A Hybrid Approach Based on Artificial Intelligence and Model Predictive Control for Enhancing Stability and Efficiency of Complex Dynamic Systems
Year: 2025
Type: article
Abstract: Stability and operational performance in uncertain and time-varying nonlinear dynamic systems, such as robotic manipulators and autonomous vehicles, is an important problem in modern nonlinear dynamical systems. This study presents a hybrid model that integrates Long Short-Term Memory (LSTM)-based artificial intelligence (AI) to identify adaptive system changes with Model Predictive Control (MPC), provided with the help of an autoencoder that leads to dimensionality reduction. The proposed AI-MPC system offers conditional stability and real-time feasibility (guaranteed by a broad Lyapunov-based theoretical framework with a clear definition of AI in the stability guarantees). Among the contributions to research are scalable and computationally efficient control architecture that adaptively models complex system dynamics, minimizes control computation time, and remains robust to disturbances and uncertainties. It is also known that the implemented framework can be useful to a wide range of simulated and experimental case studies such as robotic arms and autonomous vehicle platforms, the framework has offered improved tracking performances with a root mean square error (RMSE) by more than 58 times compared to conventional MPC and PID controllers, uses less energy by up to 50 times less and disturbances with an order of magnitude (more than 20 times) reduced. The control iterations use 30-60 milliseconds that is why it is better to use it in real time. The extension of this AI-MPC system is that the control systems can be predictive in an unpredictable and nonlinear environment that may represent the interface to resource-constrained and safety-critical systems.
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7160921570