In this paper, we consider the transport capacity of ad hoc networkswith a random flat topology under the present support of an infinitecapacity infrastructure network. Such a network architecture allowsad hoc nodes to reach each other by purely using ad hoc nodes asrelays. In addition, ad hoc nodes can also utilize the existinginfrastructure fully or partially by reaching any access point (orgateway) of the infrastructure network in a single or multi-hopfashion.

Using the same tools as in cite{gupta-capacity2000}, weshowed that the per source node capacity of $Theta(W/log(N))$ can beachieved in a random network scenario with the assumptions that thenumber of ad hoc nodes per access points is bounded above and that $N$ad hoc nodes excluding the access points, each capable of transmittingat $W$ bits/sec using a fixed transmission range, constitute aconnected graph.

This is a significant improvement over the capacityof random ad hoc networks with no infrastructure support which isfound as $Theta(W/sqrt{N log(N)})$ in cite{gupta-capacity2000}.Although better capacity figures are obtained by complex networkcoding or exploiting mobility in the network, infrastructure approachprovides a simpler mechanism that has more practical aspects.

We alsoshow that even when less stringent requirements are imposed ontopology connectivity, a per source node capacity figure that isarbitrarily close to $Theta(1)$ can not be obtained. Neverthelessunder these weak conditions, we can further improve per nodethroughput significantly.