UB - University at Buffalo, The State University of New York

Dynamics, Control, and Mechatronics

Faculty

• Crassidis, J.
• Krovi, V.
• Mook, J.
• Singh, T.
• Singla, P.
• Soom, A.

Affiliated Faculty

• Liang, Z.
• Marsh, E.

Laboratories

• Automation Robotics and Mechatronics Laboratory (ARM Lab)
• Control, Dynamics and Estimation Laboratory (CoDE)
• Lab for Autonomous and Intelligent Robotics Systems (LAIRS)
• Virtual Reality Laboratory

Research Summaries

Spacecraft and Aircraft Formation Flying Navigation

Research is being performed to develop new attitude and position determination methods for formation flying applications. Spacecraft and aircraft formation flying is an evolving technology with many possible applications, such as long baseline interferometry, stereographic imaging, synthetic apertures, and distinguishing spatial from temporal magnetospheric variations. A significant advantage of distributed platforms over a single multifunctional vehicles is that single point failures can be rectified through replacement of cheaper and smaller spacecraft to maintain mission capability, thus providing a more reliable and robust system. Many missions rely on precise relative position and attitude knowledge in order to maintain mission requirements. To date, most research studies into determining relative positions and attitudes between vehicles have involved using the Global Positioning System (GPS), which is subject to performance-limiting effects, including multipath, geometric dilution of precision, integer ambiguity resolution, and cycle slip. The main objective of the research is to provide novel, reliable and autonomous relative navigation and attitude determination systems, being fully independent of any external systems.

J. CRASSIDIS. Sponsor: US Air Force and NASA.

Attitude and Angular Velocity Estimation

Research is being performed to develop new and efficient algorithms for attitude and angular velocity determination for spacecraft applications. This involves robust methods using GPS signals, real-time calibration methods, and a novel method to determine angular rates from a star camera. The goal of the GPS research is to develop fully autonomous algorithms that are robust for any initial attitude. Specific issues to be addressed in the research include: 1) provide optimal attitudes even for coplanar baseline configurations, 2) guarantee convergence even for poor initial conditions, and 3) develop computationally efficient algorithms. The goal of the real-time calibration methods is to provide onboard solutions to spacecraft calibration problems in order to alleviate ground processing approaches, which require significant costs. The goal of the angular rate determination approach is to use star observations to determine body rates without ever identifying the stars observed. This approach, coupled with new technology that enables for fast sampling rates, can lead to a dual-use capability for a star camera, i.e. attitude and rate determination. The eventual goal is to eliminate the need for costly gyros, which are required for fine-pointing spacecraft.

J. CRASSIDIS. Sponsor: NASA.

Generalized Multiple-Model Adaptive Estimation

A new generalized approach for multiple-model adaptive estimation (MMAE) is being developed, which can be used for time-varying and nonlinear systems. This approach is based on using the autocorrelation of the measurement-minus-estimate residual. Initial results indicate that the new generalized MMAE approach can provide better convergence properties than the standard approach. MMAE schemes are widely used on a number applications, such as fault detection, robust tracking of vehicles, network centric warfare applications, information technology applications and extreme events. The new generalized MMAE approach has the potential to revolutionize these areas.

J. CRASSIDIS.

Generalizations of the Complex-Step Derivative Approximation

This works aims at deriving a general framework for the complex-step derivative approximation to compute numerical derivatives. For first derivatives the complex-step approach does not suffer subtraction cancellation errors as in standard numerical finite-difference approaches. Therefore, since an arbitrarily small step-size can be chosen, the complex-step method can achieve near analytical accuracy. However, for second derivatives straight implementation of the complex-step approach does suffer from roundoff errors. Therefore, an arbitrarily small step-size cannot be chosen. In this work we expand upon the standard complex- step approach to provide a wider range of accuracy for both the first and second derivative approximations. Higher accuracy formulations can be obtained by repetitively applying the Richardson extrapolations. The new extensions can allow the use of one step-size to provide optimal accuracy for both derivative approximations.

J. CRASSIDIS.

Cooperative Payload Transport by Robot Collectives

Our goal is to design, analyze and implement a flexible, scalable system of multiple mobile manipulators, that are individually autonomous but can team up to cooperatively transport large payloads and to perform tasks that simply cannot be performed by a single mobile platform. Such frameworks for remotely controlled or remotely supervised cooperation of multiple autonomous modules have many applications from uneven terrain exploration to material handling on the shop floor.

V. KROVI.

Nonlinear Identification

Most identification techniques require that the form and order of the dynamic model equations be assumed, and then the constant parameters in the dynamic equations are estimated using measurements. A technique that does not require a prior assumption of the model form and order is being developed and will be utilized to aid in identifying the nonlinear behavior of large space truss joints.

D.J. MOOK.

Control of Nonlinear Systems

Optimal linear feedback of linear state-space systems is well known, and is typically found from solution of a Riccatti equation. This project is investigating various techniques for determining an optimal linear or nonlinear feedback law for a nonlinear system.

D.J. MOOK.

Spacecraft Attitude Estimation and Identification

Standard spacecraft attitude estimation is performed using either a large complement of attitude and attitude rate sensors, and/or a Kalman-filter algorithm. In either of these approaches, the accuracy is limited in instances of sensor failure or a poor dynamic model. This project investigates the use of nonlinear identification techniques to determine on-orbit accurate attitude dynamic models for the SAMPEX spacecraft. Actual flight data from the SAMPEX spacecraft is made available through an agreement with NASA Goddard Spaceflight Center. SAMPEX is the first of the new Small Explorer (SMEX)-class satellites and was launched into a low-earth orbit in July 1992.

D.J. MOOK.

Data Assimilation for Chem-Bio Dispersion

The objective of this project is the systematic study and elucidation of the basic physico-mathematical principles underlying data assimilation in the chem-bio context, that is, the blending of chem-bio dispersion forecasts with uncertain sensor data. This problem is mathematically characterized by high dimensionality, profoundly nonlinear models, and uncertainties with non-Gaussian non-stationary distributions. The physical dispersion models involve a balance of numerous natural forces acting in the complex atmospheric medium and are propagated with meteorological data which is itself uncertain.

T. SINGH, P. Singla.

Sequential LP for Nonlinear Controller Design

This project deals with the development of a technique which uses linear programming for the determination of solutions of the Mayer Problem. This includes the minimum time optimal control problem and maximizing terminal state problem among others. An iterative approach which solves for a perturbation control using linear programming permits exploitation of the potential of LP in solving nonlinear control problems.

T. SINGH, P. Singla.

Robust Control of Nonlinear Systems

Modeling uncertainties which are present in any system representation have tempered the performance of controllers designed to meet stringent specifications. These uncertainties include errors in model parameters and unmodeled dynamics. The effect of parametric uncertainties is accentuated when the controllers are required to switch between their limits as in time-optimal and fuel/time optimal controllers. This work focuses on developing techniques to design robust controllers for nonlinear systems subject to control constraints.

T. SINGH.

Repetitive Control

There are numerous systems that are subject to periodic disturbances. Some applications include hard disk drives and CD players where the non-coincidence of the geometric center of the disk and the center of rotation results in a periodic disturbance where the dominant requency is the speed of rotation of the disk. Other applications include robots performing the same operation repeatedly. Repetitive control is an Internet Model Control approach that permits the rejection of such disturbance. This research project focuses on developing control techniques that are robust to uncertainties or variations in the periodicity of the disturbance.

T. SINGH.

Iterative Learning Control

This is a feed-forward control technique that is used for high precision control of systems undergoing repetitive tasks. At the end of every iteration, the tracking error is used to update the feed-forward controller with algorithms that guarantee convergence. This research focuses on development of filters for the learning part of the iterative controller to achieve performance objective such as rapid convergence, robustness to and variations in system parameters.

T. SINGH.

Simultaneous Feedback / Feed-Forward Control Design

Extensive work has been carried out related to the design of feedback controllers for vibrations attenuation of maneuvering systems. Over the past 10 years a large body of literature has also evolved related to the pre-filtering of reference input for vibration attenuation. These two design techniques have evolved independent of each other. This work focuses on the integration of the design of the feedback and the feed-forward controllers.

T. SINGH.

Hybrid Filters for Target Tracking

This work focuses on the development of a filter that permits tracking on a user specified profile. This filter is proposed to be used in conjunction with a Kalman filter to permit tracking maneuvering targets. The unscented filter will be used for the estimation of the covariance matrix.

T. SINGH.

Image-Guided Tracking of Internal Organ and Tumor Motion

This research work deals with the development of safe and effective Adaptive Conformal Radiation Therapy for cancer treatment while minimizing the relapse rate of tumor and side effects of the lethal radiation dose. The main objective of this research work is to design and test a novel framework for accurate estimation of 7-D (position + orientation) tumor target dynamics based on the correlation of real-time imagery data from external and internal fiducial markers. The core tool at the heart of our approach is recently developed adaptive multi-resolution system identification algorithm, which makes use of recent advances in Neural Networks, Finite Element Methods and Nonlinear Adaptive Control. An important aspect of our work is to adapt respiratory models in real-time which helps us in addressing many critical issues specific to image-guided radiation therapy such as distinguishing between patient movement and respiratory motion, signal deficiencies, time latency, frequency of the measurement data and uncertainty in the breathing models currently in use.

P. SINGLA, T. SINGH.

Uncertainty Propagation Through Nonlinear Dynamical Systems

This problem involves the study of the time evolution of the state probability density function corresponding to the state of a dynamical system using measurements from multiple sources. Mathematically, it is a formidable problem to solve because of the issues like positivity, normality, discretization, and most importantly, the dimensionality of the system. Conventional approaches such as Kalman filters, ensemble filters, and particle filters work well when measurement updates are available frequently, however; there is no way of updating the probability density function (characterizing uncertainty) weights during propagation. The objective of this project is to develop novel analytical and computational tools for efficient propagation of uncertainties through nonlinear dynamical systems while using Fokker Plank Equation error as feedback to update the weights in the absence of measurement data. Applications of interest include tracking of a space object for the determination of its orbit, diffusion of Chem-Bio Radioactive Nuclear (CBRN) material and mobility prediction for mobile robots in uncertain environment.

P. SINGLA.

Motion Planning for Unmanned Air-Ground Vehicles

The main focus of this project is to study the problem of real-time motion planning for autonomous aerial and ground vehicles with special focus on the development of algorithmic models for planning under uncertainty. The problem of operation in a cluttered urban environment is an especially difficult one due to limited work space available for maneuvering. Buildings and other obstructions often limit the air vehicle¡¯s line of sight, and can also hinder other required functions such as visual assessments. The key for success in real-time motion planning is their ability to recognize obstacles in real-time and take affirmative actions as quickly as possible and viable to vehicle dynamics. As a consequence, the success of such missions is highly correlated to the robustness of the algorithms that navigate these vehicles in an uncertain environment. The main objective of our work is the development of hierarchical, real-time motion planning algorithms that incorporate risk into obstacle detection, localization errors, system dynamics and real-time control of vehicles.

P. SINGLA.