This page summarizes the research projects from my graduate career. Please see the Publications page for a listing of published journal articles and conference papers.

My advisor, Nancy S. Pollard, is a collaborator on all projects. This research is supported by the National Science Foundation (CCF-0343161, IIS-0326322, ECS-0325383, CNS-0423546, and CCF-0702443). I also received support from a National Science Foundation Graduate Research Fellowship and a NASA Harriet G. Jenkins Pre-Doctoral Fellowship.

Pre-grasp interaction of movable objects
My thesis defines a type of manipulation strategy for difficult object interactions, investigates human performance of this strategy, and develops techniques for planning similar actions for a robot manipulator or animated character. The goal is to use this human-inspired strategy to improve manipulation by artificial agents through increased task success and appearance of dexterity.

Pre-grasp interaction adjusts object placement in the environment instead of directly grasping from its presented position. Humans use this strategy naturally to dexterously manipulate objects in their environment. I believe humans expect similar dexterity from intelligent robots and convincing animated characters or virtual avatars.

Perception of manipulation actions.

In current work, I am investigating how human observers respond to manipulation actions in other agents.

Planning pre-grasp manipulation for transport tasks.

Our human studies and initial robot demonstration suggest the advantages of pre-grasp interaction in task robustness and dexterous appearance. However, automation of the this technique is difficult due to practical challenges in manipulation planning. I develop a method for planning complete manipulation tasks that include pre-grasp interaction, based on a technique similar to posture-based models of human motor behavior.

Project Page (with videos) at Intel Labs website


Lillian Y. Chang, Siddhartha S. Srinivasa, and Nancy S. Pollard. Planning pre-grasp manipulation for transport tasks. IEEE International Conference on Robotics and Automation (ICRA 2010), in press, May 2010.

Selection criteria for preparatory object rotation in manual lifting actions.

We investigate what factors change the human performance of pre-grasp interaction. We found that this strategy is most relevant to difficult tasks involving heavy objects or high precision constraints. In addition, the amount of pre-grasp rotation correlated with the resulting change in lifting capability at the grasping posture. Our results suggest that pre-grasp interaction is more robust for new and repeated tasks compared to direct grasping.


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

Pre-grasp rotation is only one example of a pre-grasp interaction strategy. In this video survey of human hand activity, we filmed people performing manipulation tasks in the home and workplace. We found that there is indeed a broad class of pre-grasp interactions used naturally by humans. Our framework describes the survey examples according to two main aspects of object re-configuration and underlying purpose of the adjustment.

Workshop poster [1.6MB]


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.

Grasping with an anthropomorphic robot manipulator

We demonstrate how pre-grasp interaction can improve robot manipulation. Taking advantage of object movability can make good grasps possible for more task conditions and reduce the expense of programming a complex mechanism. Our empirical evaluation in this example showed that a simple open-loop pre-grasp rotation can increase the grasping success from 45-degree to 360-degree of input object orientations. This was achieved without programming new actions for each different input condition.

Video of robot pan rotation [5MB]


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.

Human lifting strategies under task constraints

In our first study of human pre-grasp interaction, we asked whether pre-rotation was a prevalent strategy performed naturally by humans under certain task constraints. We found that, when grasping heavy objects with one hand, humans naturally re-oriented the object handle to a preferred configuration prior to lifting the object from the surface. Interestingly, this pre-grasp rotation was not absolutely necessary, because the subjects could still complete the task even when rotation was not permitted. Our analysis suggests that the preference to use pre-grasp interaction may be due to the more natural body postures due to lower joint torques or increased stance and grasp stability.


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. In this exploratory project, we measured the complex manipulation humans use to pick up tools from a surface. Our observations of the hand interaction inspired our investigation of pre-grasp interaction in human manipulation.

Learning grasping by demonstration
In this project we were interested in developing a simple gesture interface for grasp imitation learning, where a human can teach a robot how to grasp objects.

Reduced feature set for hand surface markers

Although the human hand has a high-dimensional configuration, only a small set of features may be necessary for distinguishing functional grasp classes. We use supervised feature selection to methodically select a minimal set of hand marker locations for grasp recognition. The reduced marker set has only 5 instead of 30 markers, retains at least 92% of the prediction accuracy, and generalizes better to new subjects. With this simplified marker set, a natural interface using optical markers or vision-based tracking is a practical alternative to data gloves for grasp demonstration.


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
One aspect of human dexterity is the hand's versatility of shape. To assess the hand's complex coordination patterns, we need reliable measurement of its configuration. Standard methods for tracking skeletal motion require an expert practitioner to locate anatomic landmarks on an individual.

This series of projects investigates techniques which allow non-expert users to obtain quality motion data from generic marker placements. The algorithms I developed provide automatic estimation of kinematic parameters to fit customized skeletal models of an individual's hand.

Two axis model for the thumb carpometacarpal joint

The mobility of the thumb carpometacarpal (CMC) joint is critical for functional grasping and manipulation tasks. It is also very difficult to measure due to the non-intersecting and non-orthogonal rotation axes of the bones' saddle-joint geometry. We present an optimization technique for non-invasively determining a subject-specific kinematic model of the CMC joint. Our technique provides anatomically-plausible results and is compatible with clinical descriptions of thumb mobility. The key ideas of our method were to bias the search according to known anatomy characteristics and to formulate a low-dimensional parameterization of the optimization domain.


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

Many joints of the hand (and other parts of the body) may be represented by a hinge joint with a single dominant axis of rotation. Real joints have limited joint range of motion and additional rotation due to a second degree of freedom, causing poor model-fitting with previous approaches. We propose a simple method for robustly estimating the dominant axis of rotation by penalizing both in-plane and out-of-plane errors of surface marker trajectories. This reflects the joint constraints more accurately than previous methods, and the dominant rotation axis can be determined even in the presence of secondary motion.


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

A spherical or ball joint is a simple and useful model for describing anatomic joint center locations. Previous methods for joint center estimation work well for near-ideal motion. However, in real hand data, small range of motion and noise result in poor estimates due to plane singularities of the existing techniques. Our solution uses careful geometric analysis to relocates the plane singularity to a point singularity, which does not occur in practice. The constrained least-squares formulation provides a direct repeatable solution wihout the local minima or manually-tuned parameters typical in iterative techniques.


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.