In practice, we experience low efficiency of search and rescue (SAR) frequently in disaster relief. Here, we will optimize the SAR through agent-based simulation. In the kind of cases described here, rescue teams are characterized by different capabilities, and the tasks often require different capabilities to complete. To this end, a combinatorial auction-based task allocation scheme is used to develop a cooperative rescue plan for the heterogeneous rescue teams. Then, we illustrate the proposed cooperative rescue plan in different scenarios with the case of landslide disaster relief. The simulation results indicate that the combinatorial auction-based cooperative rescue plan would increase victims’ relative survival probability by 13.8–16.3%, increase the ratio of survivors getting rescued by 10.7–12.7%, and decrease the average elapsed time for one site getting rescued by 19.0–26.6%. The proposed rescue plan outperforms the rescue plan based on the F-Max-Sum a little bit. The robustness analysis shows that the proposed rescue plan is relatively reliable on condition that both the search radius and scope of cooperation are larger than thresholds. Furthermore, we have investigated how the number of rescue teams influences the rescue efficiency.