Two Relative-Residual-based Dynamic Schedules (RRDS) for Belief Propagation (BP) decoding of Low-Density Parity-Check (LDPC) codes are proposed, in which the Variable code-RRDS (VN-RRDS) is a greediness-reduced version of the Check code-RRDS (CN-RRDS). The RRDS only processes the variable (or check) node, which has the maximum relative residual among all the variable (or check) nodes in each decoding iteration, thus keeping less greediness and decreased complexity in comparison with the edge-based Variable-to-Check Residual Belief Propagation (VC-RBP) algorithm. Moreover, VN-RRDS propagates first the message which has the largest residual based on all check e-quations. For different types of LDPC codes, simulation results show that the convergence rate of RRDS is higher than that of VC-RBP while keeping very low computational complexity. Furthermore, VN-RRDS achieves faster convergence as well as better performance than CN-RRDS.