Meng Wang

Assistant Professor
Department of Electrical, Computer & Systems Engineering
Rensselaer Polytechnic Institute

Phone: 518.276.3842 Fax: 518-276-6261
ECSE Department, JEC 6024
Rensselaer Polytechnic Institute Troy, NY, 12180

Home   Research   Publications   Teaching   Group 

Due to the massive scale of engineering networks such as the Internet and the power systems, network operators receive terabytes of monitoring data constantly. The large amount of data provides detailed information about the network operation status, but also imposes significant computational and operational challenges to network operators. The reliable operation of the Internet and the power grid requires the development of data acquisition and transmission methods that can obtain and transmit pertinent measurements with efficient energy consumption, as well as the development of computationally efficient inference methods that can accurately extract key operational characteristics from the obtained measurements.

My research lies at the intersection of signal processing, networking monitoring and optimization. The goal is to develop a theoretical framework of data-challenged network control and monitoring and to explore the practical implications of these fundamental understandings in the monitoring of large-scale engineering networks.

Efficient Sensing and Inference by Exploiting Low-dimensional Structures

Efficient monitoring of engineering networks requires accurate estimates of key characteristics. It might not be feasible to directly measure all these network parameters due to operational constraints. Even if direct measurements are allowed, the large amount of sensor measurements would result in high sensing costs in terms of both energy consumption and transmission bandwidth requirements. Therefore, it is important to develop efficient data acquisition and inference methods that can accurately recover key network characteristics from a relatively small number of pertinent measurements.

Many practical high-dimensional datasets exhibit intrinsic low-dimensional structures. Compressed sensing theory indicates that by exploiting the low-dimensional structures(e.g. sparsity, low-rankness), the massive amount of information can be fully captured and exactly recovered from a relatively small number of linear observations. This opens up applications in image processing, medical applications, network inference, etc . One fundamental question is what is the required number of measurements for identifying signals that have certain special structures?

Weak Recovery

Strong Recovery

Connection to Internet Monitoring

Existing results in the literature of compressed sensing usually assume that an observation could be any linear combination of parameters to recover. In Internet monitoring, however, a measurement should satisfy network topological constraints. For example, one important problem in Internet monitoring is to identify the transmission delays on individual transmission links from aggregate end-to-end path delay measurements. Then feasible measurements should satisfy network topological constraints in the sense that each measurement should correspond to a valid path in the network. How can we design measurements subject to network constraints such that the number of measurements required to recover a sparse signal is as small as possible?

Connection to Power System Monitoring

State estimation solves for the states of a power grid, usually in the form of bus voltage magnitudes and angles, from the obtained meter measurements. Meter measurements may contain errors at unknown locations due to sensing errors or communication failures, and the erroneous measurements may significantly impair the accuracy of state estimation. Can we improve the data quality of the measurements by exploiting the correlations in meter measurements imposed by the interconnections of the power systems?

  • P. Gao, M. Wang, S. G. Ghiocel, and J. H. Chow.  Modeless Reconstruction of Missing Synchrophasor Measurements. accepted to IEEE Power & Energy Society General Meeting, 2014.


I gratefully acknowledge the current and past support from the following funding agencies and organizations.

picture National Science Foundation

picture Army Research Office

picture New York State Energy Research and Development Authority

picture CURENT NSF & DOE Engineering Research Center
picture Electric Power Research Institute (EPRI)