Main menu >> Electrical and Computer Engineering >> Advanced Automation

Electrical and Computer Engineering

Advanced Automation
^ Augmented Reality
N. Ahuja,* J. Ma
U.S. Army Research Laboratory, DAAL01-96-2-0003F

(Conducted in the Coordinated Science Laboratory)

The objective of this project is to develop computer vision-based approaches to augmentation of 3-D displays of real scenes. Displays may select a subset of original image features or add new ones to enhance the perception of the scene structure and dynamics. The displays may also overlay on the images information from a variety of sources to increase the situational awareness.

^ Development of Head-mounted Projective Displays for Distance Collaborative Environments
N. Ahuja,* H. Hua,* L. Brown, C. Gao
National Science Foundation

The objectives of this project are to develop a novel visualization device, called head-mounted projective display (HMPD), which allows real-time superposition of a direct image of the scene with a stored virtual view; build a multi-user interactive workbench by integrating the developed HMPD with a high-performance real-time image acquisition system; and evaluate the performance of the resulting system as a tool for remote collaboration.

^ Higher Order Feature Detection
N. Ahuja,* A. Sehgal, A. Jagmohan, K. H. Tan
Defense Advanced Research Projects Agency

(Conducted in the Coordinated Science Laboratory)

This project is concerned with higher order features for recognition of objects in multisensory images. The target characteristics are known because the target is known. The sensor characteristics are known too. A model of the target is to be created in terms of complex features that are defined in terms of more low-level, primitive features extracted by simple detectors from the data acquired by known sensors.

^ Image Matching and Interpretation
N. Ahuja,* M. H. Yang
ATR International

This project is aimed at the interpretation of moving, nonrigid surfaces carrying little or limited detail, with applications to virtual space teleconferencing. The objectives include delineation of moving parts of a scene, active selection of viewpoints and data acquisition, and integration of focus, shading, and silhouette information for functionality under a range of environmental conditions.

^ Image Segmentation
N. Ahuja,* R. Dugad
Eastman Kodak Co.

(Conducted in the Coordinated Science Laboratory)

The goal of this project is to segment an image, or an image sequence, into its constituent regions such that each region is characterized by homogeneity of a three-dimensional property. Currently, we are developing segmentation algorithms that use uniformity of three-dimensional surface texture and three-dimensional object motion as homogeneity criteria.

^ Path Planning for Robot Navigation
N. Ahuja*
Rockwell International

(Conducted in the Coordinated Science Laboratory)

This project concerns efficient generation of object representations from multiple perspectives. We are developing algorithms to generate octree representation of an object from its planar projections. We use the known representation of obstacles to plan efficient motion trajectory to move an object from one location to another.

^ Recognition and Contents-based Retrieval of Hand Gestures from Video
N. Ahuja,* M. H. Yang, S. C. Yoon
U.S. Office of Naval Research

(Conducted in the Coordinated Science Laboratory)

This project is concerned with recognition of scenes from the spatiotemporal structure of the video data. Trajectories of scene contents seen in the video sequence are used as the basis for this purpose. Objects are characterized by their spectral properties as well as temporal behavior. Such representations are used for information access as well as for recognition using methods such as support vector machines.

^ Real-Time Path Planning in Changing Environments
S. Hutchinson
National Science Foundation

New methods are proposed to generate collision-free paths for robots that operate in environments that change over time. The proposed approach is related to recent probabilistic roadmap approaches. These planners use preprocessing and query stages and are aimed at planning many times in the same environment. In contrast, the preprocessing stage for this research creates a representation of the configuration space that can be easily modified in real-time to account for changes in the environment. As with previous approaches, the proposed approach began by constructing a roadmap in the configuration space, but this roadmap is not constructed for a specific workspace. Instead, it is constructed for an obstacle-free workspace, and the mapping from workspace cells to nodes and arcs in the roadmap is encoded. When the environment changes, this mapping is used to make the appropriate modifications to the roadmap, and plans can be generated by searching the modified roadmap. At the heart of the method is the encoding for mapping workspace obstacles to configuration space obstacles. To make the proposed approach truly viable, a major component of the proposed research will focus on robustness and complexity issues. These issues will be addressed by using tools from the fields of image processing, information theory, graph theory, computational geometry, and incremental algorithms.

^ Development of a Robot that Plays Air Hockey
M. W. Spong,* S. Hutchinson, S. Kuo, S. Bunchongchuits
National Science Foundation, IRI-9216428, CMS-9712170; Electric Power Research Institute, RP 8030-14

This project is to develop a three-degree-of-freedom robot that can play air hockey. Research issues being addressed include real-time visual servoing, adaptive camera calibration and windowing, hybrid estimation, and hybrid nonlinear control. Based on the reliability of sensory information, a supervisory control system switches among a fixed set of nonlinear controllers, each designed for a particular task such as blocking or striking the puck. Future research is aimed at learning through repetitive play.

^ Nonlinear Control of Underactuated Mechanical Systems
M. W. Spong*
National Science Foundation, CMS-9712170, CMS-9840985

This project seeks to develop stability and tracking results for underactuated mechanical systems using tools from Lagrangian and Hamiltonian dynamics, geometric nonlinear control theory, hybrid control, and saturation. This work exploits the underlying structure of the nonlinear dynamics, makes efficient use of passivity and energy, and complements existing energy and passivity methods and more recent backstepping methods, all of which attempt to exploit more fully the inherent nonlinearities of the system, by "shaping" rather than by "canceling" all of the nonlinearities in the system.

^ Learning Sensorimotor Control of Balance and Locomotion
M. Spong,* G. DeJong, S. Hutchinson, K. Rosengren, R. Sreenivas
National Science Foundation, ECS-9812591

The goal of this project is to investigate computational methods for learning sensorimotor control in bipedal locomotion. The project will integrate ideas from engineering, psychology, and kinesiology. Researchers will utilize techniques from control theory and artificial intelligence to improve understanding of the dynamics and control of human movement and the mechanisms by which humans learn sensorimotor control. The research team will use studies of human movement to develop improved learning and control techniques for multi-degree-of-freedom mechanical systems. Applications include more dexterous robots, more effective diagnostic and physical therapy approaches for disabled humans, and better balance training and falls prevention programs for elderly and individuals with balance deficits.


Summary of Engineering Research