Inference-Optimal ISAC via Task-Oriented Feature Transmission and Power Allocation

TL;DR

This paper proposes a transceiver design for ISAC systems that maximizes discriminant gain to improve inference performance, leading to more power-efficient transmission especially at low SNR.

Contribution

It introduces a novel DG-based optimization framework for ISAC transceiver design, providing closed-form solutions and revealing new S extbackslash C tradeoffs.

Findings

DG-optimal design outperforms MSE-based design in power efficiency at low SNR

Closed-form solutions for DG and MSE optimal transceivers are derived

Selective power allocation to informative features enhances sensing performance

Abstract

This work is concerned with the coordination gain in integrated sensing and communication (ISAC) systems under a compress-and-estimate (CE) framework, wherein inference performance is leveraged as the key metric. To enable tractable transceiver design and resource optimization, we characterize inference performance via an error probability bound as a monotonic function of the discriminant gain (DG). This raises the natural question of whether maximizing DG, rather than minimizing mean squared error (MSE), can yield better inference performance. Closed-form solutions for DG-optimal and MSE-optimal transceiver designs are derived, revealing water-filling-type structures and explicit sensing and communication (S\&C) tradeoff. Numerical experiments confirm that DG-optimal design achieves more power-efficient transmission, especially in the low signal-to-noise ratio (SNR) regime, by selectively allocating power to informative features and thus saving transmit power for sensing.