Goal-Oriented Communication for Real-time Inference Over Networks with Two-way Delay

Date

2025-8

Author

Arı, Çağrı

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

373
views

240
downloads

Cite This

The rise of artificial intelligence (AI) has played a key role in enabling technologies such as industrial robotics and self-driving vehicles. However, these intelligent models often rely on data collected remotely, requiring the network to deliver the relevant data in a timely fashion. Motivated by this, this thesis studies a setting where an intelligent model infers the real-time value of a target signal using data samples transmitted from a remote source. The scheduler decides on i) the age of the samples to be transmitted, ii) the transmission times, and iii) the length of each packet (i.e., the number of samples contained in each transmission). The dependence of inference quality on the Age of Information (AoI) for a given packet length is modeled by a general relationship. Previous work assumed i.i.d. transmission delays with immediate feedback or were restricted to the case where inference performance degrades as the input data ages. In contrast, our formulation captures non-monotonic age dependence and covers Markovian delay processes on both the forward and feedback links. We model this problem as an infinite-horizon average-cost Semi-Markov Decision Process. First, we derive a closed-form solution that decides on (i) and (ii) for a constant packet length whose value is separately optimized. Then, we propose an index-based threshold policy for the variable packet length case, significantly reducing the computational complexity of the dynamic programming solution. Simulation results demonstrate that our goal-oriented scheduler reduces inference error by up to a factor of six compared to age-based scheduling with unit-length packets.

Subject Keywords

Remote Inference, Goal-Oriented Communication, Two-way Markov Delay, Age of Information, Dynamic Programming, Uzaktan Çıkarım/Tahmin, Hedef-Odaklı Haberleşme, İki Yönlü Markov Gecikme, Bilgi Yaşı, Dinamik Programlama

URI

https://hdl.handle.net/11511/115597

Collections

Graduate School of Natural and Applied Sciences, Thesis