Optimizing 6G Integrated Sensing and Communications (ISAC) via Expert Networks

Jiacheng Wang,Hongyang Du,Geng Sun,Jiawen Kang,Haibo Zhou,Dusit Niyato,Jiming Chen

2024-06-01

Abstract:Integrated Sensing and Communications (ISAC) is one of the core technologies of 6G, which combines sensing and communication functions into a single system. However, limited computing and storage resources make it impractical to combine multiple sensing models into a single device, constraining the system's function and performance. Therefore, this article proposes enhancing ISAC with the mixture of experts (MoE) architecture. Rigorously, we first investigate ISAC and MoE, including their concepts, advantages, and applications. Then, we explore how MoE can enhance ISAC from the perspectives of signal processing and network optimization. Building on this, we propose an MoE based ISAC framework, which uses a gating network to selectively activate multiple experts in handling sensing tasks under given communication conditions, thereby improving the overall performance. The case study demonstrates that the proposed framework can effectively increase the accuracy and robustness in detecting targets by using wireless communication signal, providing strong support for the practical deployment and applications of the ISAC system.

Signal Processing

What problem does this paper attempt to address?

The paper attempts to address the issue in 6G Integrated Sensing and Communication (ISAC) systems where limited computational and storage resources prevent the deployment and execution of multiple sensing models on a single device, thereby restricting the system's functionality and performance. Specifically, the user's needs are unpredictable, and the types and numbers of models on a single ISAC device are limited, leading to a lack of corresponding models or insufficient communication rates to support the effective operation of the models when handling specific tasks.

To solve these problems, the paper proposes an ISAC framework based on a Mixture of Experts (MoE) architecture. By introducing a gating network to select and activate multiple experts (i.e., sub-models), this framework can efficiently handle sensing tasks under different communication conditions, thereby improving the overall performance and robustness of the system. The specific contributions include:

1. **Analyzing the potential support of MoE for ISAC systems**: The paper analyzes from multiple perspectives how MoE can optimize ISAC systems, providing new insights.
2. **Proposing an MoE-enhanced ISAC framework**: This framework can dynamically select appropriate experts based on current communication conditions, input data, and other relevant information to handle user requests, thereby improving the system's performance and robustness.
3. **Evaluating the framework through case studies**: Experimental results of target detection based on Channel State Information (CSI) show that under low communication rates, the detection accuracy of this framework is at least 18% higher than that of a single expert.

Overall, the paper aims to address the resource limitations and insufficient multi-task processing capabilities in ISAC systems in practical applications by introducing the MoE architecture, thereby enhancing the overall performance and adaptability of the system.