Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CrowdTransfer: Enabling Crowd Knowledge Transfer in AIoT Community

Yan Liu, Bin Guo, Nuo Li, Yasan Ding, Zhouyangzi Zhang, Zhiwen Yu

TL;DR

This paper introduces CrowdTransfer, a crowd-based knowledge transfer framework for AIoT that addresses resource constraints, dynamic environments, and incremental tasks. It defines four crowd-transfer modes—derivation, sharing, evolution, and fusion—and presents a general framework decomposed into intra-agent, decentralized inter-agent, and centralized inter-agent transfer, with applications across HAR, urban computing, connected vehicles, multi-robot systems, and smart factories. It surveys core transfer techniques (domain adaptation, domain generalization, multi-task learning, knowledge distillation, meta-learning) and maps them to CrowdTransfer models (FTL, teacher-student, LLM fine-tuning, CTDE-based MARL, distributed IL and DFL, multimodal and continual learning, test-time adaptation). The paper discusses open issues—cognitive foundations, transferability metrics, resource-constrained learning, security, continuous evolution, and hybrid human–machine intelligence—and outlines future directions for robust, privacy-preserving, and continually evolving AIoT crowds. Overall, CrowdTransfer offers a principled path to scalable, self-learning AIoT ecosystems that harness collective intelligence to improve performance and reduce learning costs in real-world deployments.

Abstract

Artificial Intelligence of Things (AIoT) is an emerging frontier based on the deep fusion of Internet of Things (IoT) and Artificial Intelligence (AI) technologies. Although advanced deep learning techniques enhance the efficient data processing and intelligent analysis of complex IoT data, they still suffer from notable challenges when deployed to practical AIoT applications, such as constrained resources, and diverse task requirements. Knowledge transfer is an effective method to enhance learning performance by avoiding the exorbitant costs associated with data recollection and model retraining. Notably, although there are already some valuable and impressive surveys on transfer learning, these surveys introduce approaches in a relatively isolated way and lack the recent advances of various knowledge transfer techniques for AIoT field. This survey endeavors to introduce a new concept of knowledge transfer, referred to as Crowd Knowledge Transfer (CrowdTransfer), which aims to transfer prior knowledge learned from a crowd of agents to reduce the training cost and as well as improve the performance of the model in real-world complicated scenarios. Particularly, we present four transfer modes from the perspective of crowd intelligence, including derivation, sharing, evolution and fusion modes. Building upon conventional transfer learning methods, we further delve into advanced crowd knowledge transfer models from three perspectives for various AIoT applications. Furthermore, we explore some applications of AIoT areas, such as human activity recognition, urban computing, multi-robot system, and smart factory. Finally, we discuss the open issues and outline future research directions of knowledge transfer in AIoT community.

CrowdTransfer: Enabling Crowd Knowledge Transfer in AIoT Community

TL;DR

This paper introduces CrowdTransfer, a crowd-based knowledge transfer framework for AIoT that addresses resource constraints, dynamic environments, and incremental tasks. It defines four crowd-transfer modes—derivation, sharing, evolution, and fusion—and presents a general framework decomposed into intra-agent, decentralized inter-agent, and centralized inter-agent transfer, with applications across HAR, urban computing, connected vehicles, multi-robot systems, and smart factories. It surveys core transfer techniques (domain adaptation, domain generalization, multi-task learning, knowledge distillation, meta-learning) and maps them to CrowdTransfer models (FTL, teacher-student, LLM fine-tuning, CTDE-based MARL, distributed IL and DFL, multimodal and continual learning, test-time adaptation). The paper discusses open issues—cognitive foundations, transferability metrics, resource-constrained learning, security, continuous evolution, and hybrid human–machine intelligence—and outlines future directions for robust, privacy-preserving, and continually evolving AIoT crowds. Overall, CrowdTransfer offers a principled path to scalable, self-learning AIoT ecosystems that harness collective intelligence to improve performance and reduce learning costs in real-world deployments.

Abstract

Artificial Intelligence of Things (AIoT) is an emerging frontier based on the deep fusion of Internet of Things (IoT) and Artificial Intelligence (AI) technologies. Although advanced deep learning techniques enhance the efficient data processing and intelligent analysis of complex IoT data, they still suffer from notable challenges when deployed to practical AIoT applications, such as constrained resources, and diverse task requirements. Knowledge transfer is an effective method to enhance learning performance by avoiding the exorbitant costs associated with data recollection and model retraining. Notably, although there are already some valuable and impressive surveys on transfer learning, these surveys introduce approaches in a relatively isolated way and lack the recent advances of various knowledge transfer techniques for AIoT field. This survey endeavors to introduce a new concept of knowledge transfer, referred to as Crowd Knowledge Transfer (CrowdTransfer), which aims to transfer prior knowledge learned from a crowd of agents to reduce the training cost and as well as improve the performance of the model in real-world complicated scenarios. Particularly, we present four transfer modes from the perspective of crowd intelligence, including derivation, sharing, evolution and fusion modes. Building upon conventional transfer learning methods, we further delve into advanced crowd knowledge transfer models from three perspectives for various AIoT applications. Furthermore, we explore some applications of AIoT areas, such as human activity recognition, urban computing, multi-robot system, and smart factory. Finally, we discuss the open issues and outline future research directions of knowledge transfer in AIoT community.
Paper Structure (61 sections, 5 equations, 19 figures, 5 tables)

This paper contains 61 sections, 5 equations, 19 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Embracing Crowd Knowledge Transfer in AIoT
  3. Preliminary of Transfer Learning
  4. Overview of AIoT
  5. Definition of CrowdTransfer
  6. Insights of CrowdTransfer
  7. Framework of CrowdTransfer for AIoT
  8. AIoT Agent
  9. CrowdTransfer in AIoT
  10. AIoT Applications
  11. KEY TECHNIQUES OF KNOWLEDGE TRANSFER
  12. Domain Adaptation
  13. Domain Generalization
  14. Multi-task Learning
  15. Knowledge Distillation
  16. ...and 46 more sections

Figures (19)

  • Figure 2: An overview of AIoT.
  • Figure 3: Four CrowdTransfer modes.
  • Figure 4: The general framework of CrowdTransfer for AIoT.
  • Figure 5: The general framework of Federated Transfer Learning.
  • Figure 6: The framework of Fine-tuning of Large Language Models.
  • ...and 14 more figures

Theorems & Definitions (7)

  • Definition 1: Domain
  • Definition 2: Task
  • Definition 3: Transfer Learning
  • Definition 4: AIoT
  • Definition 5: AIoT Agent
  • Definition 6: AIoT Context
  • Definition 7: Crowd Knowledge Transfer in AIoT