The Ohio State University Corporate Engagement Office

Back to All Technologies

Fast and Efficient Cross Band Wireless Channel Prediction Using Machine Learning

Engineering & Physical Sciences
Software & Information Technology
Communications & Networking
Communications & Networking
Wireless Communications Antennas
Machine Learning / Artificial Intelligence
College of Engineering (COE)
Srinivasan, Kannan
Bakshi, Arjun
Parthasarathy, Srinivasan
Licensing Manager
Hong, Dongsung

T2020-058 A neural network that utilizes a simple and efficient single antenna optimization framework

The Need

Systems that use different frequency bands for uplink and downlink communication often need feedback between the clients to exchange band-specific channel information. The current state-of-the-art approach predicts the downlink channel based on the uplink channel by identifying certain variables underlying the observed uplink channel. In such applications, the channel observed in the receiving frequency band is different from those in the transmitting frequency band. Therefore, a device cannot use the channel it observes in the receiving frequency band to perform tasks like beamforming or rate selection in the transmit frequency band. Traditionally, in cellular systems, channel values are reported back to the base station by the mobile clients, either with or without compression. However, those approaches can be prohibitively expensive or can compromise accuracy. Thus, there is a need to reduce the complexity of this task and make it applicable for single-antenna devices.

The Technology

A team of researchers, led by Dr. Kannan Srinivasan, has developed a neural network trained on a standard channel model to generate coarse estimates for the variables underlying the channel. This invention is the first to apply machine learning to the problem of estimating component distances in a multipath rich channel. The technology uses a simple and efficient single antenna optimization framework to more accurately estimate variables which are used for downlink channel prediction. The program implements a software-defined radio approach and has been compared to the state-of-the-art through experiments and simulations. This technology has demonstrated a reduced time complexity ranging from 10x to 80x while maintaining a similar prediction quality.

Commercial Applications

  • Communications & Networking
  • Base Stations
  • User End Devices (Cellphones, Laptops, Tablets)
  • Smart Home Devices


  • Match performance of the current state-of-the-art approach for beamforming
  • Provides a order of magnitude reduction in runtime (10-80x)
  • Single antenna operation
  • Low complexity

Research Interest

Dr. Kannan Srinivasan is an associate professor in the Department of Computer Science and Engineering at The Ohio State University. His research develops wireless network and communication systems that are efficient, reliable and secure. His interests include networking protocols, measurements, communication systems, and security. Finally, Dr. Srinivasan is a member of the Co-Sy-Ne Communication Systems and Networking group at The Ohio State University which investigates on upcoming and fundamental research problems in Communication and Networking.