Dynamic Inference Approach Based on Rules Engine in Intelligent Edge Computing for Building Environment Control.

Area of Science:

• Computer Science
• Artificial Intelligence
• Edge Computing

Background:

• Computation offloading addresses hardware limitations in edge computing by distributing tasks.
• Deep learning inference requires substantial data and computing resources, posing deployment challenges at the network edge.
• Deploying domain-specific inference on edge devices enables intelligent services closer to users.

Purpose of the Study:

• To propose an intelligent edge computing framework for building environment control.
• To develop a dynamic inference approach for selecting and executing inference functions on edge gateways.
• To enhance edge gateway capabilities with microservices for flexibility and extensibility.

Main Methods:

• A rules engine on the edge gateway dynamically selects inference functions based on triggered rules.
• Microservices architecture supports comprehensive gateway functions: device management, proxy, client service, intelligent service, and rules engine.
• Deep learning models are trained on edge servers using building user data and deployed as inference models on the edge gateway.

Main Results:

• The proposed system enables intelligent building environment control through dynamic inference.
• Edge gateways effectively manage IoT devices and provide intelligent services on constrained hardware.
• Microservices facilitate flexible updates of inference models and intelligent services.

Conclusions:

• The intelligent edge computing approach with a dynamic inference mechanism is effective for building environment control.
• The microservices-based architecture enhances the flexibility and scalability of edge gateways.
• This framework successfully bridges IoT devices with the internet, offering intelligent services at the network edge.