The indicators and constraints of distribution system restoration (DSR) models were sorted out, and restoration strategies based on representative flexible resources (FRs) were reviewed. The methods for addressing uncertainties in renewable energy output and load fluctuations, such as stochastic programming, robust optimization, and chance-constrained programming, were systematically summarized, together with their applicable scenarios. Existing algorithms such as search algorithms, mathematical programming methods, and artificial intelligence algorithms that utilize multiple FRs to solve DSR problems, as well as their respective strengths and drawbacks were elaborated. Research findings on DSR involving the coordination of multiple FRs in coupled network configurations were elaborated, and distributed control strategies for service restoration in coupled networks were further introduced. Potential future research directions in the field of DSR were discussed, such as research on hybrid DSR algorithms and DSR for coupled network configurations based on the integration of deep reinforcement learning algorithm and multi-agent systems.