Xiao Li,Huan Li,Harry Kai-Ho Chan,Hua Lu,Christian S. Jensen
Abstract:Indoor location-based services rely on the availability of sufficiently accurate positioning in indoor spaces. A popular approach to positioning relies on so-called radio maps that contain pairs of a vector of Wi-Fi signal strength indicator values (RSSIs), called a fingerprint, and a location label, called a reference point (RP), in which the fingerprint was observed. The positioning accuracy depends on the quality of the radio maps and their fingerprints. Radio maps are often sparse, with many pairs containing vectors missing many RSSIs as well as RPs. Aiming to improve positioning accuracy, we present a complete set of techniques to impute such missing values in radio maps. We differentiate two types of missing RSSIs: missing not at random (MNAR) and missing at random (MAR). Specifically, we design a framework encompassing a missing RSSI differentiator followed by a data imputer for missing values. The differentiator identifies MARs and MNARs via clustering-based fingerprint analysis. Missing RSSIs and RPs are then imputed jointly by means of a novel encoder-decoder architecture that leverages temporal dependencies in data collection as well as correlations among fingerprints and RPs. A time-lag mechanism is used to consider the aging of data, and a sparsity-friendly attention mechanism is used to focus attention score calculation on observed data. Extensive experiments with real data from two buildings show that our proposal outperforms the alternatives with significant advantages in terms of imputation accuracy and indoor positioning accuracy.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is that in indoor positioning, radio maps have a large number of missing Received Signal Strength Indicator (RSSI) values and Reference Points (RPs) due to radio environment fluctuations during the data collection process and the asynchrony between RPs and data collection. These missing values make the radio maps sparse, thus affecting the accuracy of radio - map - based indoor positioning. Therefore, the paper proposes a complete set of techniques to fill these missing values in order to improve the accuracy of indoor positioning.
Specifically, the paper distinguishes between two types of missing RSSIs: Missing Not at Random (MNAR) and Missing at Random (MAR). To deal with these problems, the author designs a framework that includes a module for distinguishing the types of missing RSSIs and a data imputation module for filling in the missing values. Through this method, the paper aims to improve the quality of radio maps and thereby improve the accuracy of indoor positioning.
### Main contributions:
1. **Distinguishing missing RSSIs**: For the first time, the paper distinguishes between MNARs and MARs and provides a clustering - based method to identify these missing types.
2. **New encoder - decoder model**: A new encoder - decoder model (BiSIM) is designed, which can simultaneously fill in missing RSSIs and RPs, making use of the time - dependence in time series and the correlation between fingerprints and RPs.
3. **Experimental verification**: Through a large number of experiments, it is proved that the proposed framework is significantly superior to existing methods in terms of data imputation accuracy and indoor positioning accuracy.
### Technical details:
- **Missing RSSI discriminator module**:
- **Clustering algorithms**: K - means and Topology - Aware Agglomerative Clustering (TopoAC) are used to cluster Access Point (AP) profiles.
- **Distinguishing process**: Through the clustering results, MARs and MNARs are identified according to the locality of AP profiles, and a mask matrix is generated.
- **Data imputation module**:
- **Bidirectional encoder - decoder model**: The BiSIM model takes into account the timeliness of records and deals with irregular intervals in time series through a time - lag mechanism.
- **Sparsity - friendly attention mechanism**: A sparsity - friendly attention mechanism is designed to deal with the high sparsity of input features.
### Experimental results:
The paper has carried out extensive experiments on datasets of two real - world buildings, and the results show that the proposed method is significantly superior to existing methods in terms of both data imputation accuracy and indoor positioning accuracy.
In conclusion, this paper solves the problem of radio - map sparsity in indoor positioning through innovative technical means, improving the precision and reliability of indoor positioning.