TPC_IPS

Table of contents

Indoor Positioning System (IPS) is a computer network system used to locate people and things where GPS and other satellite technologies cannot provide accurate positioning, such as factories, warehouses, hospitals, care homes, buildings, airports, and underground.

Research and market forecasts that the global IPS market will be worth $6.92 billion in 2020 and $23.6 billion by 2025, growing at a CAGR of 27.9%.

We have been developing our own system called TPC_IPS since 2017, and plan to release it in the first half of 2021.

The figure below shows a conceptual diagram of TPC_IPS in a warehouse. Beacons are attached to things and people, and the electrical signals emitted by the beacons are read by a terminal, which analyzes the signal information and positions each beacon using a computer.

TPC_IPS is our original indoor positioning system that uses beacons as the transmitter and Raspberry Pi as the receiver (sensor).

When a beacon (Apple's iBeacon standard) is activated, it sends out a signal with a certain structure. TPC_IPS sends a beacon signal scanning command to all Raspberry Pi terminals (RP terminals) with a single click from the GUI of the PC terminal. Upon receiving this command, each RP terminal starts scanning for beacon signals and acquires the information necessary for positioning, such as UUID and RSSI, from the received signals and sends it to the server. The server analyzes the information, estimates the position coordinates of the beacon, and writes the coordinates to the database.

The figure below shows each device, user operation, and data processing flow in more detail.

Demo video for QuickIPS overview

This 11-minute video provides an overview of QuickIPS. Turn on the English subtitle when viewing it.

TPC_IPS has the following features:

1. Indoor positioning system using BLE beacons as transmitters and Raspberry Pi's as locators

2. Depending on the usage conditions, the position estimation algorithm can be selected from TCOT (two-circle-oriented trilateration, no prior learning required) or machine learning (kNN, prior learning required).

3. Advantages of using general-purpose equipment

• Avoids vendor lock-in by using general-purpose products such as BLE beacons and Raspberry Pi's (RP's).

• Reduces costs for expansions and extensions

4. UX-conscious user interface

• Ordinary operations such as beacon scanning, position information acquisition, and database information updating can be performed from the GUI on a single PC terminal.

5. Centralized management of a large number of RP's receiving terminals from a single PC

• Even when a large number of RP terminals (Linux OS) are installed, commands can be sent to all terminals with a single click from the GUI of the PC terminal.

6. Customizable license structure.

• Basic Feature License --- When the existing features are enough to satisfy your requirements

• Development license --- Contact us

7. Application Template --- QuickIPS bundled

• 2 types of floormaps, heatmap, and error graph

• Fast development for iPad/iPhone applications

• Customizable

Note:

• FileMaker Pro 19 or later is required for using GUI and customizing QuickIPS.

• Configuration and management of beacons themselves should be done by the dedicated service or software provided by the vendor.

Software

Hardware

Scanning Beacons

Plotting the Positions

Plotting function calculates the positional coordinates of each beacon using the data stored in the database from the scanning described above, using either our proprietary two-circle-oriented trilateration (TCOT) or machine learning (knn), and displays the positional coordinates in a scatter plot.

The plotting function is used through the process of deployment or development in order to check the degree of difference between the actual coordinates of the beacons and the calculated coordinates. This feature is available in the TPC_IPS application or in a web browser.

TCOT Algorithm

knn Algorithm

On the scatter plot, the fixed beacons are indicated by red circles (●) and the movable beacons are indicated by blue circles (●). If arrows appear in the plot, the starting point of the arrow indicates the original beacon position and the end of the arrow indicates the estimated beacon position.

QuickIPS ― IPS/RTLS application templates for FileMaker ―

TPC_IPS comes with IPS/RTLS application templates for FileMaker, allowing any FileMaker developer to easily create an IPS application.

Note:
The sample application has 81 beacon display objects. If you wish to display more than 81 beacon positions, you need to add more beacon objects by yourself.

In this section, we will introduce two types of positioning estimation algorithms used in TPC_IPS - TCOT and machine learning (kNN), as well as the new technology AoA that is currently attracting attention.

Two-Circle-Oriented Trilateration — How TCOT Works

TPC_IPS calculates the coordinates of the the iBeacons attached to people or to things, by receiving their signals using multiple locators (Raspberry Pi's: RP's) which are placed in advance.

TCOT (Two-Circle-Oriented Trilateration), one of the position estimation algorithms of TPC_IPS, calculates the distance between the terminal and each iBeacon based on the strength of the iBeacon signasl (RSSI) received by the three nearest locators, and then calculates the position of the iBeacon using trilateration.

TPC_IPS's "Plotting All the Beacon Positions" function simultaneously displays the actual coordinates of all beacons scanned and the estimated coordinates calculated by TCOT (or knn as described below).

This allows the user to understand at a glance the degree of positioning error during development and deployment.

For more details on TCOT, please refer to the following blog post:

Indoor Mobile Position Monitoring Model with iBeacon/Raspberry Pi 3 ― Enhancement of Two-Circles-Oriented Trilateration (TCOT) and Measurement Results

Positioning accuracy is an important factor when evaluating an IPS.

If positioning accuracy is defined as "the difference between the actual and measured beacon position coordinates", then the positioning accuracy of TPC_IPS is as follows:

When locators are placed at 4m x 4m intervals (4mGird): Approx. 2.6m

When locators are spaced 8m x 8m (8mGird): Approx. 3.8m

When locators are placed at 16m x 16m intervals (16mGird): Approx. 7.0m

Note:
When the locators (Raspberry Pi locators) are placed to form a square and another locator is placed in the center of the square.

• Sample results of positioning accuracy are explained in the following blog post:
  Indoor Mobile Position Monitoring Model with iBeacon/Raspberry Pi 3 ― Enhancement of Two-Circles-Oriented Trilateration (TCOT) and Measurement Results

Improving Estimated Position Accuracy with Machine Learning

Another algorithm for positioning in TPC_IPS is fingerprinting / machine learning (knn). In this algorithm, a receiving terminal stores in advance the RSSI it will receive from a beacon at a known position. For example, when terminal X receives signals from a beacon located at coordinates (0, 4), the RSSI is -77, which is stored in the database. Similarly, place beacons in all monitoring areas and register the actual position coordinates and RSSI. This is called fingerprinting.

For this fingerprint data, we apply kNN (k nearest neighbor method), one of the machine learning algorithms, to calculate the position coordinates. The figure below shows the result of plotting the estimated position coordinates of TCOT and knn, respectively.

As shown in the figure, using knn may significantly improve the positioning accuracy over TCOT shown on the left. However, fingerprinting/machine learning requires a lot of time and effort for preparation, because the computer needs to be trained (supervised learning) by repeating the process of placing and measuring beacons in advance, and then placing and measuring them again.

Please refer to the following blog post for the details:
Improving IPS Positioning Accuracy with Machine Learning

There are many different algorithms in machine learning, not only kNN.　We plan to verify and implement other algorithms, and post the results on this website.

Improving Positioning Accuracy with AoA/AoD

The figure above shows the direction detection method of AoA.　Two or more antennas installed/connected to the locator receive beacon signals, and the direction (angle) of the beacon as seen from the locator is calculated based on the distance between the antennas, phase difference, and wavelength. If the exact angle is known, the position coordinates of the beacon can be calculated with higher accuracy than simple trilateration that relies solely on distance.

As of May 2020, no "major" vendors have released locators or transmitters (beacons) equipped with Bluetooth 5.1 SoC's. However, we are conducting research and development to improve positioning accuracy by integrating TCOT with direction finding using AoA/AoD.

Note:
Quuppa, a Finnish company, has been commercializing AoA-based IPS (RTLS) with its own technology for many years before the Bluetooth 5.1 specification was formulated (early 2019), and has an established reputation for high-precision positioning. A video showing Quuppa's RTLS tracking the movements of a fast-moving ice hockey player is also available.

Here are some possible use cases for our TPC_IPS solution.

*TPC_IPS provides the position coordinates of each beacon via API or database. Applications such as map display and inventory display need to be developed separately.

Warehouses

You can keep track of products, workers, forklifts, and other transportation equipment at a glance in a large warehouse.

Care homes

Care home for the elderly.

Real-time display of residents, wheelchairs, staff, and the occupancy of common areas such as lounges and dining rooms.

Hospitals

To manage the location of doctors, patients, medical staff, hospital beds, and medical equipment in hospitals

Lost and Found at Amusement Parks

Security

For office security at night.

Real-time display of security staff deployment and patrol positions.