Abstract:Integrated sensing and communications (ISAC) is envisioned as one of the key enablers of next-generation wireless systems, offering improved hardware, spectral, and energy efficiencies. In this paper, we consider an ISAC transceiver with an impaired uniform linear array that performs single-target detection and position estimation, and multiple-input single-output communications. A differentiable model-based learning approach is considered, which optimizes both the transmitter and the sensing receiver in an end-to-end manner. An unsupervised loss function that enables impairment compensation without the need for labeled data is proposed. Semi-supervised learning strategies are also proposed, which use a combination of small amounts of labeled data and unlabeled data. Our results show that semi-supervised learning can achieve similar performance to supervised learning with 98.8% less required labeled data.

What problem does this paper attempt to address?

This paper attempts to address the problem of optimizing the performance of Integrated Sensing and Communication (ISAC) systems in the presence of hardware impairments using semi-supervised and unsupervised learning methods. Specifically, the researchers focus on how to achieve performance comparable to fully supervised learning with a small amount of labeled data and a large amount of unlabeled data in single-target detection and localization as well as multiple-input single-output communication scenarios, while reducing the need for labeled data. The paper proposes a learning method based on a differentiable model, optimizing the end-to-end design of the transmitter and sensing receiver, and introduces an unsupervised loss function to compensate for the impact of hardware impairments. ### Main Issues 1. **Impact of Hardware Impairments**: Existing model-driven signal processing algorithms suffer significant performance degradation when encountering unexpected hardware impairments (e.g., antenna distortion, sampling jitter, or calibration errors). 2. **Need for Labeled Data**: Traditional supervised learning methods require a large amount of labeled data, which can be very difficult and time-consuming to obtain in practical sensing environments. 3. **End-to-End Optimization**: How to optimize the performance of ISAC systems through end-to-end learning methods in the presence of hardware impairments while reducing the reliance on labeled data. ### Solutions 1. **Unsupervised Loss Function**: The paper proposes an unsupervised loss function that can compensate for hardware impairments without labeled data. 2. **Semi-Supervised Learning Strategy**: By combining a small amount of labeled data with a large amount of unlabeled data, the semi-supervised learning method achieves performance comparable to fully supervised learning. 3. **Model-Driven Machine Learning**: Utilizing existing signal processing models to initialize, structure, and train learning methods, reducing the need for labeled data. ### Experimental Results - **Unsupervised Learning**: Unsupervised learning converges quickly in the initial stage but does not achieve the final performance of supervised learning. - **Semi-Supervised Learning**: Semi-supervised learning can achieve performance comparable to fully supervised learning after 1,000 iterations of supervised learning, requiring only 1.2% of labeled data. - **ISAC System Performance**: When jointly optimizing transmission and sensing tasks, the semi-supervised learning method can achieve performance comparable to fully supervised learning with a small amount of labeled data. ### Conclusion The paper demonstrates that the performance of ISAC systems can be effectively optimized using unsupervised and semi-supervised learning methods in the presence of hardware impairments, significantly reducing the need for labeled data. Future work will consider unsupervised learning methods in multi-target sensing scenarios.

Semi-Supervised End-to-End Learning for Integrated Sensing and Communications

End-to-End Learning for SLP-Based ISAC Systems

Distributed Unsupervised Learning for Interference Management in Integrated Sensing and Communication Systems

Integrated Sensing and Communication Beamforming Design Based on Mutual Information

Learn to Sense: A Meta-Learning-Based Sensing and Fusion Framework for Wireless Sensor Networks

ISAC-NET: Model-driven Deep Learning for Integrated Passive Sensing and Communication

A Memory-Based Reinforcement Learning Approach to Integrated Sensing and Communication

Integrated Sensing and Communication in Coordinated Cellular Networks

Multi-Point Integrated Sensing and Communication: Fusion Model and Functionality Selection

Unsupervised Learning for Gain-Phase Impairment Calibration in ISAC Systems

End-to-End Learning for RIS-Aided Communication Systems.

Integrated Sensing and Communications: Recent Advances and Ten Open Challenges

Integrated Sensing and Communication for Symbiotic Radio With Multiple IoT Devices

Integrated Sensing and Communications for 3D Object Imaging via Bilinear Inference

Short-Length Code Designs for Integrated Sensing and Communications Using Deep Learning

Deep-Learning Channel Estimation for IRS-Assisted Integrated Sensing and Communication System

Multicarrier ISAC: Advances in Waveform Design, Signal Processing and Learning under Non-Idealities

End-to-End Learning for Task-Oriented Semantic Communications Over MIMO Channels: An Information-Theoretic Framework