Pedram Kheirkhah Sangdeh
Michigan State University, Ph.D. Candidate
sangdeh at msu.edu

Biography

I am a Ph.D. candidate in the Department of Computer Science and Engineering at Michigan State University (MSU), skilled in in wireless communications and networking, signal processing, and machine learning. My research interests include design, simulation, and analysis of networking protocols for wireless networks.

NEWS

  • 08/2021: Won best paper award in IEEE SECON 2021
  • 02/2021: DeepMux accepted by IEEE JSAC--ML series
  • 01/2021: VehCom accepted by IEEE Trans. Wrireless Commun. (TWC)
  • 10/2020: DM-COM accepted by IEEE Internet Things J. (IoTJ)
  • 08/2020: LB-SciFi accepted by IEEE ICNP 2020
  • 05/2020: One paper accepted by IEEE/ACM Trans. Netw. (ToN)
  • 04/2020: TCCI accepted by ACM MobiHoc 2020

Fall 2020 - Present
Ph.D. student in Computer Science

at Michigan State University

Fall 2017 - Summer 2020
Ph.D. student in Electrical Engineering

at University of Louisville

Fall 2011 - Spring 2014
M.S. in Electical and Computer Engineering

at University of Tehran

Fall 2006 - Spring 2011
B.S. in Electrical and Computer Engineering

at Iran University of Science and Technology

Selected Publications

Full List of Publications →

DeepMux: Sounding and Resource Allocation for WiFi6

DeepMux is deep-learning-based MU-MIMO-OFDMA transmission scheme for 802.11ax networks. DeepMux comprises DLCS and DLRA. DLCS uses DNNs to reduce the airtime overhead of 802.11 protocols and DLRA employs a DNN to solve the mixed integer resource allocation problem.

UD-MIMO: Uplink Distributed MIMO for Wireless LANs
SECON 2021-Best Paper of SECON'21

UD-MIMO enables concurrent uplink transmission in the absence of fine-grained synchronization. Its enabling technique is a scheme to decoding uplink packets from asynchronous users. UD-MIMO makes it possible for WLANs to significantly improve their uplink throughput without requiring synchronization.

A Spectrum Sharing Scheme for CRNs: Design and Experiments

We proposed a spectrum sharing scheme for cognitive radio networks where secondary users have no knowledge about the primary network. We have built a prototype of our scheme that showed the secondary network could coexist with commercial Wi-Fi and LTE devices without degrading their performance.

LB-SciFi: Online Learning-Based Channel Feedback in WLANs

LB-SciFi is a learning-based feedback framework for MU-MIMO in WLANs. It employs DNN autoencoders to compress CSI in WiFi networks, thereby conserving airtime and improving spectral efficiency. Experimental results show that it offers 73% airtime overhead reduction and 69% throughput gain.

A Practical Downlink NOMA Scheme for WLANs


We proposed a downlink NOMA scheme for WLANs including precoder design, user grouping, and successive interference cancellation (SIC). We prototyped the proposed scheme and experimental results show that the proposed downlink NOMA significantly improves the weak user’s date rate and sum rate.

TCCI: Taming Co-Channel Interference for Wireless LANs

TCCI is an interference management scheme enabling concurrent transmission in WLANs. TCCI requires neither network-wide synchronization nor inter-network data sharing, and therefore is amenable to implementation. We have illustrated TCCI compatibility with commercial devices.

A Practical Underlay Spectrum Sharing Scheme for CRNs


This paper proposes a practical underlay spectrum sharing scheme for CRNS where the primary users are oblivious to secondary users. The key components of our scheme are two MIMO-based interference cancellation techniques, blind beamforming and blind interference cancellation techniques.

EE-IoT: An energy-efficient IoT communication for WLANs


In this paper, we propose an energy-efficient IoT communication scheme by taking advantage of the existing WiFi infrastructure. EE-IoT will not only avoid monthly service charge for the end users but also maintain a low power consumption for IoT devices.

Current Projects

FL acceleration in VANETs

In this project, we aim to increase the convergence of federated learning in a deadline-constrained networks such as VANETs, where local models need to be polled considering vehicles deadines.

Storage-free federated learning

In most federated learning frameworks, the non-selected users store data samples. This paradigm is not suited for low-end storage-limited users. In this project, we aim to revisit this paradigm.

Experimental Data Sets

If you need real experimental datasets to evaluate your ideas, you can use the following datasets. No permission is required.

CSI for 3600 user pairs: two users and one two-antenna BS/AP at 2.4GHz and 5MSps


This test measures the channel gain difference for 3600 users pairs to investigate the statistics of the gain difference among all possible user pairs for using in NOMA. . We have measured the average channel gain of users on 2.485 GHz with 5MHz bandwidth.

8x8 MIMO for indoor WLANs at 2.1GHz and 25MHz: freq. response and correlation


This test aims at measuring the channel frequency response of an indoor wireless environment over 25 MHz bandwidth in 2.1 GHz band. Based on the measured results, we can study the rank deficiency of MIMO channels to have a better DoF allocation for MIMO transmissions.

Channel Reciprocity Test for USRP N210: DL/UL Mistmach in 7 Hours

This test examines the channel reciprocity for USRP N210 at 2.4 GHz with 5MSps for 7 hours. Wi-Fi legacy frame is used to measure the uplink and downlink channels over 52 valid subcarriers. We used Argos as a relative calibration method to measure and compensate the mismatch for 4 N210 USRPs with SBX daughterboards.

Last updated: 10/27/2021-21:41:52