Passive Internet of Things (IoT) builds on backscatter mechanism and exchanges data between passive tags, which is regarded as the key to realize the beautiful vision of Internet of Things due to the characteristics of low cost, low energy consumption, long lifetime, and low maintenance. It has extremely high theoretical research significance, application value, and wide development prospects in a broad spectrum of applications, which span from augmented reality, automatic book identification, warehousing management to industrial IoT. Although passive backscatter networks-based sensing has drawn extensive attention from both academy and industry, there are still many challenges. First, in general, different links possess different statistical characteristics. It is necessary to couple the link heterogeneity into location estimators to decrease the sensitivity to noise. Second, Cramér-Rao lower bound provides a metric of localization performance limit. It is necessary to determine the distributions under which a location estimator achieves the bound. Third, there are many pairs of tags incapable of communicating with each other because of limited communication range, resulting in a broken pairwise distance matrix. It is necessary to perform cooperative localization with such a matrix of high-frequency loss of range measurements. Fourth, under semi-distributed localization schemes, after breaking down the full network into a series of fragments, it is necessary to determine the (sub)optimal assembling order from O(N!) choices. This chapter answers above questions in the multidimensional scaling framework.