Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems

Amna Tahir; Sotirios K. Goudos; Pavlos I. Lazaridis; Panagiotis Sarigiannidis; Thomas Lagkas; Zaharias Zaharis

doi:10.1109/ICT70370.2026.11594720

Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems

Amna Tahir, Sotirios K. Goudos, Pavlos I. Lazaridis, Panagiotis Sarigiannidis, Thomas Lagkas, Zaharias Zaharis

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Integrated sensing and communication (ISAC) enables joint communication and environmental sensing using shared resources. This paper investigates the communication-sensing tradeoff in a reconfigurable intelligent surface (RIS)assisted multi-user MISO system using deep reinforcement learning (DRL). A joint reward formulation balances communication and sensing through a sensing weight parameter. Deep Deterministic Policy Gradient (DDPG) and Twin Delayed DDPG (TD3) are applied for continuous beamforming and RIS configuration. In the communication-only setting, TD3 achieves more stable convergence. Performance is evaluated across communicationdominant, balanced, and sensing-dominant regimes by varying the sensing weight. It is observed that TD3 performs best in balanced scenarios, while DDPG is more effective when sensing is prioritized. These results show that the sensing weight acts as a control parameter for ISAC optimization.

Original language	English
Title of host publication	2026 32nd International Conference on Telecommunications (ICT)
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	166-169
Number of pages	4
ISBN (Electronic)	9798319546012
ISBN (Print)	9798319546029
DOIs	https://doi.org/10.1109/ICT70370.2026.11594720
Publication status	Published - 9 Jul 2026
Event	32nd International Conference on Telecommunications - Thessaloniki, Greece Duration: 20 May 2026 → 22 May 2026 https://ict2026thessaloniki.org/

Publication series

Name	International Conference on Telecommunications (ICT)
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISSN (Print)	2771-4233
ISSN (Electronic)	2993-1916

Conference

Conference	32nd International Conference on Telecommunications
Abbreviated title	ICT 2026
Country/Territory	Greece
City	Thessaloniki
Period	20/05/26 → 22/05/26
Internet address	https://ict2026thessaloniki.org/

Access to Document

10.1109/ICT70370.2026.11594720Licence: Unspecified

Cite this

Tahir, A., Goudos, S. K., Lazaridis, P. I., Sarigiannidis, P., Lagkas, T., & Zaharis, Z. (2026). Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems. In 2026 32nd International Conference on Telecommunications (ICT) (pp. 166-169). Article 11594720 (International Conference on Telecommunications (ICT)). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICT70370.2026.11594720

@inproceedings{d155c8b862c74b3aa610074c24e2cc7f,

title = "Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems",

abstract = "Integrated sensing and communication (ISAC) enables joint communication and environmental sensing using shared resources. This paper investigates the communication-sensing tradeoff in a reconfigurable intelligent surface (RIS)assisted multi-user MISO system using deep reinforcement learning (DRL). A joint reward formulation balances communication and sensing through a sensing weight parameter. Deep Deterministic Policy Gradient (DDPG) and Twin Delayed DDPG (TD3) are applied for continuous beamforming and RIS configuration. In the communication-only setting, TD3 achieves more stable convergence. Performance is evaluated across communicationdominant, balanced, and sensing-dominant regimes by varying the sensing weight. It is observed that TD3 performs best in balanced scenarios, while DDPG is more effective when sensing is prioritized. These results show that the sensing weight acts as a control parameter for ISAC optimization.",

keywords = "Integrated Sensing and Communication, Reconfigurable Intelligent Surfaces, Deep Reinforcement Learning, Deep Deterministic Policy Gradient, Twin Delayed Deep Deterministic Policy Gradient, Beamforming Optimization, Communication-Sensing Tradeoff",

author = "Amna Tahir and Goudos, \{Sotirios K.\} and Lazaridis, \{Pavlos I.\} and Panagiotis Sarigiannidis and Thomas Lagkas and Zaharias Zaharis",

note = "Publisher Copyright: {\textcopyright} 2026 IEEE.; 32nd International Conference on Telecommunications, ICT 2026 ; Conference date: 20-05-2026 Through 22-05-2026",

year = "2026",

month = jul,

day = "9",

doi = "10.1109/ICT70370.2026.11594720",

language = "English",

isbn = "9798319546029",

series = "International Conference on Telecommunications (ICT)",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "166--169",

booktitle = "2026 32nd International Conference on Telecommunications (ICT)",

address = "United States",

url = "https://ict2026thessaloniki.org/",

}

Tahir, A, Goudos, SK, Lazaridis, PI, Sarigiannidis, P, Lagkas, T & Zaharis, Z 2026, Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems. in 2026 32nd International Conference on Telecommunications (ICT)., 11594720, International Conference on Telecommunications (ICT), Institute of Electrical and Electronics Engineers Inc., pp. 166-169, 32nd International Conference on Telecommunications, Thessaloniki, Greece, 20/05/26. https://doi.org/10.1109/ICT70370.2026.11594720

Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems. / Tahir, Amna; Goudos, Sotirios K.; Lazaridis, Pavlos I. et al.
2026 32nd International Conference on Telecommunications (ICT). Institute of Electrical and Electronics Engineers Inc., 2026. p. 166-169 11594720 (International Conference on Telecommunications (ICT)).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems

AU - Tahir, Amna

AU - Goudos, Sotirios K.

AU - Lazaridis, Pavlos I.

AU - Sarigiannidis, Panagiotis

AU - Lagkas, Thomas

AU - Zaharis, Zaharias

PY - 2026/7/9

Y1 - 2026/7/9

N2 - Integrated sensing and communication (ISAC) enables joint communication and environmental sensing using shared resources. This paper investigates the communication-sensing tradeoff in a reconfigurable intelligent surface (RIS)assisted multi-user MISO system using deep reinforcement learning (DRL). A joint reward formulation balances communication and sensing through a sensing weight parameter. Deep Deterministic Policy Gradient (DDPG) and Twin Delayed DDPG (TD3) are applied for continuous beamforming and RIS configuration. In the communication-only setting, TD3 achieves more stable convergence. Performance is evaluated across communicationdominant, balanced, and sensing-dominant regimes by varying the sensing weight. It is observed that TD3 performs best in balanced scenarios, while DDPG is more effective when sensing is prioritized. These results show that the sensing weight acts as a control parameter for ISAC optimization.

AB - Integrated sensing and communication (ISAC) enables joint communication and environmental sensing using shared resources. This paper investigates the communication-sensing tradeoff in a reconfigurable intelligent surface (RIS)assisted multi-user MISO system using deep reinforcement learning (DRL). A joint reward formulation balances communication and sensing through a sensing weight parameter. Deep Deterministic Policy Gradient (DDPG) and Twin Delayed DDPG (TD3) are applied for continuous beamforming and RIS configuration. In the communication-only setting, TD3 achieves more stable convergence. Performance is evaluated across communicationdominant, balanced, and sensing-dominant regimes by varying the sensing weight. It is observed that TD3 performs best in balanced scenarios, while DDPG is more effective when sensing is prioritized. These results show that the sensing weight acts as a control parameter for ISAC optimization.

KW - Integrated Sensing and Communication

KW - Reconfigurable Intelligent Surfaces

KW - Deep Reinforcement Learning

KW - Deep Deterministic Policy Gradient

KW - Twin Delayed Deep Deterministic Policy Gradient

KW - Beamforming Optimization

KW - Communication-Sensing Tradeoff

UR - https://www.scopus.com/pages/publications/105045427959

UR - https://ieeexplore.ieee.org/xpl/conhome/11594638/proceeding

U2 - 10.1109/ICT70370.2026.11594720

DO - 10.1109/ICT70370.2026.11594720

M3 - Conference contribution

AN - SCOPUS:105045427959

SN - 9798319546029

T3 - International Conference on Telecommunications (ICT)

SP - 166

EP - 169

BT - 2026 32nd International Conference on Telecommunications (ICT)

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 32nd International Conference on Telecommunications

Y2 - 20 May 2026 through 22 May 2026

ER -

Tahir A, Goudos SK, Lazaridis PI, Sarigiannidis P, Lagkas T, Zaharis Z. Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems. In 2026 32nd International Conference on Telecommunications (ICT). Institute of Electrical and Electronics Engineers Inc. 2026. p. 166-169. 11594720. (International Conference on Telecommunications (ICT)). doi: 10.1109/ICT70370.2026.11594720

Deep Reinforcement Learning-based Communication-Sensing Tradeoff Optimization in RIS-Assisted ISAC Systems

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this