Hybrid estimation of complex systems

Area of Science:

• Control Systems Engineering
• Stochastic Systems Theory
• Artificial Intelligence in Engineering

Background:

• Modern automated systems exhibit continuous and discrete dynamics, exceeding standard Kalman Filter capabilities.
• Multiple Model (MM) estimation, while effective for discrete modes, becomes infeasible for complex systems with numerous interacting components and modes.
• Complex stochastic hybrid models are necessary for systems with many operational and failure modes, posing challenges for real-time estimation.

Purpose of the Study:

• To analyze the limitations of current Multiple Model (MM) estimation techniques for complex automated systems.
• To introduce an efficient alternative hybrid estimation scheme for systems with a large number of modes.
• To develop a robust estimation approach for unforeseen behavioral modes and degraded system conditions.

Main Methods:

• Analysis of shortcomings in existing Multiple Model (MM) estimation schemes.
• Development of a novel hybrid estimation scheme leveraging model-based reasoning search techniques.
• Integration of a new approach for handling unknown behavioral modes within the hybrid estimation framework.

Main Results:

• The proposed hybrid estimation scheme efficiently estimates complex systems with a large number of modes.
• Search techniques focus estimation on the most probable modes, effectively handling system noise.
• The scheme robustly manages unknown behavioral modes, ensuring fail-safe operation in degraded conditions.

Conclusions:

• A new hybrid estimation scheme offers an efficient and robust solution for complex automated systems with numerous modes.
• This approach overcomes the infeasibility of traditional MM estimation for large-scale, dynamic systems.
• The developed method enhances system reliability and safety by adapting to unforeseen situations.