[2603.02109] Orchestrating Multimodal DNN Workloads in Wireless Neural Processing

Electrical Engineering and Systems Science > Signal Processing arXiv:2603.02109 (eess) [Submitted on 2 Mar 2026] Title:Orchestrating Multimodal DNN Workloads in Wireless Neural Processing Authors:Sai Xu, Kai-Kit Wong, Yanan Du, Hyundong Shin View a PDF of the paper titled Orchestrating Multimodal DNN Workloads in Wireless Neural Processing, by Sai Xu and 3 other authors View PDF HTML (experimental) Abstract:In edge inference, wireless resource allocation and accelerator-level deep neural network (DNN) scheduling have yet to be co-optimized in an end-to-end manner. The lack of coordination between wireless transmission and accelerator-level DNN execution prevents efficient overlap, leading to higher end-to-end inference latency. To address this issue, this paper investigates multimodal DNN workload orchestration in wireless neural processing (WNP), a paradigm that integrates wireless transmission and multi-core accelerator execution into a unified end-to-end pipeline. First, we develop a unified communication-computation model for multimodal DNN execution and formulate the corresponding optimization problem. Second, we propose O-WiN, a framework that orchestrates DNN workloads in WNP through two tightly coupled stages: simulation-based optimization and runtime execution. Third, we develop two algorithms, RTFS and PACS. RTFS schedules communication and computation sequentially, whereas PACS interleaves them to enable pipeline parallelism by overlapping wireless data transfer...