Throughput Region of Spatially Correlated Interference Packet Networks

Abstract

In multi-user wireless packet networks interference, typically modeled as packet collision, is the throughput bottleneck. Users become aware of the interference pattern via feedback and use this information for contention resolution and for packet retransmission. Conventional random access protocols interrupt communication to resolve contention which reduces network throughput and increases latency and power consumption. In this work we take a different approach and we develop opportunistic random access protocols rather than pursuing conventional methods. We allow wireless nodes to communicate without interruption and to observe the interference pattern. We then use this interference pattern knowledge and channel statistics to counter the negative impact of interference. We prove the optimality of our protocols using an extremal rankratio inequality. An important part of our contributions is the integration of spatial correlation in our assumptions and results. We identify spatial correlation regimes in which inherently outdated feedback becomes as good as idealized instantaneous feedback, and correlation regimes in which feedback does not provide any throughput gain. To better illustrate the results, and as an intermediate step, we characterize the capacity region of finite-field spatially correlated interference channels with delayed channel state information at the transmitters.

Capacity Results for Intermittent X-Channels with Delayed Channel State Feedback

“…Remark 1. The capacity region of the two-user erasure interference channel with delayed CSIT is known only under certain conditions [12], [20], [21], and remains unsolved for the general case. Our results for the X-Channel rely on the achievability strategy of other well-known problems, including the two-user erasure interference channel with delayed CSIT.…”

Content Delivery over Broadcast Erasure Channels with Distributed Random Cache

“…length of phase, number of bits received by each user, etc). A more precise statement would use a concentration theorem result such as the Bernstein inequality to show the omitted terms do not affect the overall result and the achievable rates as done in [29], [30].…”