“In-hand manipulation” is the ability to reposition an object in the hand, for example when adjusting the grasp of a hammer before hammering a nail. The common approach to in-hand manipulation with robotic hands, known as dexterous manipulation [1], is to hold an object within the fingertips of the hand and wiggle the fingers, or walk them along the object's surface. Dexterous manipulation, however, is just one of the many techniques available to the robot. The robot can also roll the object in the hand by using gravity, or adjust the object's pose by pressing it against a surface, or if fast enough, it can even toss the object in the air and catch it in a different pose. All these techniques have one thing in common: they rely on resources extrinsic to the hand, either gravity, external contacts or dynamic arm motions. We refer to them as “extrinsic dexterity”. In this paper we study extrinsic dexterity in the context of regrasp operations, for example when switching from a power to a precision grasp, and we demonstrate that even simple grippers are capable of ample in-hand manipulation. We develop twelve regrasp actions, all open-loop and hand-scripted, and evaluate their effectiveness with over 1200 trials of regrasps and sequences of regrasps, for three different objects (see video [2]). The long-term goal of this work is to develop a general repertoire of these behaviors, and to understand how such a repertoire might eventually constitute a general-purpose in-hand manipulation capability.

This video presents the application of Extrinsic Dexterity to change the pose of an object in the hand, i.e., to regrasp the object.

We explore the control synthesis problem for a robot dynamically manipulating an object in the presence of multiple frictional contacts. Contacts occur both between the object and the robot, and between the object and the environment. Two sets of constraints govern the evolution of the system — contact velocity constraints that prevent separation and cause rolling, and, contact force constraints that arise from Coulomb friction. We combine the constraints in the space of contact accelerations, obtaining bounds on the robot acceleration as a function of the system state. We solve the motion planning problem by providing a feasible path for the system and generating the controls and the system trajectory by time-scaling the feasible path. We provide examples that illustrate the merits and limitations of our technique and discuss some of the open problems.

This paper addresses the planning and control of dynamic contact manipulation. In an earlier paper (Srinivasa et al., 2005), we derived a constraint on the robot joint accelerations that needed to be satisfied to obtain a desired contact mode and a desired dynamic motion of the object. We proposed a technique for trajectory planning which involved planning a path in the system configuration space followed by time-scaling the path to satisfy dynamic constraints. This paper tackles a problem where only a small set of paths can be time-scaled to satisfy the constraints. We note that the dynamic constraints depend only on a subspace of the system state space. Projecting the dynamics and the constraints onto the subspace allows us to compute an analytical solution for the trajectory generation problem. We generate controllable simulations by allowing the user to control the system in the space orthogonal to the projection. We also demonstrate the construction of feedback controllers using dynamic programming.

Computer Science Department

We explore the control of a nonholonomic robot subject to additional constraints on the state variables. In our problem, the user specifies the path of a subset of the state variables (the task freedoms X/sub P/), i.e. a curve X/sub P/(s) where s/spl isin/[0,1] is a parameterization that the user chooses. We control the trajectory of the task freedoms by specifying a bilateral time-scaling s(t) which assigns a point on the path for each time t. The time-scaling is termed bilateral because there is no restriction on s(t), the task freedoms are allowed to move backwards along the path. We design a controller that satisfies the user directive and controls the remaining state variables (the shape freedoms X/sub R/) to satisfy the constraints. Furthermore, we attempt to reduce the number of control switchings, as these result in relatively large errors in our system state. If a constraint is close to being violated (at a switchings point), we back up X/sub P/ along the path for a small time interval and move X/sub S/ to an open region. We show that there are a finite number of switching points for arbitrary task freedom paths. We implement our control scheme on the Mobipulator and discuss a generalization to arbitrary systems satisfying similar properties.

Motion sketching is an approach for creating realistic rigid-body motion. In this approach, an animator sketches how objects should move and the system computes a physically plausible motion that best fits the sketch. The sketch is specified with a mouse-based interface or with hand-gestures, which move instrumented objects in the real world to act out the desired behaviors. The sketches may be imprecise, may be physically infeasible, or may have incorrect timing. A multiple-shooting optimization estimates the parameters of a rigid-body simulation needed to simulate an animation that matches the sketch with physically plausible timing and motion. This technique applies to physical simulations of multiple colliding rigid bodies possibly connected with joints in a tree (open-loop) topology.

We present new results on reconstruction of the shape and motion of an unknown object using tactile sensors without requiring object immobilization. A robot manipulates the object with two flat palms covered with tactile sensors. We model the full dynamics and prove local observability of the shape, motion and center of mass of the object based on the motion of the contact points as measured by the tactile sensors.

Orienting parts that measure only a few micrometers in diameter introduces several challenges that need not be considered at the macro-scale. First, there are several kinds of sticking effects due to Van der Waals forces and static electricity which complicate hand-off motions and release of a part. Second, the degrees of freedom of micromanipulators are limited. The paper proposes a pair of manipulation primitives and a complete algorithm that addresses these challenges. We show that a sequence of these two manipulation primitives can uniquely orient any asymmetric part while maintaining contact without sensing. This allows us to apply the same plan to many (identical) parts simultaneously. For asymmetric parts we can find a plan of length O(n) in O(n) time that orients the part, where n is the number of vertices.

This paper summarizes ongoing work with a mo- bile manipulator (Mobipulator ). We describe the sys- tem architecture of the latest version of the robot, a hierarchy of robot motion commands (the Mobipula- tion library) that can be snapped together to generate complicated paths easily, a configuration space plan- ner that plans wheel motions to manipulate paper, and a visual servoing system to monitor and correct errors in robot motion.

We present a new method to reconstruct the shape of an unknown object using tactile sensors without requiring object immobilization. Instead, the robot manipulates the object without prehension. The robot infers the shape, motion and the center of mass of the object based on the motion of the contact points as measured by tactile sensors. Our analysis is supported by simulation and experimental results.

For assembly tasks parts often have to be oriented before they can be put in an assembly. We present a new approach to parts orienting through the manipulation of pose distributions. Through dynamic simulation we can determine the pose distribution for an object being dropped from an arbitrary height on an arbitrary surface. By varying the drop height and the shape of the support surface we can find the initial conditions that will result in a pose distribution with minimal entropy. We attempt to uniquely orient a part with high probability just by varying the initial conditions. We derive a condition on the pose and velocity of an object in contact with a sloped surface that will allow us to quickly determine the final resting configuration of the object. This condition can then be used to quickly compute the pose distribution. We also show simulation and experimental results which confirm that our dynamic simulator can be used to find the true pose distribution of an object.

Physical simulation of dynamic objects has become commonplace in computer graphics because it produces highly realistic animations. In this paradigm the animator provides few physical parameters such as the objects' initial positions and velocities, and the simulator automatically generates realistic motions. The resulting motion, however, is difficult to control because even a small adjustment of the input parameters can drastically affect the subsequent motion. Furthermore, the animator often wishes to change the end-result of the motion instead of the initial physical parameters.

This paper describes a mobile manipulator that uses its wheels for manipulation as well as locomotion. This robot, named the mobipulator, looks like a small car with four independently powered wheels, none of them steered. It is designed to manipulate paper and other objects on the surface of a desk. The wheels are used for locomotion or for manipulation, switching functions dynamically as the task demands. So far we have preliminary demonstrations of a variety of motions, and performance data for the task of moving a sheet of paper in a square while maintaining constant orientation.

Investigates how to "observe" a planar object being pushed by a finger. The pushing is governed by a nonlinear system that relates through contact the object pose and motion to the finger motion. Nonlinear observability theory is employed to show that the contact information is often sufficient for the finger to determine not only the pose but also the motion of the object. Therefore a sensing strategy can be realized as an observer of the nonlinear dynamical system, which is subsequently introduced. The observer based on the result of Gauthier et al. (1992), has its "gain" determined by the solution of a Lyapunov-like equation. Simulations have been done to demonstrate the feasibility of the observer. A sensor has been implemented using strain gauges and mounted on an Adept robot with which preliminary experiments have been conducted from a general perspective, this work presents an approach for acquiring geometric and dynamical information about a task from a small amount of tactile data, with the application of nonlinear observability theory.

The authors previously (1996) presented a model of nonprehensile manipulation, using two one-degree-of-freedom palms. Under the assumptions of low friction and quasistatic motion, we developed a planning method for part reorientation with our model, starting from a known initial state. Our method finds feasible paths through the space of equivalent state configurations of the object in the palms, without requiring that the palms maintain stable support of the object over the entire path. We have shown that such a device can reliably orient parts in the plane. In this paper we extend our method to the case of reorienting a part to a desired goal from an unknown initial state. In addition to the all sliding contacts case which the model is based upon, we look at extensions to rolling contacts. We include the results of tests with example plans.

In the absence of vision, grasping an object often relies on tactile feedback from the fingertips. Before force closure is formed, where on the object a fingertip touches can usually be felt from the motion of contact on the fingertip during a small amount of pushing. In this paper the authors investigate the first stage of such "blind" grasping. More specifically the authors study the problem of determining the pose of a known planar object by pushing. Assuming sliding friction in the plane, a dynamic analysis of pushing results in a numerical algorithm that compares the object pose from three instantaneous contact positions on a fingertip. Simulations and experiments (with an Adept robot) have been conducted to demonstrate the sensing feasibility. Inspired by the way a human hand grasps, this work can be viewed as a primitive step in exploring interactive sensing in grasping tasks.

Manipulation without prehension is a natural way of handling objects for both humans and machines. Nonprehensile operations are appropriate when complete constraint over the object to be manipulated is either undesirable or impractical, but some control over the object is desired over its entire trajectory in order to bring the object reliably to a desired final state. Research to date has explored only a small portion of this class. We are interested in controlling the shape of the constraint surfaces so that constraint and external forces naturally attract the system to the desired state, even if the object momentarily loses stability during the motion. We present a preliminary analysis of the nonprehensile orientation of planar objects by two low friction palms joined at a central hinge. These palms support an object in a gravitational field, without grasping or gripping. We determine connected regions of stable states of the object, and give a method of planning part orientation based on a graph search over these regions, allowing nonequilibrium transitions between them. We conclude with the results of simulations and tests of an example plan.

Explores the use of stable support in robotic manipulation. Stable support describes any contact configuration which balances a known applied force and is stable with respect to small perturbations in the handled object's pose. Specifically, the authors address the problem of manipulating a planar polygonal object in a fixed gravitational field, stably supported by two frictionless contacts. Representing each contact configuration as a force focus point defined by the intersection of the lines of action of the contact forces, the authors derive geometric regions of permissible force focus points for contacts on each pair of polygon edges. Each permissible force focus point maps to a unique configuration of the object and contacts in stable equilibrium. In turn, paths in the space of permissible force focus points map to real space motions of the contact points which induce quasi-static rotations of the object. The authors also present two graph search based strategies for planning hand-offs between pairs of contacts, thus enabling larger rotations than can be executed by a single contact pair. Finally, the authors describe an implementation of one of these planners.

Industrial assembly involves sensing the pose (orientation and position) of a part. Efficient and reliable sensing strategies can be developed for an assembly task if the shape of the part is known in advance. In this paper the authors investigate the problem of determining the pose of a convex n-gon from a set of m supporting cones, i.e., cones with both sides supporting the polygon. An algorithm with running time O(nm) which almost always reduces to O(n+m log n) is presented to solve for all possible poses of the polygon. As a consequence, the polygon inscription problem of finding all possible poses for a convex n-gon inscribed in another convex m-gon, can be solved within the same asymptotic time bound. The authors prove that the number of possible poses cannot exceed 6n, given m/spl ges/2 supporting cones with distinct vertices. Experiments demonstrate that two supporting cones are sufficient to determine the real pose of the n-gon in most cases. The authors' results imply that sensing in practice can be carried out by obtaining viewing angles of a planar part at multiple exterior sites in the plane. As a conclusion, the authors generalize this and other sensing methods into a scheme named sensing by inscription.>

While most previous work in planning manipulation tasks relies on the assumption of quasi-static conditions, there can be situations where the quasi-static assumption may not hold, and the assumptions about the environment must be relaxed. This is true, for example, in a situation where objects are making and breaking contact at high enough velocities that contact dynamics play a significant effect in the motion of the colliding objects. There has been some work studying model of collision, in particular, for the design and analysis of systems with intermittent constraints, and for the design of juggling robots. Our work extends previous studies in planar juggling to the case of a polygonal object, using the model of rigid body impulsive collision. Simulations verify the results of a linearized analysis.>

This article summarizes ideas on representing translational friction and computing reaction forces in configuration space. The configuration space of an object is the parameter space describing the object's degrees of freedom. Kinematic constraints imposed on the object by obstacles in the environment may be represented as hypersurfaces in the configuration space. Physical and dynamical constraints such as those imposed by friction may be represented as constraints in the generalized force space that accompanies the configuration space. This article develops a representation of friction in this generalized force space that is analogous to the classical friction cone in real space.

The problem of orienting a part resting on a table by tilting the table is considered. The initial orientation of the part is assumed to be completely unknown. The objective is to tilt the table in a manner that reduces the uncertainty in the part's orientation. This work focuses on three-dimensional polyhedral parts, with finite friction between the parts and the table, and for which all transitions between different face-table contacts may be regarded as rotations across edges. A planner that determines a sequence of tilting operations designed to minimize the uncertainty in the part's orientation is proposed. The planner runs in time O(n/sup 4/), where n is the number of faces of the polyhedron. The planner produces a sequence of O(n) distinct tilts.>

A programmable parts feeder, a mechanism that can be reprogrammed to handle differently shaped parts, is discussed. The authors present a planning algorithm that accepts an n-sided polygonal part as input and, in time O(n/sup 2/), generates a program (plan) for the feeder that maximizes expected feedrate. They have implemented the planner and verified some of the resulting plans in the laboratory. This work illustrates a stochastic framework for manipulation planning.>

Provides a geometric representation of friction for a rigid planar part with two translational and one rotational degrees of freedom. The construction of a generalized friction cone is accomplished by imbedding into the part's configuration space the constraints that define the classical friction cone in real space. The resulting representation provides a simple computational method for determining the possible motions of a part subjected to an applied force and torque. The representation has been used both for simulating part motions and for planning assembly operations. Generalizations to the six-dimensional configuration space of a three-dimensional part are possible.>

The use of motion strategies to eliminate uncertainty, without the use of sensors, is considered. The approach is demonstrated within the context of a simple method to orient planar objects. A randomly oriented object is dropped into a tray. When the tray is tilted, the object can slide into walls, along walls, and into corners, sometimes with the effect of reducing the number of possible orientations. For some objects a sequence of tilting operations exists that leaves the object's orientation completely determined. An automatic planner is described that constructs such a tilting program, using a simple model of the mechanics of sliding. The planner has been implemented, the resulting programs have been executed using a tray attached to an industrial manipulator, and sometimes the programs work. The authors explore the issue of sensorless manipulation, tray tilting in particular, within the context of a formal framework described by T. Lozano-Perez, M.T. Mason, and Rolf Taylor (1984). It is observed that sensorless motion strategies perform conditional actions using mechanical decisions in place of environmental inquiries.>

This paper explores the motion planning problem for multiple moving objects. The approach taken consists of assigning priorities to the objects, then planning motions one object at a time. For each moving object, the planner constructs a configuration space-time that represents the time-varying constraints imposed on the moving object by the other moving and stationary objects. The planner represents this space-time approximately, using two-dimensional slices. The space-time is then searched for a collision-free path. The paper demonstrates this approach in two domains. One domain consists of translating planar objects; the other domain consists of two-link planar articulated arms.

An autonomous robotic manipulator can reduce uncertainty in the locations of objects in either of two ways: by sensing, or by motion strategies. This paper explores the use of motion strategies to eliminate uncertainty, without the use of sensors. The approach is demonstrated within the context of a simple method to orient planar objects. A randomly oriented object is dropped into a tray. When the tray is tilted, the object can slide into walls, along walls, and into corners, sometimes with the effect of reducing the number of possible orientations. For some objects a sequence of tilting operations exists that leaves the object's orientation completely determined. The paper describes an automatic planner that constructs such a tilting program, using a simple model of the mechanics of sliding. The planner has been implemented, the resulting programs have been executed using a tray attached to an industrial manipulator, and sometimes the programs work. The paper also explores the issue of sensorless manipulation, tray-tilting in particular, within the context of a formal framework first described by Lozano-Pérez, Mason, and Taylor [1984]. It is observed that sensorless motion strategies perform conditional actions using mechanical decisions in place of environmental inquiries.

This paper outlines a method for planning motions in the presence of uncertainty. Tasks are modelled as geometrical goals in configuration space. The planning process consists of determining regions from which particular motions are guaranteed to successfully reach a desired goal. An algorithm is presented for backprojecting from desired goal states. The backprojection regions are computed by erecting constraints that geometrically capture the uncertainty in motion.