Topology Sensing of Non-Collaborative Wireless Networks With Conditional Granger Causality

Topology sensing of non-collaborative wireless networks is a challengingtask due to the limited available information resulting from the inherent non-collaborative characteristics. To address this issue, this paper investigates the topology sensing by effectively exploiting a time series of packet arrival times, but without package decoding. Firstly, we develop a topology sensing framework for a non-collaborative wireless network from a statistical perspective, in which the signal arrival time is converted into discrete time series as the original data of topology sensing, and the connectivity problem is transformed into the correlation problem. Following that, we propose a conditional Granger causality (CGC) based topology sensing scheme using the causal inference of binary time series without packets decoding. The main idea of the algorithm is to find the potential neighbor set by Granger causality (GC) first and then refine the ultimate neighbor set by CGC. Then, to deal with the problem of small samples in real case, we adopt a high-order Markov chain to model the communication behavior between nodes. Finally, extensive simulations under various parameter configurations are presented to validate the effectiveness of the proposed scheme. It shows that the proposed scheme outperforms the GC-based approach and using Markov chain to enhance data can increase the accuracy of topology sensing with small samples.

Automated summary

This paper investigates a topology sensing scheme for non-collaborative wireless networks using a time series of packet arrival times and causal inference. It proposes a framework for topology sensing from a statistical perspective, transforms the connectivity problem into a correlation problem, and utilizes a high-order Markov chain to model communication behavior. Extensive simulations are conducted to validate the effectiveness of the proposed scheme.