The Impact of Imperfect Spectrum Sensing on the Performance of LTE Licensed Assisted Access Scheme

The energy detection technology is adopted as the detection method of the unlicensed channel in LTE release 13. However, the detection may be imperfect in actual scenario due to the simplicity of the detection method. In order to eliminate the controversy over LTE licensed assisted access (LAA) scheme, the paper deeply investigates the performance of the scheme under imperfect spectrum sensing (ISS). Considering possible ISS during the enhanced clear channel assessment (eCCA), the LAA scheme is modeled as a new two dimensional discrete time markov chain. In order to capture ISS, different from the definition in Bianchi's model, the discrete time scale is defined as the end of backoff slot time in the proposed model. The definition enables LAA small base stations to enter the defer state from backoff state when the channel is detected to be busy, and further possible ISS in the defer period and backoff slot time (BST) of eCCA can be considered. Based on the proposed model, the expressions of collision probabilities and throughput under three backoff mechanisms based LAA schemes are derived to deeply analyze their performance by comparisons. A large number of experimental and analytical results prove the validity of the proposed model. The analytical results also show that not only the backoff mechanisms but also ISS can greatly affect the network performance. In the respect of fairness (i.e. the impact on WiFi users), the WiFi collision probability always increases with the decrease of the false alarm probability (FAP). This means the WiFi collision probabilities under three LAA schemes may be higher than the baseline level for the graceful coexistence when FAP is very little. In this sense, the LAA schemes cannot be referred to as being graceful though they have been proven to be deterministic graceful coexistence schemes under perfect spectrum sensing (PSS). In the respect of throughput, LAA throughput always decreases and WiFi throughput increases with the growth of FAP, and total throughput always decreases. Therefore, based on the fixed SNR and sampling rate in actual scenario, the network performance can be tuned by energy threshold, as is consistent with the viewpoint in other literatures.