Volatility-Based Measurements Allocation for Distributed Data Storage in Mobile Crowd Sensing

For the implementation of distributed storage frameworks in the context of mobile crowd sensing (MCS), compressed sensing (CS) theory provides significant support, mainly because of the essential characteristic that CS theory will contain global information when encoding any measurements. Therefore, with limited measurement resources, the rational allocation of measurement resources becomes the most critical factor affecting recovery accuracy when using CS to recover the data. Unfortunately, the latest distributed storage frameworks do not take into account the importance of measurement resource allocation, which directly leads to a significant loss of data recovery accuracy. Therefore, to address this issue, this article proposes a volatility-based allocation strategy for the measurement resource. First, we process the target monitoring region in blocks. Next, we calculate the magnitude of fluctuations between adjacent reconstructed data by volatility, which is used to assess the importance of the different areas. Finally, a volatility-based measurement allocation scheme is proposed by fully considering the importance of different areas. It is important to note that the introduction of the concept of “volatility” in the context of MCS makes it feasible to correctly differentiate the importance of individual parts of the target monitoring region without any prior knowledge by employing extremely fuzzy recovery data. In addition, extensive experiments show that our measurement allocation scheme improves data recovery accuracy by 44% for uneven data distribution scenarios and 25% for even data distribution scenarios, compared with the random measurement allocation used in the state-of-the-art MCS distributed storage framework.