In wireless networks, bandwidth is likely to remain a scarce resource. We foresee scenarios wherein mobile hosts will require simultaneous data transfer across multiple links to obtain higher overall bandwidth. A call request of a connection in a wireless network is blocked if there exits no bandwidth route. This blocking does not mean that the total system bandwidth capacity is less than that requested, but that there is no path in which each link has enough residual unused bandwidth to satisfy the requirement. Like routing in a datagram network, if packets of a virtual circuit can stream across multiple paths, we can select multiple bandwidth routes such that the total bandwidth can meet the requirement of a source-destination pair. Therefore, even though there is no feasible single path for a bandwidth-constrained connection, we may still have a chance of success this one if we can find multiple bandwidth routes to meet the bandwidth constraint. In this paper, we propose a bandwidth-constrained routing algorithm to aggregate the bandwidth of multiple wireless links by splitting a data flow across multiple paths at the network layer. That is, the algorithm allows the packet flow of a source-destination pair to be delivered over multiple bandwidth routes with enough overall resources to satisfy a certain bandwidth requirement. Our algorithm considers not only the QoS requirement, but also the cost optimality of the routing paths to improve the overall network performance. Extensive simulations show that high call-admission ratios and resource utilization are achieved with modest routing overheads. This algorithm can also tolerate node moving, joining, and leaving.