Abstract
In current reconfigurable intelligent surface (RIS)-aided systems, adding a new pair of transceivers often requires the RIS phases to be entirely re-calculated for all elements, inducing great overhead, especially for large-scale RISs. In this letter, we propose to only adjust part of RIS elements to serve the newly added pair, while the other elements stay unchanged, providing more flexibility to balance performance and overhead. Simulation results show that the method, with modifying only a small portion of elements, can achieve most of the RIS gain that would be achieved if optimizing phases of all the elements.