This paper applies Information Theoretic analysis to packet-based random
multiple access communication systems. A new channel coding approach is
proposed for coding within each data packet with built-in support for bursty
traffic properties, such as message underflow, and for random access
properties, such as packet collision detection. The coding approach does not
require joint communication rate determination either among the transmitters or
between the transmitters and the receiver. Its performance limitation is
characterized by an achievable region defined in terms of communication rates,
such that reliable packet recovery is supported for all rates inside the region
and reliable collision detection is supported for all rates outside the region.
For random access communication over a discrete-time memoryless channel, it is
shown that the achievable rate region of the introduced coding approach equals
the Shannon information rate region without a convex hull operation. Further
connections between the achievable rate region and the Shannon information rate
region are developed and explained.