Abstract: Based on the correlation between performance and topology in a large-scale heterogeneous ad hoc network, a cluster-based topology reconfiguration algorithm was proposed. For the active reconfiguration, the densely deployed nodes within every cluster will discover local topology by acquiring adjacent information to maintain the k-connectivity of neighboring topology so as to implement it. While the intercluster boundary nodes and relevant links will form the weighted bipartite graphs via computation, and they are matched optimally with each other to maintain the k nonadjacent links, where k is the number of paths selected and its value shall be adjusted dynamically according to the characteristics of local topology and requirements for applications. The passive reconfiguration is applied to the network failure by way of recovering the diffusive connectivity in the neighborhood around the failure where k hops are to be implemented. Then, after recovery, the failure domain is optimized topologically and locally. Simulation results showed that the active topology reconfiguration scheme in combination with the passive one can improve the invulnerability, reliability and validity of networks.