CoTwin: Collaborative improvement of digital twins enabled by blockchain

Integrating digital twin technology in Cyber-Physical Systems and Internet of Things can boost their intelligence. Given the current maturity of digital twin technology (yet in progress), improving the models that these systems use is typically achieved off-line, requiring the system to stop and reconfigure to run each new model version. In fact, most works use cloud back -ends to run heavy machine learning algorithms, imposing strict requirements on the data exchange between the physical system and the cloud. We address the online improvement of digital twin models in cyber-physical systems by supporting model refinement without disrupting nor stopping the normal operation of the system. This improves the dynamicity of the system that may turn into a major competitive advantage in a number of industrial scenarios. Precisely, we exploit the collaborative expectation of next generation cyber-physical systems based on highly -connected cells enabled by 5G and 6G networking; and on top of these, we design a shared space properly managed to deliver the needed temporal behaviour required by cyber-physical systems. For this, we present the design of CoTwin framework as a middleware that allows cells to collectively improve digital twin models seamlessly. CoTwin manages the interaction of cells with a blockchain-based collaborative space offering a built-in trusted storage model. We integrate neural network algorithms as they provide fast execution that meet the time -sensitivity requirements of cyber-physical systems. Our contribution is validated by means of its implementation and deployment on an actual blockchain network, and an exhaustive set of experiments to analyse the resulting overhead and temporal behaviour. Results show that CoTwin achieves stable execution times across all its functional pieces; and it exhibits stable service time for large sets of cells.