Control Synthesis Techniques

Control Design Optimization Using Evolutionary Algorithms

Evolutionary Algorithms

In the control design process, controller parameters often must be optimized to meet the system performance requirements. Evolutionary algorithms (EA’s), search and optimization procedures based on the mechanics of natural genetics, are an emerging technology for parameter optimization. EA search combines a Darwinian survival-of- the-fittest strategy to eliminate unfit characteristics and uses random information exchange, with knowledge from old solutions, to effect a fast and powerful search mechanism. Some advantages EA’s offer over conventional optimization techniques are convergence to a global optimum, faster convergence, and application over a large class of problems.

Evolutionary Algorithm Capabilities

• Convergence to Global Optima.
• Faster convergence than traditional numerical techniques.
• Applicable for a large class of problems.
• Allow non-traditional and multiple performance measures.

University of Alabama/Glenn Technology

The University of Alabama, under a grant from NASA Glenn, has developed:

• Software with graphical user interface (GUI) for applying Evolutionary Algorithms to control optimization problems, based on commercially available software.
• New techniques implemented that allow faster solution to large size problems.
• Technology demonstrated on Integrated Flight/Propulsion.
• Control Optimization and Integrator Wind-up Protection Optimization.
• Software released to university and industry researchers all over the country.

Contact: Sanjay Garg
Phone: (216) 433-2685
email: sanjay.garg@grc.nasa.gov

responsible official: sanjay garg
site maintenance: edmond wong
nasa privacy statement
last updated: 2.29.08