Optimization for wireless communication system is a hot topic, consisting of many sub-topics. In this work, what we focus on is the balance between communication performance & inference accuracy in ISAC.

we aim to propose an enhanced wireless communication system via Semantic inpainting (Semantic inpainting in wireless communications), as we noticed that there is almost no such existing work in the field of wireless communication. The propose of such an ICT infrastructure, for communication & sensing via semantic inpainting as a more solid step.

jepa-based joint embeddning

for the next-generatioon communication system

to reach the goal, we need to model the system carefully (we did not achieve in the last work in details), especially, how can we evaluate the communication cost, details must be given.

Since systematic orchestration is required for integrated sensing and communication, we have to face questions like:

model training strategy (data from all the sensors must be aggregated)

fine-tuning design for high-accuracy maintenance (when? who? how?)

The training & fine-tuning asks data aggregation from multiple sensory terminals, leading to inevitable increased communication overhead or traffic. Therefore, there is a trade-off between sensing accuracy and communication cost.