Particle filter technology is state-of-the-art in tracking moving objects that follow complex dynamics in noisy and uncertain environments. The military uses it for radar tracking. It is used in video monitoring for surveillance, security and traffic control. It has also been used for global positioning of robots. The power and wide applicability of this technology stem from its simple implementation that enfolds a mathematical foundation with established optimal properties of convergence and locking into the target. However, these mathematical guarantees fail to hold when the moving target obeys additional constraints that are imposed by the environment or the underlying dynamics of the system. This research established the mathematics of the particle filter technology for tracking moving targets under additional (stress) constraints.

In addition, we collaborated with several institutions to evaluate this new technology in several applications, including prosthetic devices, brain source imaging, molecular networks and machine learning under uncertainty. Specifically, we collaborated with the Rehabilitation Institute of Chicago to address the problem of movement identification for the forearm prosthesis from Electromyogram (EMG) data. EMG provides an insight into the neural processes taking place in the central nervous system (CNS) for planning and execution of voluntary tasks. The CNS generates, coordinates, controls and executes voluntary movements in a highly complex manner, where the neural drive and muscle activations are intertwined in a highly nonlinear fashion. We took these nonlinearities and constraints into account, producing more realistic models of the neural drive-myoelectric map and leading to improved performance on neuromuscular models and actual prosthesis devices.

This project was integrated into the curriculum through the Engineering Clinic Program at Rowan University, which provides a unique mechanism for involving undergraduate students in research. An engineering clinic group of Juniors and Seniors participated in the Idea Challenge sponsored by Rowan Innovation Venture Fund & the Center for Innovation & Entrepreneurship. They won 2nd place for “Smart Bionics”, which aims at controlling prosthetic arms. This cross-disciplinary clinic became very popular with many engineering students (electrical, computer, biomedical and mechanical) requesting it for subsequent academic years. A total of 40 undergraduate students participated in this Clinic project; some of them joined the PhD Program in Engineering after graduation.

This research took an international dimension through a collaboration with The University of Aveiro, Portugal, and the University of Sheffield, UK, on the problem of dynamic Electroencephalography (EEG) brain source localization during execution of multisensory activities. In dynamic EEG source localization, the source brains, also called dipoles, are not stationary but move in the brain depending on the activity being performed. By constraining their movement to relevant areas in the brain, we showed that we can improve the tracking accuracy of brain dipoles during a continuum of activities.

With the University of Arizona, we introduced a new approach to tracking sparse molecular networks, which evolve over time in response to cellular development and environmental changes. Understanding the dynamical behavior of living cells from their complex genomic regulatory networks is a challenge posed in systems biology; yet it is one of critical importance (i.e., morphogenesis). While investigating the problem of tracking dynamic networks under sparsity constraints, we revisited the framework of compressive sensing, which aims at exact reconstruction of a signal from a limited number of measurements. The proposed technology yields significantly smaller errors than current methods. We inferred the wing muscle gene regulatory network of the Drosophila Melanogaster (fruit fly), during four developmental phases of its life cycle, and successfully identified all seven known interactions reported in Flybase.

This grant was instrumental in building our entrepreneurship activities. We applied to an NSF I-Corps supplement through this project to gain skills in entrepreneurship through training in customer discovery and guidance from established entrepreneurs. This I-Corps led to an NIH SBIR Phase I grant in September 2020.

This project funded some of the inaugurating PhD students to the new PhD in Engineering Program. Two women PhD students, Ms. Nesrine Amor and Ms. Dimah Dera, were directly funded by this grant.  Ms. Amor is an exchange student from the National Superior School of Engineers of Tunis, Tunisia. She graduated in May 2018 and is now a Research Fellow with the Technical University of Liberec in Czech Republic. Ms. Dera graduated in May 2020 and will be joining the Department of Electrical and Computer Engineering at the University of Texas Rio Grande Valley as a Tenure-Track Assistant Professor starting August 2021. A 3rd woman PhD student (Giuseppina Carannante) is currently working on the extension of this framework for propagating uncertainty in machine learning systems to enhance their reliability and trustworthiness.

The results were disseminated through 5 journal publications, 7 peer reviewed conference publication, including 1 Best Paper Award, as well as invited talks at international institutions, Government Labs, and industry, including The Federal Aviation Administration, Lockheed Martin, and the University of Sheffield, UK.

Last Modified: 05/13/2021
Modified by: Nidhal Bouaynaya