Unlike current clustering methods, the presented vote-based clustering (VC) algorithm uses not only node location and ID information, but also battery time information. In VC, each mobile host (MH) counts Hello messages from its neighbors. At the same time it calculates its own vote that is the weighted sum of the normalized number of valid neighbors and its normalized remaining battery time. The one with higher vote than its neighbors will be selected preferentially as a cluster head (CH). When the number of dominated MHs of a CH is more than a balance threshold, neither of new coming MHs will be permitted to participate in the current cluster. Analysis and simulation results show that VC method can improve cluster structure than Lowest ID (LID) algorithm and Highest Degree (HD) algorithm.