WECAR: An End-Edge Collaborative Inference and Training Framework for WiFi-Based Continuous Human Activity Recognition

2025-03-09

AI Summary

To address the triple challenges of dynamic scenario adaptation, catastrophic forgetting, and stringent computational constraints in WiFi-based continuous human activity recognition on edge devices, this paper proposes an end-edge collaborative dynamic continual learning framework. Methodologically: (1) we design a task-adaptive model expansion mechanism coupled with stability-aware selective retraining to enable incremental learning while preserving previously acquired knowledge; (2) we introduce a two-stage hierarchical relation distillation—comprising multi-head self-attention relation distillation and prefix relation distillation—to compress the model while retaining discriminative capacity. Built upon a Transformer architecture, our framework is validated on heterogeneous hardware platforms (Jetson Nano for edge and ESP32 for end devices). Experiments across three public WiFi datasets demonstrate state-of-the-art performance: model parameters are reduced by 42%–68% without accuracy degradation, significantly enhancing the practicality and efficiency of continual learning at the edge.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

WiFi-based human activity recognition (HAR) holds significant promise for ubiquitous sensing in smart environments. A critical challenge lies in enabling systems to dynamically adapt to evolving scenarios, learning new activities without catastrophic forgetting of prior knowledge, while adhering to the stringent computational constraints of edge devices. Current approaches struggle to reconcile these requirements due to prohibitive storage demands for retaining historical data and inefficient parameter utilization. We propose WECAR, an end-edge collaborative inference and training framework for WiFi-based continuous HAR, which decouples computational workloads to overcome these limitations. In this framework, edge devices handle model training, lightweight optimization, and updates, while end devices perform efficient inference. WECAR introduces two key innovations, i.e., dynamic continual learning with parameter efficiency and hierarchical distillation for end deployment. For the former, we propose a transformer-based architecture enhanced by task-specific dynamic model expansion and stability-aware selective retraining. For the latter, we propose a dual-phase distillation mechanism that includes multi-head self-attention relation distillation and prefix relation distillation. We implement WECAR based on heterogeneous hardware using Jetson Nano as edge devices and the ESP32 as end devices, respectively. Our experiments across three public WiFi datasets reveal that WECAR not only outperforms several state-of-the-art methods in performance and parameter efficiency, but also achieves a substantial reduction in the model's parameter count post-optimization without sacrificing accuracy. This validates its practicality for resource-constrained environments.

Problem

• Enable dynamic adaptation to evolving scenarios in WiFi-based HAR.
• Prevent catastrophic forgetting while learning new activities on edge devices.
• Achieve efficient parameter utilization and reduce storage demands.

Innovation

• Dynamic continual learning with parameter efficiency
• Hierarchical distillation for end deployment
• End-edge collaborative inference and training