Performance-Complexity Tradeoff for ISAC Transceiver Design: A Deep Unfolding Method

• Published in: IEEE International Conference on Communications (2003) pp. 4433 - 4438
• Main Authors: Zhang, Jifa, Zhu, Yongxu, Zhao, Nan, Jin, Shi, Wang, Xianbin, Ng, Derrick Wing Kwan, Al-Dhahir, Naofal
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.06.2025
• Subjects: ADMM; Artificial neural networks; Benchmark testing; Computational complexity; deep unfolding; Integrated sensing and communication; Interference; ISAC; Iterative methods; Optimization; RIS; Signal to noise ratio; Simulation; Transceivers; waveform design
• ISSN: 1938-1883
• DOI: 10.1109/ICC52391.2025.11161873

Integrated sensing and communication (ISAC) can boost the spectrum efficiency and facilitate the diverse emerging applications via sharing the same spectrum and hardware between communication and sensing. However, it may suffer from high complexity. In this paper, we develop a low-complexity deep unfolding learning aided transceiver design for ISAC. Particularly, the weighted sum of multi-user interference power and the reciprocal of sensing signal-to-interference-plus-noise ratio is minimized subject to the constraints of constant modulus signal and waveform similarity by transceiver design. An alternating direction method of multipliers (ADMM)-based iterative algorithm is first developed to solve this non-convex optimization problem. To reduce the complexity, we propose a deep unfolding neural network (NN), which can unfold the underlying ADMMbased iterative algorithm to a lightweight NN with some learnable parameters and circumvent the bisection method using the projected gradient descent. Simulation results demonstrate the effectiveness of our proposed deep unfolding NN.