Projects

Computer Vision - Control Theory - Locomotion - Medical Imaging - Robotics


Computer Vision

Research into computer vision is driven by a controls perspective. Projects include vision-based closed-loop control, estimation and filtering of computer vision algorithms, and computer vision for decision making and knowledge generation. Some research explores the role of statistical machine learning in computer vision.

Active Vision:
- vision in the feedback loop.
As-Built Modeling:
- using vision to understand the built world.
Biomechanics:
- tracking the kinematics of people and animals.
Gait Analysis:
- analyzing movement to aid rehabilitation.
Observe:
- estimation theory and contour-based tracking.
Surveillance:
- systems for understanding work processes and social interactions.
Visual SLAM:
- observability theory meets localization and mapping.
Visual Tracking:
- statistical machine learning and tracking.
- top -

Control Theory

Recent research has revolved around neuro-adaptive control and how the theory behind single-layer neural networks used in adaptive control could be improved by insights or techniques from the machine learning community. Older research was on geometric nonlinear control theory. the work sought to connect averaging theory to control-oriented series expensions for the purpose of controling underactuated systems.

Neuro-Adaptive Control:
- online learning in a controls setting.
Geometric Control Theory:
- where differential geometry meets nonlinear control theory.
- top -

Locomotion

Research on the design of control strategies for biologically-inspired robotic systems. We know how to engineer them, but do we really know how to control them? Considers both the practical aspects of biologically-inspired robotics and the theoretical aspect, with the intent to unite these two facets for the purpose of engineering autonomous robots.

Feedback Control of Gaits:
- using gaits and geometric control for closed-loop motion.
Optimal Trajectory Gait Synthesis:
- using gait dynamics to snythesize gait-optimal trajectories.
Robotic Snakes:
- design, dynamics, and control of robotic snakes.
- top -

Medical Imaging

Applies some of the IVALab research to medical imaging.

Non-rigid registration:
- connecting the dots from one scan to the next.
Interactive Segmentation:
- human-in-the-loop dynamic segmentation.
- top -

Robotics

Touches on robotic system components where the solutions being studied in our vision and control research domains transfer naturally to robotic systems.

Robot Planning:
- planning actions for robots and manipulators.
Root-Locus Path Generation:
- globally valid reactive path generation for planar robots.
- top -