Title: Throughput Scaling Laws in Random Wireless Networks
Speaker: Bert Hochwald (joint work with Radhika Gowaikar, Babak Hassibi) 4:20-4:40

ABSTRACT

We depart from the purely geometric models of Kumar and Gupta and consider a wireless network where the fading connections between any two nodes are drawn independently from a given distribution f(.). Such models may be better suited for networks of small physical size, and with many scatterers, where the strength of a connection depends not so much on the distance between the transmitter and receiver, but rather on a random event, such as the existence of an obstacle. For such networks, we show that the total throughput achieved by multi-hop routing is a strong function of the distribution f(.) from which the connections are drawn. In particular, we show that for a wide class of distributions the throughput scales as O(n/log^c n), where n is the number of nodes and c is a constant independent of n (typically between 1 and 4), which is a significant improvement over the O(\sqrt{n}) results obtained from geometric models. Our results therefore shed light on what kind of connectivity is needed in wireless networks to obtain high throughput. We will also discuss various other implications of the result.