Multi-objective Optimization for Over-the-Air Federated Edge Learning-enabled Collaborative Integrated Sensing and Communications

TL;DR

This paper proposes a multi-objective framework combining over-the-air federated learning and collaborative sensing, improving sensing accuracy while maintaining learning performance through novel signal processing and optimization techniques.

Contribution

It introduces a new integrated sensing and communications framework with a novel pulse shaping method and a multi-objective optimization approach for joint sensing and learning.

Findings

Enhances sensing accuracy without harming federated learning performance.

Achieves the Cramer-Rao bound for sensing error under the proposed framework.

Demonstrates improved multi-task performance through numerical simulations.

Abstract

This paper introduces a novel multi-objective integrated sensing and communications (ISAC) framework to enable collaborative wireless sensing in conjunction with over-the-air federated-edge learning (OTA-FEEL). The framework enables multi-task OTA aggregation to handle sensing and learning simultaneously, while benefiting from dual-purpose uplink signals for both communications and target sensing. Starting from characterizing the local sufficient statistics at each edge device and establishing its stationary, we develop a tractable analytical expression for the local sufficient statistics. To suppress the interference from uplink transmissions of other devices through matched filtering, we then propose a novel orthogonal pulse shaping method. Then, we derive the optimal unbiased estimate of the target's coordinates by casting the centralized problem of joint likelihood function maximization of all devices as the distributed likelihood maximization of each device (which requires only local sufficient statistics). A lower bound on the sensing error variance is then characterized using the Cramer-Rao bound (CRB). We then formulate a multi-objective optimization (MOOP) problem to minimize the mean square error (MSE) and sensing error bound simultaneously. The considered problem is then solved using the epsilon-constrained method. Numerical results demonstrate that the proposed dual-purpose OTA-FEEL-enabled collaborative ISAC framework enhances sensing accuracy without adversely affecting the performance of the primary OTA-FEEL task. While conventional single-shot collaborative sensing schemes are limited by the average error of local estimators, the proposed algorithm achieves the CRB of the considered problem.