Following a research initiative, researchers at Frankfurt School and ETH Zurich have developed an artificial neural network for solving challenging control problems. Optimization of production processes and supply chains are some applications the self-learning system can be used for as well as for traffic control systems and smart grids.
For industrial processes, supply chain disruptions, power cuts, and financial network failures are some among many problems that are typically encountered in complex systems. These challenges are difficult or even impossible to control using existing techniques.
Artificial Intelligence(AI)-based control systems can help to optimize complex processes and can be used to develop new business models.
The team of researchers have developed a versatile AI-based control system named AI Pontryagin which is created to maneuver complex systems and networks toward desired target states. The use of combination of analytical and numerical methods led researchers to demonstrate how the AI-based method automatically learns to control work systems in a near-optimal ways.
Meanwhile, disruptions in complex systems can trigger cascades and blackouts. In order to avoid this and improve resilience of work systems, a wide spectrum of control mechanisms and regulations devised by system specialists. Some typical applications of the system include stress testing in financial organizations and voltage control in power grids.
Despite this, it is not always possible to control complex dynamic systems by human intervention.
The ability of AI Pontryagin to automatically learn quasi-optimal control signals for complex dynamic systems is demonstrated by researchers in their paper. The analysis presented by researchers lays much of the foundational work, but further research is required to understand the applicability of the system for specific real-world cases.
At present, protection of power grids from fluctuations and disruptions and optimize supply chains are some examples control methods are currently used for.