I want to know the main differences between optimization and reoptimization concept.
Then can we say model predictive control (MPC) as a rolling horizon control (RHC) method is a reoptimization strategy because it measures outputs and runs online optimization over the next prediction horizon for some instances (after control horizon)?
Not sure if I have understood the question. I assume you mean is MPC keeps re-optimizing?
Yes, MPC re-optimizes the optimal input vector at every sample time. The reason is that there are noises and disturbances and the previous optimized values are not necessarily good for the current moment. In addition, the horizon is moving and new optimization is necessary. In practice, The system can drift from theory significantly and avoid re-optimization can cause instability.