Lillian Y. Chang and Nancy S. Pollard

This research is supported by the National Science Foundation (CCF-0343161, IIS-0326322, ECS-0325383, CNS-0423546, and CCF-0702443). L. Y. Chang has received support from a National Science Foundation Graduate Research Fellowship and a NASA Harriet G. Jenkins Pre-Doctoral Fellowship.

[Back to home page]

Preparatory manipulation of movable objects (in progress)

Selection criteria for preparatory object rotation in manual lifting actions.

We extend our previous study of pre-grasp interaction as a human manipulation strategy. In this work, we investigate whether pre-grasp rotation could be affected by the posture-specific lifting capability, as well as task difficulty factors of object mass and required task precision.

Participants lifted a canister by its handle while balancing a ball on the lid. Experiment 1 allowed object rotation prior to lifting. A lifting comfort zone was measured by the variability in object orientation at lift; its size depended on the object mass and required task precision. The amount of pre-lift rotation correlated with the resulting change in lifting capability, as measured for different object orientations. Experiment 2 required direct grasping, without preparatory rotation. Task completion time and success rate decreased, and initial object orientation affected pre-lift time. Results suggest that lifting from the comfort zone produces more robust performance at a cost of slower completion; moreover, physical rotation could be replaced by mental planning when direct grasping is enforced.

Citation

Lillian Y. Chang, Roberta L. Klatzky, and Nancy S. Pollard. Selection criteria for preparatory object rotation in manual lifting actions. Journal of Motor Behavior, in press, July 2009.

Pre-grasp interactions in natural manipulation actions

Preparatory rotation is only one example of a pre-grasp interaction strategy. In the video survey of human hand activity, we filmed people performing manipulation tasks in natural settings such as the home or place of occupation. We found that there is indeed a broad class of pre-grasp interactions where the object is not grasped directly from its presented placement in the environment. Our framework describes the survey examples according to two main aspects of the pre-grasp interaction. The first aspect is the type of object re-configuration resulting from the interaction. The second aspect is the underlying intent of the interaction to improve the posture quality or grasp quality of the manipulation action.

Citation

Lillian Y. Chang and Nancy S. Pollard. Video survey of pre-grasp interactions in natural hand activities. Robotics: Science and Systems (RSS) 2009 Workshop: Understanding the Human Hand for Advancing Robotic Manipulation, University of Washington, Seattle, USA, June 28 2009.

Poster

Workshop poster [1.6MB]

Grasping with an anthropomorphic robot manipulator

We investigate how the pre-rotation strategy observed in humans could be used to improve robot manipulator performance. Taking advantage of object movability may increase robustness by making good grasps possible for a greater variety of initial object configurations. The expense of tuning control parameters can be reduced because a single well-tuned canonical grasp can be robustly applied to multiple conditions. In addition, mimicking human manipulation could lead to more natural-looking and socially-acceptable motions for an anthropomorphic robot interacting in a human living space.

We designed a preparatory rotation strategy for an anthropomorphic robot manipulator as a method of extending the capture region of a specific grasp prototype. The strategy was implemented as a sequence of two open-loop actions mimicking the human motion: a preparatory rotation action followed by a grasping action. The grasping action alone can only successfully lift the object from a 45-degree region of initial orientations (4 of 24 tested conditions). Our empirical evaluation of the robot preparatory rotation shows that even using a simple open-loop rotation action enables the reuse of the grasping action for a 360-degree capture region of initial object orientations (24 of 24 tested conditions).

Citation

Lillian Y. Chang, Garth J. Zeglin, and Nancy S. Pollard. Preparatory object rotation as a human-inspired grasping strategy. IEEE International Conference on Humanoid Robots (Humanoids 2008), December 2008. 527-534.

Video

robot pan rotation [5MB]

Human lifting strategies under task constraints

We investigated how human manipulation strategies change in response to task constraints. Human subjects were captured lifting a set of heavy, handled objects presented in a variety of initial conditions. When only instructed to lift the object with their right hand, humans naturally re-oriented the object handle to a preferred configuration prior to lifting the object from the surface. This results in similar body poses at object liftoff for all of the presented object orientations. In contrast, when the subject is not permitted to pre-rotate the object, the poses at object liftoff differ more between the multiple object orientations presented. The lifting task is more difficult because the constraint requires unnatural body poses to successfully lift the object. In the unconstrained lifting tasks, the amount of rotation depends on the initial configuration of the object. This pre-rotation strategy allows humans to change the object such that it can be grasped from the canonical lifiting pose, rather than being constrained to the different object configurations presented. The canonical lifting pose may be preferable to the unnatural poses in the constrained tasks because of lower joint torques, increased kinematic reachability, or increased stability of stance and grasp.

Citation

Lillian Y. Chang, Garth J. Zeglin, and Nancy S. Pollard. Preparatory object rotation as a human-inspired grasping strategy. To appear at IEEE International Conference on Humanoid Robots (Humanoids 2008), Daejeon, Korea, December 2008.

Lillian Y. Chang and Nancy S. Pollard. On preparatory object rotation to adjust handle orientation for grasping Tech. Report CMU-RI-TR-08-10, Robotics Institute, Carnegie Mellon University, April, 2008.

Tool acquistion from a work surface

Direct whole-hand grasps of tools from a work surface are not possible because the object first needs to be lifted before the fingers can envelope the handle. We measured the complex manipulation humans use to pick up tools from a surface. Initial contact with the object starts with a precision fingertip grasp. As the tool is lifted from the surface, the fingers manipulate the tool in-hand to transition from the fingertip grasp to the whole-hand grasp. After adjusting the tool handle so that it is aligned with the palm's natural oblique grasping axis, the final whole-hand power grasp is possible.

Learning grasping by demonstration

Reduced feature set for hand surface markers

Although the human hand is a complex biomechanical system, only a small set of features may be necessary for observation learning of functional grasp classes. We explore how to methodically select a minimal set of hand pose features from optical marker data for grasp recognition. Supervised feature selection is used to determine a reduced feature set of surface marker locations on the hand that is appropriate for grasp classification of individual hand poses. Classifiers trained on the reduced feature set of five markers retain at least 92% of the prediction accuracy of classifiers trained on a full feature set of thirty markers. The reduced model also generalizes better to new subjects. The dramatic reduction of the marker set size and the success of a linear classifier from local marker coordinates recommend optical marker techniques as a practical alternative to data glove methods for observation learning of grasping.

Citation

Lillian Y. Chang, Nancy S. Pollard, Tom M. Mitchell, and Eric P. Xing. Feature Selection for Grasp Recognition from Optical Markers. Proceedings of the 2007 IEEE/RSJ Intl. Conference on Intelligent Robots and Systems (IROS 2007), October, 2007. 2944-2950.
Fitting joint models to motion capture data

Two axis model for the thumb carpometacarpal joint

The mobility of the thumb carpometacarpal (CMC) joint is critical for functional grasping and manipulation tasks. We present an optimization technique for determining from surface marker measurements a subject-specific kinematic model of the in vivo CMC joint that is suitable for measuring mobility. Our anatomy-based cost metric scores a candidate joint model by the plausibility of the corresponding joint angle values and kinematic parameters rather than only the marker trajectory reconstruction error. The proposed method repeatably determines CMC joint models with anatomically-plausible directions for the two dominant rotational axes and a lesser range of motion (RoM) for the third rotational axis. We formulate a low-dimensional parameterization of the optimization domain by first solving for joint axis orientation variables which then constrain the search for the joint axis location variables. Individual CMC joint models were determined for 24 subjects. The directions of the flexion-extension (FE) axis and adduction-abduction (AA) axis deviated on average by 9 degrees and 22 degrees, respectively, from the mean axis direction. The average RoM for FE, AA, and pronation-supination (PS) joint angles are 76, 43, and 23 degrees for active CMC movement. The mean separation distance between the FE and AA axes was 4.6 mm, and the mean skew angle was 87 degrees from the positive flexion axis to the positive abduction axis.

Citation

Lillian Y. Chang and Nancy S. Pollard. Method for determining kinematic parameters of the in vivo thumb carpometacarpal joint. IEEE Transactions on Biomedical Engineering, 2008, 55 (7): 1897-1906.

Dominant axis of rotation for a hinge joint

A simple method is developed for robustly estimating a fixed dominant axis of rotation (AoR) of anatomical joints from surface marker data. Previous approaches which assume a model of circular marker trajectories use plane-fitting to estimate the direction of the AoR. However, when there is limited joint range of motion and rotation due to a second degree of freedom, minimizing only the planar error can give poor estimates of the AoR direction. Optimizing a cost function which includes the error component within a plane, instead of only the component orthogonal to a plane, leads to improved estimates of the AoR direction for joints which exhibit additional rotational motion from a second degree of freedom. Results from synthetic data validation show the ranges of motion where the new method has lower estimation error compared to plane-fitting techniques. Estimates of the flexion-extension AoR from empirical motion capture data of the knee and index finger joints were also more anatomically plausible than alternative techniques.

Citation

Lillian Y. Chang and Nancy S. Pollard. Robust estimation of dominant axis of rotation. Journal of Biomechanics, 2007. 40 (12): 2707-2715.

Center of rotation for a spherical joint

We present a new direct method for estimating the average center of rotation (CoR). An existing least-squares (LS) solution has been shown by previous works to have reduced accuracy for data with small range of motion (RoM). Alternative methods proposed to improve the CoR estimation use iterative algorithms. However, in this paper we show that with a carefully chosen normalization scheme, constrained least-squares solutions can perform as well as iterative approaches, even for challenging problems with significant noise and small RoM. In particular, enforcing the normalization constraint avoids poor fits near plane singularities that can affect the existing LS method. Our formulation has an exact solution, accounts for multiple markers simultaneously, and does not depend on manually-adjusted parameters. Simulation tests compare the method to four published CoR estimation techniques. The results show that the new approach has the accuracy of the iterative methods as well as the short computation time and repeatability of a least-squares solution. In addition, application of the new method to experimental motion capture data of the thumb carpometacarpal (CMC) joint yielded a more plausible CoR location compared to the previously reported LS solution and required less time than all four alternative techniques.

Citation

Lillian Y. Chang and Nancy S. Pollard. Constrained least-squares optimization for robust estimation of center of rotation. Journal of Biomechanics, 2007, 40 (6): 1392-1400.