NEMS 2006

Program

Click here for a PDF version of the program.

Talks should be 15 minutes, allowing for 5 minutes of questions and discussion.

9:55-10:00 Welcome

10:00-10:20 Computation reuse in statics and dynamics problems for assemblies of rigid bodies
Anne Loomis and Devin Balkom
Dartmouth

PPT slides
The problem of determining the forces among contacting rigid bodies is fundamental to many areas of robotics, including manipulation planning, control, and dynamic simulation. In considering problems of this type over discrete or continuous time, we often encounter a sequence of problems with similar substructure. The primary contribution of our work is the observation that in many cases, common physical structure can be exploited to solve a sequence of related problems more efficiently than if each problem were considered in isolation.
We examine three general problems concerning rigid-body assemblies: dynamic simulation, assembly planning, and assembly stability given limited knowledge of the structure's geometry. Our approach to both dynamic simulation and assembly planning is based on a theoretical result regarding computation reuse in a known method for solving the system dynamics.

10:20-10:40 Time-optimal trajectories for omni-directional vehicles
Paritosh Kavathekar
Dartmouth

PPT slides
Unlike steered cars, an omni-directional vehicle can move in any direction without needing to turn first. In this talk I will describe the geometric structure of time-optimal trajectories for a kinematic model of omni-directional vehicles. A constant endeavor of research is to use minimum resources for achieving maximum output. Time-optimal trajectories achieve this goal for perhaps the most fundamental resource, time. Optimal trajectories are an intrinsic property of any vehicle design. They serve as a benchmark for comparing different planners, and help understand the tradeoffs that a particular design makes. Although the problem of obtaining time-optimal trajectories is easy to state, the analytical optimal trajectories are known for only two classes of vehicles: steered cars and differential drive. We use Pontryagin's maximum principle to show that optimal trajectories for an omni-directional vehicle are well-behaved, contain no more than 18 control switches, and can be divided into four classes. I will also describe some possible extensions of the techniques we use to areas such as mechanism design and stable pushing.

10:40-11:00 2D Subspaces for Sparse Control of High-DOF Robots
Chad Jenkins
Brown University

PDF slides
We have explored the use of dimension reduction for estimating 2D subspaces for controlling high DOF robotic systems. Our aim in this work has been to provide a means for controlling high dimensional systems, such as prosthetic arms, with sparse 2D input with an emphasis towards leveraging current capabilities for decoding EMG, EEG, and neural signals. We have taken a data-driven approach to this problem by using dimension reduction to embed motion performed by humans into 2D. We discuss the use of five dimension reduction methods and their application to motion capture data of human performed power and precision grasps. We additionally discuss how the use of shape descriptors for representing hand pose affects the dimension reduction process. We present results from controlling a physically simulated hand using 2D mouse control and thoughts for how autonomous control can improve control.

11:00-11:20 From Robotic Hands to Human Hands: A Visualization and Simulation Engine for Grasping Research
Peter Allen and Matei Ciocarlie
Columbia University

PPT slides
At Columbia University, we have created GraspIt!, a publicly available simulator to serve as a useful tool for grasping research. In this presentation, we introduce some of the main features of GraspIt!, including computation of numerical grasp quality measures, visualization methods that allow a user to see the weak point of a grasp, automatic grasp planning and dynamic simulation for studying grasp formation. We then present our research focused on building a functional, biomechanically realistic human hand model, integrated with the GraspIt! analysis system. Such a model will enable studies of the functional abilities of the intact and impaired human hand using the rigorous mathematical framework of robotics. We discuss some of the features of the human hand which are important to include in a simulation model, such as correct kinematic parameters, indirect actuation and soft contact areas, and present a method for performing grasp quality computations involving deformable fingertips.

11:20-11:40 Manipulation in Human Environments
Aaron Edsinger and Charlie Kemp
MIT CSAIL

PPT slides
Robots that can work alongside us in our homes and workplaces, assisting in tasks that are important to us, will transform how we live and work. They may extend the time an elderly person can live at home, provide physical assistance to a worker on an assembly line, or simply help us with household chores. Manipulation in human environments is at the heart of all these domains. In this talk we present results for several bimanual manipulation tasks on a humanoid robot. This work supports three main observations about manipulation in human environments. First, the robot's physical embodiment can be leveraged to simplify many tasks. Second, human environments are constrained to match our cognitive and physical abilities. Manipulation tasks in these settings can often be reduced to the perception and control of simple, task-relevant features. Third, many manual tasks can be treated as a collaboration between the robot and a person. During this collaboration, the person will intuitively compensate for the robot's physical and perceptual limitations.

11:40-12:00 Cooperative Payload Transport by Robot Collectives
Venkat Krovi
SUNY Buffalo

PDF slides
Cooperative material-handling by a fleet of decentralized manipulation agents has many applications ranging from hazardous waste removal, material handling on the shop floor, to robot work crews for planetary colonization. Our long-term goal is the development of a theoretical and operational framework to model, analyze, implement and validate cooperative payload transport capabilities in such distributed robot collectives.
Our particular focus is on creation, control and active reconfiguration of marching formations of wheeled mobile robots for cooperative payload transportation tasks. The selection of the underlying physical/informational infrastructure, system architecture, and mechanisms of cooperation creates many alternatives. A systematic framework is therefore critical for evaluation/selection of suitable implementations with quantifiable cooperative-performance benefits and forms the focus of our research activities. In particular, we will present our efforts to develop methodologies for design and optimization of formations for apriori known tasks, adaptation of formations for changing tasks and scalable schemes for control under the common theoretical but computationally tractable framework.

12:00-2:00 Lunch, Demos, Discussion

2:00-2:20 daVinci Code: A Multi-Model Simulation and Analysis Tool for Multi-Body Systems
Steve Berard and Jeff Trinke
RPI

PDF slides
We will describe the design and current capabilities of a software tool we are developing, "daVinci Code", capable of simulating and animating planar systems of bodies experiencing intermittent and steady unilateral contacts. Since different problems require different levels of accuracy, dVC provides user-selectable body types (rigid or locally-compliant), motion models (first-order, quasi-static, dynamic), and time-stepping methods. One can also choose to include friction between the bodies and the plane of motion. To support optimal and robust part design, dVC also supports on-the-fly changes to the geometric and physical parameters of the bodies.

2:20-2:40 Designing Open-Loop Plans for Planar Micro-manipulation
Jon Fink and Dave Cappelleri
UPenn

PDF slides
We will describe a test-bed for planar micro manipulation tasks and a framework for planning based on quasi-static models of mechanical systems with frictional contacts. We show how planar peg-in-the-hole assembly tasks can be designed using randomized motion planning techniques with Mason's models for quasi-static manipulation. Finally, we present simulation and experimental results in support of our methodology.

2:40-3:00 Coordinating Droplets in Digital Microfluidic Systems
Eric Griffith and Srinivas Akella
RPI

PDF slides
Digital Microfluidic Systems (DMFS) are a newer alternative to continuous flow microfluidics. In a typical DMFS, discrete droplets of fluids are moved around planar arrays of electrodes. A variety of biochemical reactions and analyses can be performed with such a system by moving droplets around, mixing droplets together, and splitting droplets apart. This talk presents an overview of our approach to tackling the coordination problem. We reduce the complexity of the problem by defining specific areas on the array where mixing and splitting can occur and specific paths for the droplets to follow between and around them. We then decide which of these locations to send droplets to for mixing and splitting and attempt to efficiently route them there. Complexities arise when droplets enter the system too quickly or when dealing with hardware limitations on the control of electrodes.

3:00-3:20 Design and Implementation of a Dynamically Balancing Mobile Manipulator
Patrick Deegan, Bryan Thibodeau, and Rod Grupen
University of Massachusetts Amherst

PPT slides
Whole-body postural control allows bi-manual mobile manipulators to create large interaction forces. To explore this problem domain, we have set out to create an embodied, dynamically balancing robot and to contribute a software architecture for exploiting dynamics in mobile manipulation tasks. A significant component of this work is the design and construction of a new mobile manipulator, the uBot-4. First in the series, uBot-1 was a statically stable, differential drive robot with a small footprint for team exploration in maze-like environments. Balancing on two wheels, uBot-2 demonstrated high mobility and the ability to support large weights above its center of mass. Further enhancing these abilities, uBot-3, featured a kinematic prototype for two arms. Finally, uBot-4 integrates a complete redesign with a rotating trunk and four degree of freedom arms. Furthermore, the arm actuators will feature intrinsic force sensing and compliance. By comparing the magnitude of various forces that can be applied to the environment, we have demonstrated some of the advantages of whole body postural control with the new uBot-4 design. The results presented suggest that for pushing, pulling, or carrying types of tasks, using whole body postural control can lead to higher performance by allowing a platform to apply more force to the environment.

3:20-3:40 A Basic Level of Attentional Behavior for Manipulation, Interaction, and Learning
John Sweeney, Shichao Ou, Steve Hart, and Rod Grupen
University of Massachusetts Amherst

PPT slides
This talk provides an overview of our learning and manipulation framework for humanoid robots. It allows quick and efficient learning of complex manipulative and interactive behavior through autonomous exploration. In addition, it provides a way to program the robot via demonstration. The framework initially gives the robot a small number of native control primitives, or reflexes, that the robot can assemble stochastically in a hierarchical fashion. The robot is driven to learn new behavior by an inherent desire to generate new stimuli on its multi-modal receptors. We performed experiments on Dexter, the UMass humanoid, to generate robust and fault-tolerant manipulation behaviors, such as grasping, sorting, pointing and nodding. Teleological programming by demonstration uses the robot's own controllers as filters for interpreting the demonstrator, filters that focus on the goals of actions. Moreover, the robot can use its controllers to predict actions of the demonstrator; this can lead to faster, more efficient demonstrations. We will present results from teleoperation experiments on Dexter that show this approach to programming by demonstration.

3:40-4:00 Elastic Roadmaps: Globally Task-Consistent Motion for Autonomous Mobile Manipulation in Dynamic Environments
Yuandong Yang and Oliver Brock
University of Massachusetts Amherst

PPT slides
The autonomous execution of manipulation tasks in unstructured, dynamic environments requires the consideration of various motion constraints. Any motion performed during the manipulation task has to satisfy constraints imposed by the task itself, but also has to consider kinematic and dynamic limitations of the manipulator, avoid unpredictably moving obstacles, and observe constraints imposed by the global connectivity of the workspace. Furthermore, the unpredictability of unstructured environments requires the continuous incorporation of feedback to reliably satisfy these constraints. We present a novel motion generation approach, called elastic roadmap framework, capable of satisfying all of the motion constraints that arise in autonomous mobile manipulation and their respective feedback requirements. This framework is validated with simulation experiments using a mobile manipulation platform and a stationary manipulator.

4:00-4:05 Closing

9:55-10:00	Welcome
10:00-10:20	Computation reuse in statics and dynamics problems for assemblies of rigid bodies Anne Loomis and Devin Balkom Dartmouth PPT slides The problem of determining the forces among contacting rigid bodies is fundamental to many areas of robotics, including manipulation planning, control, and dynamic simulation. In considering problems of this type over discrete or continuous time, we often encounter a sequence of problems with similar substructure. The primary contribution of our work is the observation that in many cases, common physical structure can be exploited to solve a sequence of related problems more efficiently than if each problem were considered in isolation. We examine three general problems concerning rigid-body assemblies: dynamic simulation, assembly planning, and assembly stability given limited knowledge of the structure's geometry. Our approach to both dynamic simulation and assembly planning is based on a theoretical result regarding computation reuse in a known method for solving the system dynamics.
10:20-10:40	Time-optimal trajectories for omni-directional vehicles Paritosh Kavathekar Dartmouth PPT slides Unlike steered cars, an omni-directional vehicle can move in any direction without needing to turn first. In this talk I will describe the geometric structure of time-optimal trajectories for a kinematic model of omni-directional vehicles. A constant endeavor of research is to use minimum resources for achieving maximum output. Time-optimal trajectories achieve this goal for perhaps the most fundamental resource, time. Optimal trajectories are an intrinsic property of any vehicle design. They serve as a benchmark for comparing different planners, and help understand the tradeoffs that a particular design makes. Although the problem of obtaining time-optimal trajectories is easy to state, the analytical optimal trajectories are known for only two classes of vehicles: steered cars and differential drive. We use Pontryagin's maximum principle to show that optimal trajectories for an omni-directional vehicle are well-behaved, contain no more than 18 control switches, and can be divided into four classes. I will also describe some possible extensions of the techniques we use to areas such as mechanism design and stable pushing.
10:40-11:00	2D Subspaces for Sparse Control of High-DOF Robots Chad Jenkins Brown University PDF slides We have explored the use of dimension reduction for estimating 2D subspaces for controlling high DOF robotic systems. Our aim in this work has been to provide a means for controlling high dimensional systems, such as prosthetic arms, with sparse 2D input with an emphasis towards leveraging current capabilities for decoding EMG, EEG, and neural signals. We have taken a data-driven approach to this problem by using dimension reduction to embed motion performed by humans into 2D. We discuss the use of five dimension reduction methods and their application to motion capture data of human performed power and precision grasps. We additionally discuss how the use of shape descriptors for representing hand pose affects the dimension reduction process. We present results from controlling a physically simulated hand using 2D mouse control and thoughts for how autonomous control can improve control.
11:00-11:20	From Robotic Hands to Human Hands: A Visualization and Simulation Engine for Grasping Research Peter Allen and Matei Ciocarlie Columbia University PPT slides At Columbia University, we have created GraspIt!, a publicly available simulator to serve as a useful tool for grasping research. In this presentation, we introduce some of the main features of GraspIt!, including computation of numerical grasp quality measures, visualization methods that allow a user to see the weak point of a grasp, automatic grasp planning and dynamic simulation for studying grasp formation. We then present our research focused on building a functional, biomechanically realistic human hand model, integrated with the GraspIt! analysis system. Such a model will enable studies of the functional abilities of the intact and impaired human hand using the rigorous mathematical framework of robotics. We discuss some of the features of the human hand which are important to include in a simulation model, such as correct kinematic parameters, indirect actuation and soft contact areas, and present a method for performing grasp quality computations involving deformable fingertips.
11:20-11:40	Manipulation in Human Environments Aaron Edsinger and Charlie Kemp MIT CSAIL PPT slides Robots that can work alongside us in our homes and workplaces, assisting in tasks that are important to us, will transform how we live and work. They may extend the time an elderly person can live at home, provide physical assistance to a worker on an assembly line, or simply help us with household chores. Manipulation in human environments is at the heart of all these domains. In this talk we present results for several bimanual manipulation tasks on a humanoid robot. This work supports three main observations about manipulation in human environments. First, the robot's physical embodiment can be leveraged to simplify many tasks. Second, human environments are constrained to match our cognitive and physical abilities. Manipulation tasks in these settings can often be reduced to the perception and control of simple, task-relevant features. Third, many manual tasks can be treated as a collaboration between the robot and a person. During this collaboration, the person will intuitively compensate for the robot's physical and perceptual limitations.
11:40-12:00	Cooperative Payload Transport by Robot Collectives Venkat Krovi SUNY Buffalo PDF slides Cooperative material-handling by a fleet of decentralized manipulation agents has many applications ranging from hazardous waste removal, material handling on the shop floor, to robot work crews for planetary colonization. Our long-term goal is the development of a theoretical and operational framework to model, analyze, implement and validate cooperative payload transport capabilities in such distributed robot collectives. Our particular focus is on creation, control and active reconfiguration of marching formations of wheeled mobile robots for cooperative payload transportation tasks. The selection of the underlying physical/informational infrastructure, system architecture, and mechanisms of cooperation creates many alternatives. A systematic framework is therefore critical for evaluation/selection of suitable implementations with quantifiable cooperative-performance benefits and forms the focus of our research activities. In particular, we will present our efforts to develop methodologies for design and optimization of formations for apriori known tasks, adaptation of formations for changing tasks and scalable schemes for control under the common theoretical but computationally tractable framework.
12:00-2:00	Lunch, Demos, Discussion
2:00-2:20	daVinci Code: A Multi-Model Simulation and Analysis Tool for Multi-Body Systems Steve Berard and Jeff Trinke RPI PDF slides We will describe the design and current capabilities of a software tool we are developing, "daVinci Code", capable of simulating and animating planar systems of bodies experiencing intermittent and steady unilateral contacts. Since different problems require different levels of accuracy, dVC provides user-selectable body types (rigid or locally-compliant), motion models (first-order, quasi-static, dynamic), and time-stepping methods. One can also choose to include friction between the bodies and the plane of motion. To support optimal and robust part design, dVC also supports on-the-fly changes to the geometric and physical parameters of the bodies.
2:20-2:40	Designing Open-Loop Plans for Planar Micro-manipulation Jon Fink and Dave Cappelleri UPenn PDF slides We will describe a test-bed for planar micro manipulation tasks and a framework for planning based on quasi-static models of mechanical systems with frictional contacts. We show how planar peg-in-the-hole assembly tasks can be designed using randomized motion planning techniques with Mason's models for quasi-static manipulation. Finally, we present simulation and experimental results in support of our methodology.
2:40-3:00	Coordinating Droplets in Digital Microfluidic Systems Eric Griffith and Srinivas Akella RPI PDF slides Digital Microfluidic Systems (DMFS) are a newer alternative to continuous flow microfluidics. In a typical DMFS, discrete droplets of fluids are moved around planar arrays of electrodes. A variety of biochemical reactions and analyses can be performed with such a system by moving droplets around, mixing droplets together, and splitting droplets apart. This talk presents an overview of our approach to tackling the coordination problem. We reduce the complexity of the problem by defining specific areas on the array where mixing and splitting can occur and specific paths for the droplets to follow between and around them. We then decide which of these locations to send droplets to for mixing and splitting and attempt to efficiently route them there. Complexities arise when droplets enter the system too quickly or when dealing with hardware limitations on the control of electrodes.
3:00-3:20	Design and Implementation of a Dynamically Balancing Mobile Manipulator Patrick Deegan, Bryan Thibodeau, and Rod Grupen University of Massachusetts Amherst PPT slides Whole-body postural control allows bi-manual mobile manipulators to create large interaction forces. To explore this problem domain, we have set out to create an embodied, dynamically balancing robot and to contribute a software architecture for exploiting dynamics in mobile manipulation tasks. A significant component of this work is the design and construction of a new mobile manipulator, the uBot-4. First in the series, uBot-1 was a statically stable, differential drive robot with a small footprint for team exploration in maze-like environments. Balancing on two wheels, uBot-2 demonstrated high mobility and the ability to support large weights above its center of mass. Further enhancing these abilities, uBot-3, featured a kinematic prototype for two arms. Finally, uBot-4 integrates a complete redesign with a rotating trunk and four degree of freedom arms. Furthermore, the arm actuators will feature intrinsic force sensing and compliance. By comparing the magnitude of various forces that can be applied to the environment, we have demonstrated some of the advantages of whole body postural control with the new uBot-4 design. The results presented suggest that for pushing, pulling, or carrying types of tasks, using whole body postural control can lead to higher performance by allowing a platform to apply more force to the environment.
3:20-3:40	A Basic Level of Attentional Behavior for Manipulation, Interaction, and Learning John Sweeney, Shichao Ou, Steve Hart, and Rod Grupen University of Massachusetts Amherst PPT slides This talk provides an overview of our learning and manipulation framework for humanoid robots. It allows quick and efficient learning of complex manipulative and interactive behavior through autonomous exploration. In addition, it provides a way to program the robot via demonstration. The framework initially gives the robot a small number of native control primitives, or reflexes, that the robot can assemble stochastically in a hierarchical fashion. The robot is driven to learn new behavior by an inherent desire to generate new stimuli on its multi-modal receptors. We performed experiments on Dexter, the UMass humanoid, to generate robust and fault-tolerant manipulation behaviors, such as grasping, sorting, pointing and nodding. Teleological programming by demonstration uses the robot's own controllers as filters for interpreting the demonstrator, filters that focus on the goals of actions. Moreover, the robot can use its controllers to predict actions of the demonstrator; this can lead to faster, more efficient demonstrations. We will present results from teleoperation experiments on Dexter that show this approach to programming by demonstration.
3:40-4:00	Elastic Roadmaps: Globally Task-Consistent Motion for Autonomous Mobile Manipulation in Dynamic Environments Yuandong Yang and Oliver Brock University of Massachusetts Amherst PPT slides The autonomous execution of manipulation tasks in unstructured, dynamic environments requires the consideration of various motion constraints. Any motion performed during the manipulation task has to satisfy constraints imposed by the task itself, but also has to consider kinematic and dynamic limitations of the manipulator, avoid unpredictably moving obstacles, and observe constraints imposed by the global connectivity of the workspace. Furthermore, the unpredictability of unstructured environments requires the continuous incorporation of feedback to reliably satisfy these constraints. We present a novel motion generation approach, called elastic roadmap framework, capable of satisfying all of the motion constraints that arise in autonomous mobile manipulation and their respective feedback requirements. This framework is validated with simulation experiments using a mobile manipulation platform and a stationary manipulator.
4:00-4:05	Closing