Ultra-Low-Latency Edge Intelligent Sensing: A Source-Channel Tradeoff and Its Application to Coding Rate Adaptation

TL;DR

This work establishes a theoretical foundation for the source-channel tradeoff by quantifying the effects of source coding on feature discriminant gains and channel reliability on packet loss, and designs the coding rate control by optimizing the tradeoff to minimize the E2E sensing error probability, leading to a low-complexity algorithm for ultra-low-latency EI-Sense.

Abstract

The forthcoming sixth-generation (6G) mobile network is set to merge edge artificial intelligence (AI) and integrated sensing and communication (ISAC) extensively, giving rise to the new paradigm of edge intelligent sensing (EI-Sense). This paradigm leverages ubiquitous edge devices for environmental sensing and deploys AI algorithms at edge servers to interpret the observations via remote inference on wirelessly uploaded features. A significant challenge arises in designing EI-Sense systems for 6G mission-critical applications, which demand high performance under stringent latency constraints. To tackle this challenge, we focus on the end-to-end (E2E) performance of EI-Sense and characterize a source-channel tradeoff that balances source distortion and channel reliability. In this work, we establish a theoretical foundation for the source-channel tradeoff by quantifying the effects of source coding on feature discriminant gains and channel reliability on packet loss. Building on this foundation, we design the coding rate control by optimizing the tradeoff to minimize the E2E sensing error probability, leading to a low-complexity algorithm for ultra-low-latency EI-Sense. Finally, we validate our theoretical analysis and proposed coding rate control algorithm through extensive experiments on both synthetic and real datasets, demonstrating the sensing performance gain of our approach with respect to traditional reliability-centric methods.

Figures (9)

• Figure 1: An illustration of an E2E ultra-low-latency EI-Sense system. (a) System architecture and operations at the sensor and edge server. (b) Parallelization between sensing and communication in the sequential sensing scenario.
• Figure 2: An illustration of statistical inference. (a) Gaussian Mixture Model with the red line indicating the decision boundary. (b) Distribution of the discriminant score function fitted with two Gaussian distributions.
• Figure 3: Histogram of quantization noise and source distortion for a single feature. The quantization level is set to 4 bits. (a) Quantization error, defined as the difference between the input and output of the uniform scalar quantizer. (b) Feature error, defined as the difference between the decoded feature and the original feature. The data is fitted using a zero-mean Gaussian distribution.
• Figure 4: Comparison between the exact and approximated average packet loss probability as a function of (a) transmit SNR and (b) coding rate. The parameters are set to $L=4$ and $N=100$.
• Figure 5: An illustration of the source-channel tradeoff. Both the effective discriminant gain and the number of successfully received observations are plotted as a function of the coding rate. The parameters are set to $L=2$, $\gamma_0=1$ dB, $N=100$, $d=50$, $D_0=1$, $U=5$, and $K=20$.

Theorems & Definitions (15)

• Lemma 1: Block quantization noise
• proof
• Theorem 1: Discriminant gain reduction
• proof
• Theorem 2: Sensing performance
• proof
• Proposition 1: Average packet loss probability
• proof
• Lemma 2: Monotonicity of discriminant gain and packet loss probability
• Lemma 3: Concavity of effective discriminant gain