Section: New Results

Interference modeling in multi-hop wireless networks

As interference plays a crucial role in ad hoc networks, these networks are often referred to as interference limited networks. Indeed, the network capacity is related to the radio channel reuse which is limited by the interference. As interference results from the summation of signals issued by concurrent transmitters, it strongly depends on the transmitters location. The point process used to model the nodes location is thus fundamental in any study of such networks. In previous works, the spatial distribution of the radio nodes has generally been modeled thanks to a Poisson point process. However, the Poisson point process is accurate for very sparse networks only and suffers from a lack of realism in dense ones. It relies on the assumption that transmitters are independently distributed whereas in practice, the Medium Access Control (MAC) protocol defines medium access rules to avoid collisions and by this way introduces a spatial correlation between the transmitters. Therefore, the spatial independence does not hold. Recent works already tackle this problem by modifying the initial Poisson process according to the mechanisms of a CSMA/CA protocol and use the well known Matérn point process. However, this approach still suffers from two strong limitations. First, it tends to underestimate the number of concurrent transmitters and as a consequence, the interference level is strongly under-estimated in dense networks. Second, it assumes that a transmitter defers its transmission if it detects a busy medium due to the nearest previously selected transmitter. Albeit, medium access policies may rely either on carrier detection or on energy level. In the case of carrier detection, the inhibition model can be effectively related to the strongest signal, according to the Matérn selection process. However, in the case of energy detection, the candidate transmitter does not defer its transmission according to the strongest interference but rather to the interference level induced by all effective transmitters. In this case, the inhibition process should be related to all concurrent transmitters and inhibition balls are not adequate. In [12] , we investigate original processes that comply with the CSMA/CA policies. We propose the use of the Simple Sequential Inhibition point process to model CSMA/CA networks where carrier detection depends on the strongest emitter only. This point process is then extended to a family of new point processes modeling effective transmitters in a busy medium detection mode based on the strength of all concurrent signals. The impact of these different models on the interference distribution is evaluated and we highlight large variations with respect to the different point processes. In particular, we show that the use of a Poisson process is generally inaccurate to model CSMA/CA networks. We finally study the theoretical Bit Error Rate resulting from the different point processes.