Abstract: Opportunistic mobile sensor networks can be applied in some scenarios such as wild animal monitoring and urban sensing utilizing sensors embedded into handheld devices to collect urban information. In these applications, the gathered data usually need to be transmitted from the source node to one of multiple base stations (sink nodes). The authors propose the VSR (virtual space-based routing) scheme adapting “store-carry-forward” paradigm to transmit the data message to the base stations. In VSR, each sensor estimates the expected delivery delay to all base stations based on the meeting history between the node and the base stations. Then, each sensor is mapped into a coordinate point of a high dimensional space according to its delivery delays to all sinks. All sink nodes are corresponding to the origin of the space. The forwarding metric is defined as the Euclidean distance of the node to the origin. When two nodes encounter, the node with the higher such metric forwards the carried messages to the peer with the lower metric, until the messages are delivered to any sink nodes. VSR is robust to dynamic network because of its fine-grained forwarding decision and is appropriate for sensor node due to its low computing and storage overhead. Experiments under two random scenarios show that VSR outperforms the history-based routing proposed in ZebraNet and random forwarding scheme.