Multimodal Virtual Semantic Communication for Tiny-Machine-Learning-Based UAV Task Execution

Chao Ren

doi:10.1109/JIOT.2024.3416253

Return

Multimodal Virtual Semantic Communication for Tiny-Machine-Learning-Based UAV Task Execution

2024-10-01

PRE

DOI：10.1109/JIOT.2024.3416253

Original

Original request for help

Save

Abstract

En 中文

In the 6G integrated air-ground network, the process of accomplishing complex tasks through the integrated multimodal communication faces challenges induced by unmanned aerial vehicles (UAVs), such as limited communication, storage and computing capabilities, and the existence of heterogeneous UAV multimodal information and carriers. Inspired by the process of semantic communication, we view successful execution of advanced UAV tasks as semantic recognition and pragmatic execution. Tiny machine learning (TinyML) provides the UAV advanced algorithms and models that can be run on the low-power and resource-constrained platforms. In this article, from the perspective of semantic communication and leveraging the applicability of TinyML for UAVs, we map the heterogeneous multimodal communication and UAV task execution processes aiming to better utilize the capabilities of machine learning and semantic communication to enhance the pragmatic task execution of UAVs. Multimodal virtual semantic communication can provide task-related auxiliary information, enabling the complementary integration of multiple independent modalities in the task domain. The proposed scheme and model achieve a deep integration of communication, sensation, and computation ultimately enhancing the practical task execution capability of UAVs.

Keywords:

Task analysis

Semantics

Autonomous aerial vehicles

Wireless communication

Computational modeling

Wireless sensor networks

Vectors

Multimodal communication

semantic communication

tiny machine learning (TinyML)

wireless unmanned aerial vehicle (UAV) communication