Reducing Latency for Radio Co-Processor Design with Dynamic Network Stack Offloading and Spinel Optimization

With the continuous increase in communication demands and the constant advancement of information network technology, the Internet of Things is being increasingly applied across various fields of life and production. Some works give a certain contribution to it, such as Thread. It supports a Co-Processor (RCP) design to enhance the transmission between nodes, combining with Spinel protocol, which is a general management protocol aiming at enabling interoperability between simple and complex devices. However, it doesn't work very well in some scenarios like high packet loss rate environments. We propose a solution that dynamically offload part of the computational tasks between the two chips in the Radio Co-Processor architecture and give an algorithm to optimize a specific parameters in the Spinel protocol to give a positive effect on reducing latency. We evaluate the mechanism with true nodes. It can reduce the latency by about 8%, which is often crucial, especially in scenarios with a large number of tasks.