On the View-and-Channel Aggregation Gain in Integrated Sensing and Edge AI

Sensing and edge
artificial intelligence
(AI) are two key features of the
sixth-generation
(6G) mobile networks. Their natural integration, termed
Integrated sensing and edge AI
(ISEA), is envisioned to automate wide-ranging
Internet-of-Tings
(IoT) applications. To achieve a high sensing accuracy, features of multiple sensor views are uploaded to an edge server for aggregation and inference using a large-scale AI model. The view aggregation is realized efficiently using over-the-air computing (AirComp), which also aggregates channels to suppress channel noise. As ISEA is at its nascent stage, there still lacks an analytical framework for quantifying the fundamental performance gains from view-and-channel aggregation, which motivates this work. Our framework is based on a well-established distribution model of multi-view sensing data where the classic Gaussian-mixture model is modified by adding sub-spaces matrices to represent individual sensor observation perspectives. Based on the model and linear classification, we study the End-to-End sensing (inference) uncertainty, a popular measure of inference accuracy, of the said ISEA system by a novel, tractable approach involving designing a scaling-tight uncertainty surrogate function, global discriminant gain, distribution of receive Signal-to-Noise Ratio (SNR), and channel induced discriminant loss. As a result, we prove that the E2E sensing uncertainty diminishes at an
exponential
rate as the number of views/sensors grows, where the rate is proportional to global discriminant gain. Given AirComp and channel distortion, we further show that the exponential scaling remains but the rate is reduced by a linear factor representing the channel induced discriminant loss. Furthermore, in the case of many spatial degrees of freedom, we benchmark AirComp against equally fast, traditional analog orthogonal access. The comparative performance analysis reveals a sensing-accuracy crossing point between the schemes corresponding to equal receive array size and sensor number. This leads to the proposal of a scheme for adaptive access-mode switching to enhance ISEA performance. Last, the insights from our framework are validated by experiments using a convolutional neural network model and real-world dataset.