parts feeders are relatively easy to build, and we report on extensive simulation results which indicate that these devices may work very well in practice. We believe that the results in this paper could be the basis for a new generation of efficient, open-loop, parallel parts feeders.

Experiments in Constrained Prehensile Manipulation: Distributed Manipulation with Ropes

Bruce R. Donald, Larry Gariepy, Daniela Rus, Dartmouth College

Abstract:

This paper describes our experiments with a distributed manipulation system. We study a system in which multiple robots cooperate to move large objects such as furniture and boxes using a constrained prehensile manipulation mode, by wrapping ropes around them. The system consists of three manipulation skills: tieing ropes around objects, affecting translations using a flossing manipulation gait, and affecting rotations using a ratcheting manipulation gait. We present experimental data and discuss the non-holonomic nature of this system.

Simultaneous Planar Transport of Multiple Objects on Individual Trajectories Using Friction Forces

Peter U. Frei, ETH Zürich
Markus Wiesendanger, EPF Lausanne
Roland Büchi, ETH Zürich
Lorenz Ruf, SIG Pack Systems AG

Abstract:

Friction forces can be used to move objects along a planar horizontal surface by vibrating the surface with two degrees of freedom. The method presented here uses a combination of horizontal and vertical oscillations to produce a non-zero resultant friction force. Using this principle, objects can be moved along any horizontal direction with variable speed.
If many small surface elements are placed next to each other to form an array, moving multiple objects independently on individual trajectories at the same time is possible. The two-dimensional oscillations of the surface elements can be separated into a common horizontal motion of all surfaces and an individual vertical vibration. Therefore, one common actuator is used to create the horizontal motion of the surfaces and for the vertical motion only one actuator with one degree of freedom is needed for each array element.
Both open loop and closed loop control strategies are possible. The use of appropriate force fields allow the positioning of objects without position feedback. The trajectories achievable are however limited by the dimensions and the geometry of the array cells. Using position feedback, multiple objects can travel on arbitrary trajectories as long as they are not too close to each other.

Hybrid Approach of Centralized Control and Distributed Control for Flexible Transfer System

Toshio Fukuda, Isao Takagawa, Kosuke Sekiyama, Yasuhisa Hasegawa, Nagoya University

Abstract:

The exible transfer system (FTS) is a self-organizing manufacturing system composed of autonomous robotic modules, which transfer a palette carrying machining parts. The central issue is realization of both higher effciency and flexibility to cope with environmental change, such as a sudden change of machining plan or breakdowns of the modules. Through the self-organization of a multi-layered strategic vector field corresponding to a task, the FTS can generate quasi-optimal transfer path with Learning Automata. Also, the optimal planning is attempted by use of Genetic Algorithms, which bases on the global information on the system. In this paper, we propose a hybridization method between the distributed and centralized approaches. Simulation is conducted to evaluate the basic system performance and the results show the effectiveness.

Autonomous Distributed System for Cooperative Micromanipulation

Satoshi Konishi, Yoshio Mita, Ritsumeikan University
Hiroyuki Fujita, University of Tokyo

Abstract:

A micromanipulation system in a plane is described as an example of autonomous distributed micromachines (ADM). We propose and demonstrate a method to obtain macroscopic work out of distributed microactuators fabricated by IC-compatible micromachining. We have previously developed several kinds of microactuator arrays for micromanipulation tasks.
This paper presents a practical design of ADM composed of many micro cells integrated with actuators, sensors and circuits that can be fabricated by IC-compatible micromachining. ADM consist of many micro cells that are smart enough to control their own motions and to cooperate with each other. The simple and small motion of an individual micro cell is coordinated in order to perform a system task. The behavior of the manipulation system is examined by demonstrations on a computer model and experiments on real control circuits.

Discreteness Issues in Actuator Arrays

Jonathan E. Luntz, William Messner, Howie Choset, Carnegie-Mellon University

Abstract:

An actuator array is a form of distributed manipulation where an object being transported and manipulated rests on a large number of stationary supporting actuators. The authors have developed a macroscopic actuator array consisting of many motorized wheels. The analysis of such an array as opposed to a MEMS array requires the explicit modeling of the discreteness in the system, including the set of supports, distribution of weight, and generation of traction forces. Using an open-loop wheel velocity field, discreteness causes undesirable behavior such as unstable rotational equilibria, suggesting the use of object feedback. Discrete distributed control algorithms are derived by inverting the dynamics of manipulation.

Design and Simulation of a Miniature Mobile Parts Feeder

Arthur E. Quaid, Ralph L. Hollis, Carnegie-Mellon University

Abstract:

In this work, a miniature mobile vibratory parts feeder is presented. This feeder is designed to reorient, singulate, and position parts by exploiting the horizontal vibration capabilities of a recently developed closed-loop planar motor. The actuators used for generating the vibrations are also capable of large planar motions, allowing the feeder to present parts to multiple overhead robots. It is designed with a minimum of part-specific features, allowing different parts to be fed with only software changes. The basic feed principle is presented and demonstrated experimentally. Although a complete prototype has not yet been fabricated, a model for the motion of parts on the complete feeder is derived and simulation results are presented that indicate successful operation.

Building a Universal Planar Manipulator

Dan Reznik, Emil Moshkovich, John Canny, UC Berkeley

Abstract:

Distributed manipulation devices make use of a large number of actuators, organized in array fashion, to manipulate a small number of parts. Inspired by minimalism we look at a complementary question: can a device with few degrees of actuation freedom be used to flexibly manipulate a large number of parts? In previous publications we have shown that a single horizontally-vibrating plate is just such a device. This suggests that actuator count can be traded for control complexity. In this paper we review our theory of minimalist manipulation and describe implementation solutions towards a working prototype.

Distributed Agent Programming in the Architecture for Agile Assembly

Alfred A. Rizzi, Jay Gowdy, Carnegie-Mellon University

Abstract:

The goal of the Architecture for Agile Assembly (AAA) is to enable rapid deployment and reconfiguration of automated assembly systems through the use of cooperating, modular, and robust robotic agents. The programs for these agents operate in a completely distributed fashion and must efficiently specify precision cooperative behavior. To allow this, the structure of agent programs is carefully designed to enable the automatic encapsulation of the information necessary for execution when a program is down-loaded to a physical agent. The programming model supports the compact specification and robust execution of potentially complex and fragile cooperative behaviors by making use of ordered sets of control strategies which allow a real-time hybrid control system to automatically sequence their execution. This abstraction allows the agent program to describe the high-level semantic behavior of the agent while relying on a set of formally correct control strategies to properly execute and sequence the necessary continuous behaviors.

CMOS Integrated Organic Ciliary Actuator Arrays for General-Purpose Micromanipulation Tasks

John W. Suh, Stanford University / Xerox PARC
Bruce Darling, Karl F. Böhringer, University of Washington
Bruce R. Donald, Dartmouth College
Henry Baltes, ETH Zürich
Gregory T. A. Kovacs, Stanford University

Abstract:

The first micromachined bimorph organic ciliary array with on-chip CMOS circuitry is presented. This device is composed of an 8 x 8 array of cells each having four orthogonally oriented actuators in an overall die size of 9.4mm x 9.4mm. The polyimide based actuators were fabricated directly above the selection and drive circuitry. Selection and activation of actuators in this array shows that integration was successful.
The array was programmed to perform several kinds of manipulation tasks, including linear translation, diagonal motion, as well as vector field operations such as squeeze field and radial field orienting and centering. Preliminary experiments were also performed with the first implementation of a "universal field" that uniquely positions and orients any non-symmetric part without programming or sensor feedback. All tasks were demonstrated using thin silicon dice of about 3mm x 3mm x 0.5mm size as the object being moved.

Distributed Actuation Devices Using Soft-Gel Actuators

Satoshi Tadokoro, Satoshi Fuji, Toshi Takamuri, Keisuke Oguro, Kobe University

Abstract:

ICPF actuator is an electro-active polymer gel actuator. Distinguishing points of this new material are low driving voltage (1.5 V), high-speed response (> 100 Hz), softness (E = 2.2 x 10^8 Pa), and its ability to function under wet conditions. This paper introduces distributed actuation devices using the ICPF actuators. EFD is an element which drives objects by elliptic motion. A system with a number of EFD cilia which cooperatively drive or transfer objects was studied. A face actuator which freely changes its shape was developed using a pattern plating method. It has distributed ICPF actuators having some degrees of freedom (DOF). Soft manipulation is an application of this device.

Two Approaches to Distributed Manipulation

Mark Yim, Jim Reich, Andrew A. Berlin, Xerox PARC

Abstract:

Two radically different approaches to distributed manipulation are reviewed. They each address scalability and manufacturing issues while producing forces sufficient to move macro-scale objects in different ways. The airjet system achieves scalability and manufacturability through macro-scale planar batch fabrication technology while PolyBot is modular, enabling mass production. Where PolyBot is suited to couple to non-planar objects through variable out-of-plane motion of the cilia, airjets are optimized for manipulation of planar objects with delicate surface features. The designs of both systems are well suited to hierarchical computation and communication to enable scalability without an explosion in the resource requirements.

Back to main page...


© Karl F. Böhringer, Department of Electrical Engineering, Box 352500, Seattle, WA 98195-2500, USA