Konsep Inti
Integrated sensing, communication, and computation is of paramount significance for improving resource utilization and achieving customized goals of edge AI tasks, such as federated edge learning and edge AI inference.
Abstrak
The article presents the vision, motivations, and principles of ISCC designs for edge AI tasks from both the task-level resource management and the physical-layer waveform design perspectives.
For federated edge learning (FEEL) tasks:
- Digital ISCC for FEEL involves sequential data acquisition, local model training, and global model aggregation. Challenges include characterizing the impact of sensing, communication, and computation on learning performance, and developing task-oriented resource allocation schemes.
- Analog ISCC for FEEL exploits over-the-air computation to enable simultaneous data sensing and model aggregation, requiring synchronization of sensing and communication processes and transceiver design to align signals for efficient aggregation while suppressing interference.
For edge AI inference tasks:
- An ISCC-based edge inference framework is introduced, which adopts inference accuracy as the design goal and models the distortion caused by sensing, computation, and communication on the received feature vector.
- ISCC schemes are proposed for multi-device edge inference, considering both narrow-view sensing (with task-oriented resource management) and wide-view sensing (with accuracy-oriented broadband AirComp).
Physical layer ISCC techniques, such as beamforming design for dual-functional and triple-functional signals, are investigated to fully exploit radio resources and support edge AI applications.
Statistik
The implementation of edge AI tasks requires the fusion of physical, biological, and cyber worlds, involving sensing, communication, and computation.
Edge AI features a task-oriented property that concerns the effectiveness and efficiency instead of traditional design criteria like system throughput and SNR.
The design complexity of ISCC schemes is much higher than that of the separated schemes due to the joint design of the highly coupled sensing, communication, and computation modules.
Kutipan
"Integrated sensing-communication-computation (ISCC) is of paramount significance for improving resource utilization as well as achieving the customized goals of edge AI tasks."
"The ISCC schemes enjoy the benefits of better network resource coordination among the three modules and hardware sharing between sensing and communication on devices for saving their physical spaces."