Gradient-free parameter identification in repetitive processes

Particle swarm based parameter identification (tracking) in non-stationary repetitive process.

A DOP-capable PSO has been employed to track parameters of the non-stationarry repetitive process. Such an algorithm can be used to retune on the fly controller gains if they are defined as known functions of the parameter(s) being tracked. Please see inside the m-file for more information. This submission enables you to play with different system configurations, e.g. some flags are provided for fast reconfiguration of the swarm update rule. The solution was inspired by the concept of plug-in direct particle swarm repetitive controller (see PDPSRC or PDMSRC on Matlab Central). The similarity between controller and observer design present in selected "classic" control schemes obviously provoked such a solution. This submission might be of your interest if you deal with: parameter identification in repetitive processes, iterative learning estimation, dynamic optimization problems and population-based evolutionary optimization. Such estimation tasks are encountered, e.g., in robotics and power electronics and it is not rare that these systems are subject to parameter variations.