6G Orchestration of AI Tasks for IoT

Important results from the project

The project successfully met its objective by implementing and evaluating a custom WebAssembly orchestrator on edge devices. The results are promising, showing a 7% performance improvement in average completion time. These results confirm that application-level orchestration is an efficient method for optimizing resource usage on hardware with limited capacity. Another valuable results was the development of a custom test suite for WebAssembly components.

Expected long term effects

The project significantly advances edge computing by demonstrating that WebAssembly is a secure, efficient, and promising alternative to traditional containers. Long-term, this empowers IoT devices to handle complex AI and data processing locally. This shift reduces reliance on cloud infrastructure, lowers latency and bandwidth usage, and supports the transition toward more sustainable, decentralized computing architectures.

Approach and implementation

The project was conducted using a rigorous quantitative research methodology. It began with a literature review and the creation of specific WebAssembly tasks, including machine learning inference. A custom orchestrator was developed in Rust, iteratively integrating enhancements for core pinning, memory management, and task sorting. Experiments were performed on a Raspberry Pi 5 with artificial resource constraints, yielding clear and valuable data that validated the system´s performance.