Fundamental Constraints on Multicast Capacity Regions

TL;DR

This paper explores the fundamental geometric constraints of the capacity region in multicast broadcast channels with multiple receivers, revealing new bounds and the necessity of network coding for certain configurations.

Contribution

It characterizes the achievable rate region for multicast broadcast channels with multiple receivers, highlighting the role of message type swapping and network coding in the region's structure.

Findings

For L=2, the achievable region is characterized by message swapping operations.

For L=3, network coding becomes necessary to describe the capacity region.

Fundamental geometric constraints limit the achievable rate regions in multicast channels.

Abstract

Much of the existing work on the broadcast channel focuses only on the sending of private messages. In this work we examine the scenario where the sender also wishes to transmit common messages to subsets of receivers. For an L user broadcast channel there are 2L - 1 subsets of receivers and correspondingly 2L - 1 independent messages. The set of achievable rates for this channel is a 2L - 1 dimensional region. There are fundamental constraints on the geometry of this region. For example, observe that if the transmitter is able to simultaneously send L rate-one private messages, error-free to all receivers, then by sending the same information in each message, it must be able to send a single rate-one common message, error-free to all receivers. This swapping of private and common messages illustrates that for any broadcast channel, the inclusion of a point R* in the achievable rate region implies the achievability of a set of other points that are not merely component-wise less than R*. We formerly define this set and characterize it for L = 2 and L = 3. Whereas for L = 2 all the points in the set arise only from operations relating to swapping private and common messages, for L = 3 a form of network coding is required.