# Research Proposals

## Fingerprint-Based Indoor Positioning without Predefined Reference Coordinate

### Background

In modern society, having a dependable positioning system is crucial for various instances, and GPS stands out as the most commonly used system. However, challenges arise in environments with obstacles like buildings or tree canopies, making GPS less reliable. To overcome these challenges, indoor positioning systems have emerged, utilizing more accurate methods customized to specific local environments.
There are various technologies used for indoor positioning systems, currently the ones that widely researched are Wifi and Bluetooth. The technologies are using Received Signal Strength Indication (RSSI) measurement of wireless signals from stationary emitters to pinpoint the positions of specific objects, which we currently refer to as receivers. The main challenge with RSSI is the signal disruption caused by environmental objects, which impacts the accuracy of distance measurements. Some of the strategy to overcome the challenge is to increase the number of emitters [Mendoza-Silva, G.M.; Matey-Sanz, M.; Torres-Sospedra, J.; Huerta, J. BLE RSS Measurements Dataset for Research on Accurate Indoor Positioning. Data 2019, 4, 12] and introducing redundant emitter in every reference point to obtain more accuracy [Torres-Sospedra, J.; Montoliu, R.; Martínez-Usó, A.; Avariento, J.P.; Arnau, T.J.; Benedito-Bordonau, M.; Huerta, J. UJIIndoorLoc: A new multi-building and multi-floor database for WLAN fingerprint-based indoor localization problems. In Proceedings of the 2014 International Conference on Indoor Positioning and Indoor Navigation (IPIN), Busan, Korea, 27–30 October 2014; pp. 261–270].
In terms of methods, the common used one are lateration, angulation and fingerprinting. Lateration is a method to find the best fit of object based on distance of static reference points. Similarly angulation is using angles instead of distance in lateration. The fingerprinting encompasses two stages. In the first stage, also known as offline stage, the signal quantity of each detected emitter at a given time and position (a fingerprint) is measured at several places the target scenario and stored to create a characterization of the signals in that scenario as comprehensive as possible. The collected database is called the training database. If the measured signals are radio frequencies (RF), the database is also called radio map. The collection process of the database is called site survey, war-driving, radio map creation or training fingerprints collection. In the second stage, also known as online stage, the position corresponding to new measured signal quantities is estimated using the positions associated with the stored fingerprints that are the most similar when compared to the new measurements.
During the offline phase, site surveying, in which the fingerprints of the area of interest are sampled at predefined reference points (RPs), is performed [Deep Learning Methods for Fingerprint-Based Indoor Positioning: A Review].

### Scope of the research

- RSSI (either Wifi or BLE)
- Using lateration to do fingerprint
- Assumed no altitude variation during offline stage

### Limitation of existing approaches

### Questions to Solve

The hypothesis suggests that randomly placing emitters without prior knowledge of it's coordinate will provide the relative position of a receiver among the transmitter using literation method. At the same time, multiple position measurement can be used to generate radio map of signal attenuation in the space. A radio map here encapsulates the information about how much signal disruption occurs at all points in the space. The radio map can be analogous to an imaginary cloud in empty space. Whenever there are objects that disrupt the signal propagation from the emitter to the receiver, it can be imagined as if there is a thickening cloud at the position of the disruptive object.

Practically, collecting signal data at all possible points is difficult to achieve. A reasonable approach is to collect data through a receiver that randomly moves in space, similar to a training step in machine learning. The data is processed in such a way that the combination of several emitters at the same time generates position information about the object. The inconsistency in predicting adjacent points generates the radio map. As more data is received from the receiver's position, the shape of the radio map transforms. Eventually, the radio map reaches a stable shape, capturing all information about the environment's signal attenuation and enabling adjustment of position information to achieve maximum positioning accuracy. This radio map allows us to determine precise object positions relative to the emitters, which do not provide coordinate information initially. Integration with a floor plan becomes a straightforward task when needed.

Key questions for this research include:

1. How can we build mathematical models to capture all relevant information related to the radio map of signal attenuation?
2. What's the most efficient algorithm for collecting and processing data into a radio map of signal attenuation given a set of data. Moreover can the algorithm distinguish the map of signal attenuation from internal variations within signal emitters?
3. How can we incorporate altitude information into the model, extending the radio map of signal attenuation into three-dimensional space?

### Methodology

1. Explore machine learning techniques for spatial data training.
2. Investigate existing positioning algorithms and error control measures.
3. Carry out experiments for data collection and processing.
4. Build the mathematical model and the algorithm to capture and process the data.
5. Analyze the data to validate the effectiveness of the proposed model and algorithms.

### Related background

This research is an extension of my prior work during my bachelor's degree in Information System and Technology at Institut Teknologi Bandung, Indonesia. The earlier research, titled "Indoor Positioning System Using iBeacon," primarily focused on implementation and feasibility analysis, naturally leading to the exploration of the current proposal topic.

In addition to my academic background, I have significant practical experience in building GPS tracking systems. In 2016, I contributed to Vessel Tracking, and in 2018, I participated in similar projects involving Trucks and Cars.

From then until now, I have predominantly worked in IoT data processing, refining my skills in efficiently programming high frequency data processing.

Over the past two years, I have pursued undergraduate Mathematics in part-time, studying Algebra, Analysis, Geometry, and Differential Equations. This coursework has boosted my skills in working with mathematical concepts and methodologies.