Comparing stochastic and Markov decision process approaches for predicting radio frequency interference

Jacob A. Kovarskiy, Mark Kozy, Charles Thornton, Anthony F. Martone, Ram Mohan Narayanan, R. Michael Buehrer, Kelly D. Sherbondy

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This work evaluates the performance of a cognitive radar system which predicts and avoids radio frequency interference (RFI) through an alternating renewal process (ARP) model-based and Markov Decision Process (MDP) approach. As radio frequency (RF) environments grow more crowded, the need for such a system becomes necessary. The cognitive radar monitors the RF activity to train a model for RFI prediction and avoidance. By modeling activity as an alternating renewal process, the stochastic approach calculates the likelihood of interference from measured RFI statistics. Alternatively, the MDP uses reinforcement learning to determine the optimal sequence of decisions given measured RF activity. Both methods eventually select the widest radar transmit bandwidth to minimize interference. The performance of each approach is evaluated by the number of collisions and missed opportunities. A hardware implemented test-bed deploys both methods on a set of synthetic and real measured RFI spectra in real-time to compare performance with the goal of determining when each process is more beneficial (in terms of performance and complexity).

Original languageEnglish (US)
Title of host publicationRadar Sensor Technology XXIII
EditorsKenneth I. Ranney, Armin Doerry
PublisherSPIE
ISBN (Electronic)9781510626713
DOIs
StatePublished - Jan 1 2019
EventRadar Sensor Technology XXIII 2019 - Baltimore, United States
Duration: Apr 15 2019Apr 17 2019

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume11003
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceRadar Sensor Technology XXIII 2019
CountryUnited States
CityBaltimore
Period4/15/194/17/19

Fingerprint

radio frequency interference
Markov Decision Process
Interference
radar
radio frequencies
Alternating Renewal Process
Radar
interference
avoidance
test stands
reinforcement
learning
hardware
statistics
bandwidth
Reinforcement learning
collisions
Radar systems
Reinforcement Learning
Testbed

All Science Journal Classification (ASJC) codes

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Computer Science Applications
  • Applied Mathematics
  • Electrical and Electronic Engineering

Cite this

Kovarskiy, J. A., Kozy, M., Thornton, C., Martone, A. F., Narayanan, R. M., Buehrer, R. M., & Sherbondy, K. D. (2019). Comparing stochastic and Markov decision process approaches for predicting radio frequency interference. In K. I. Ranney, & A. Doerry (Eds.), Radar Sensor Technology XXIII [1100318] (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11003). SPIE. https://doi.org/10.1117/12.2519675
Kovarskiy, Jacob A. ; Kozy, Mark ; Thornton, Charles ; Martone, Anthony F. ; Narayanan, Ram Mohan ; Buehrer, R. Michael ; Sherbondy, Kelly D. / Comparing stochastic and Markov decision process approaches for predicting radio frequency interference. Radar Sensor Technology XXIII. editor / Kenneth I. Ranney ; Armin Doerry. SPIE, 2019. (Proceedings of SPIE - The International Society for Optical Engineering).
@inproceedings{6f8dd828a943486982278baddf7bda23,
title = "Comparing stochastic and Markov decision process approaches for predicting radio frequency interference",
abstract = "This work evaluates the performance of a cognitive radar system which predicts and avoids radio frequency interference (RFI) through an alternating renewal process (ARP) model-based and Markov Decision Process (MDP) approach. As radio frequency (RF) environments grow more crowded, the need for such a system becomes necessary. The cognitive radar monitors the RF activity to train a model for RFI prediction and avoidance. By modeling activity as an alternating renewal process, the stochastic approach calculates the likelihood of interference from measured RFI statistics. Alternatively, the MDP uses reinforcement learning to determine the optimal sequence of decisions given measured RF activity. Both methods eventually select the widest radar transmit bandwidth to minimize interference. The performance of each approach is evaluated by the number of collisions and missed opportunities. A hardware implemented test-bed deploys both methods on a set of synthetic and real measured RFI spectra in real-time to compare performance with the goal of determining when each process is more beneficial (in terms of performance and complexity).",
author = "Kovarskiy, {Jacob A.} and Mark Kozy and Charles Thornton and Martone, {Anthony F.} and Narayanan, {Ram Mohan} and Buehrer, {R. Michael} and Sherbondy, {Kelly D.}",
year = "2019",
month = "1",
day = "1",
doi = "10.1117/12.2519675",
language = "English (US)",
series = "Proceedings of SPIE - The International Society for Optical Engineering",
publisher = "SPIE",
editor = "Ranney, {Kenneth I.} and Armin Doerry",
booktitle = "Radar Sensor Technology XXIII",
address = "United States",

}

Kovarskiy, JA, Kozy, M, Thornton, C, Martone, AF, Narayanan, RM, Buehrer, RM & Sherbondy, KD 2019, Comparing stochastic and Markov decision process approaches for predicting radio frequency interference. in KI Ranney & A Doerry (eds), Radar Sensor Technology XXIII., 1100318, Proceedings of SPIE - The International Society for Optical Engineering, vol. 11003, SPIE, Radar Sensor Technology XXIII 2019, Baltimore, United States, 4/15/19. https://doi.org/10.1117/12.2519675

Comparing stochastic and Markov decision process approaches for predicting radio frequency interference. / Kovarskiy, Jacob A.; Kozy, Mark; Thornton, Charles; Martone, Anthony F.; Narayanan, Ram Mohan; Buehrer, R. Michael; Sherbondy, Kelly D.

Radar Sensor Technology XXIII. ed. / Kenneth I. Ranney; Armin Doerry. SPIE, 2019. 1100318 (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11003).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Comparing stochastic and Markov decision process approaches for predicting radio frequency interference

AU - Kovarskiy, Jacob A.

AU - Kozy, Mark

AU - Thornton, Charles

AU - Martone, Anthony F.

AU - Narayanan, Ram Mohan

AU - Buehrer, R. Michael

AU - Sherbondy, Kelly D.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - This work evaluates the performance of a cognitive radar system which predicts and avoids radio frequency interference (RFI) through an alternating renewal process (ARP) model-based and Markov Decision Process (MDP) approach. As radio frequency (RF) environments grow more crowded, the need for such a system becomes necessary. The cognitive radar monitors the RF activity to train a model for RFI prediction and avoidance. By modeling activity as an alternating renewal process, the stochastic approach calculates the likelihood of interference from measured RFI statistics. Alternatively, the MDP uses reinforcement learning to determine the optimal sequence of decisions given measured RF activity. Both methods eventually select the widest radar transmit bandwidth to minimize interference. The performance of each approach is evaluated by the number of collisions and missed opportunities. A hardware implemented test-bed deploys both methods on a set of synthetic and real measured RFI spectra in real-time to compare performance with the goal of determining when each process is more beneficial (in terms of performance and complexity).

AB - This work evaluates the performance of a cognitive radar system which predicts and avoids radio frequency interference (RFI) through an alternating renewal process (ARP) model-based and Markov Decision Process (MDP) approach. As radio frequency (RF) environments grow more crowded, the need for such a system becomes necessary. The cognitive radar monitors the RF activity to train a model for RFI prediction and avoidance. By modeling activity as an alternating renewal process, the stochastic approach calculates the likelihood of interference from measured RFI statistics. Alternatively, the MDP uses reinforcement learning to determine the optimal sequence of decisions given measured RF activity. Both methods eventually select the widest radar transmit bandwidth to minimize interference. The performance of each approach is evaluated by the number of collisions and missed opportunities. A hardware implemented test-bed deploys both methods on a set of synthetic and real measured RFI spectra in real-time to compare performance with the goal of determining when each process is more beneficial (in terms of performance and complexity).

UR - http://www.scopus.com/inward/record.url?scp=85072614289&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072614289&partnerID=8YFLogxK

U2 - 10.1117/12.2519675

DO - 10.1117/12.2519675

M3 - Conference contribution

AN - SCOPUS:85072614289

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Radar Sensor Technology XXIII

A2 - Ranney, Kenneth I.

A2 - Doerry, Armin

PB - SPIE

ER -

Kovarskiy JA, Kozy M, Thornton C, Martone AF, Narayanan RM, Buehrer RM et al. Comparing stochastic and Markov decision process approaches for predicting radio frequency interference. In Ranney KI, Doerry A, editors, Radar Sensor Technology XXIII. SPIE. 2019. 1100318. (Proceedings of SPIE - The International Society for Optical Engineering). https://doi.org/10.1117/12.2519675