Table of contents
overview
The real-time positioning of mobile devices (users) is the technical foundation of many current location-based services (Location Based Service, LBS), and is also the basis of navigation and other functions. Therefore, to achieve as accurate and low-cost positioning as possible is the current research goal of many scholars and companies. Outdoor positioning using GPS has been able to achieve high positioning accuracy, and the mature usage of GPS has been integrated in many platforms (for example, there are library functions dedicated to outdoor positioning in the Android platform). Indoor positioning cannot use GPS positioning due to weak indoor GPS signal and complex indoor space structure, but there are many other indoor positioning methods. This paper summarizes various common indoor positioning methods.
Common Indoor Positioning Methods
Triangulation (distance measurement)
Specifically, it includes based on signal time of arrival (Time of Arrival, TOA), based on time difference of arrival (Time Difference of Arrival, TDOA), based on angle of arrival (Angle of Arrival, AOA), etc. The basic principle is to first measure the distance between multiple devices , and then use the Euclidean formula to measure the user's location (such as by constructing a triangle, taking the centroid of the triangle as the device's location, etc.).
There are many ways to measure distance between devices, such as the method based on WiFi or BLE signal strength attenuation, because the attenuation of signal strength is positively correlated with distance; but the current mapping relationship between signal strength and distance is still based on the huge amount Data collection, which consumes a lot of deployment costs.
Since the triangulation method can often obtain multiple sets of data, many positioning methods also combine greedy algorithms (finding approximate optimal solutions), KNN algorithms (nearest neighbor algorithms), etc., to assign different weights to multiple sets of data, and to The resulting data are final measurements.
proximity method
The basic principle is to pre-install many beacons with specific built-in information (fingerprint information), such as iBeacon, in the positioning scene; when the user enters the signal range of a certain beacon, it is deemed that the user is located at the location of the beacon. The disadvantage is that the error is large, and the improved algorithm has triangulation based on beacons, which can improve the positioning accuracy.
It is worth mentioning that this type of positioning method can add some services at the user content level, such as the push of store information in the mall: when the user enters the mall, the beacon in the mall detects the user's existence and actively Push the relevant information of the mall.
scene analysis
WiFi fingerprint positioning method is a typical example. The basic principles of this type of positioning method are as follows:
1. Some specific information (such as WiFi Fingerprinter) of the positioning scene is collected in advance and stored in the cloud database, which is generally called the offline stage;
2. When the user sends a positioning request, the mobile device first Collect the WiFi fingerprint data of the current scene and transmit it to the server. The server searches and matches according to the database. After finding the location with the highest WiFi fingerprint similarity, it returns the location information. This is generally called the online stage.
The disadvantage of this positioning method is that a huge amount of information needs to be collected in the offline phase. Using MCS (Mobile Crowdsensing, Mobile Crowd Sensing) can alleviate this problem, but it still consumes a lot of deployment costs.
Pedestrian Dead Reckoning (PDR)
The basic principle is to use the rich built-in sensors of the mobile device, based on the accurate positioning obtained by the user last (such as the position obtained when using GPS to locate outdoors for the last time before entering the room), and according to the mathematical formula, continuously calculate the user's location. The displacement of the user, and then deduce the real-time position of the user.
When positioning, it mainly measures the displacement of the user and the direction angle (heading) between the user's direction and the specified coordinates. The disadvantage of PDR is that there is a cumulative error and it is more obvious.
In terms of displacement, the error mainly comes from the judgment of the user's action, that is, how to judge the user's "one step forward", and the prediction of the user's step length, that is, how long the user "walks in one step".
In terms of direction angle (heading), the main error is that the built-in sensors of different mobile devices have different precision, which is easy to cause errors. On the other hand, different ways and angles of users holding the device will also lead to measurement deviations.
Now the research on PDR mainly focuses on the elimination of the above two types of errors.
Further development of indoor positioning
Indoor positioning initially collects some initial data, such as latitude and longitude, altitude, etc. The further use of these data is also one of the current research hotspots, and indoor visual maps are a typical direction.
indoor visual map
The basic principle is to approximate the indoor structure through the activity information uploaded by a large number of users, such as judging where is the wall, where is the elevator, where is the corridor, etc.
Common methods are based on the user's activity trajectory. For example, the user cannot pass through the wall, so it can be inferred that there is a wall in the place the user has not reached. There are gaps in activity, so you can tell if a place is an "elevator" by measuring acceleration.
Another type of method is to use the camera of the user equipment to continuously take pictures, upload the pictures to the cloud server, and reconstruct (simulate) the indoor environment where the user lives through image processing.
Personally, I think the difficulty of indoor visual maps is the need to collect a large amount of basic data.
directions for in-depth research
- Improvement of positioning accuracy (the most intuitive);
- Reduction of positioning cost (same effect but less cost consumption and high practicability);
- Utilization of existing positioning data (LBS, location-based services, eg generation of visual maps).
——————————————————————————————
Finally, post my personal public account: WeChat search "Chaqian" or scan the picture below. Usually, some programming-related articles will be updated, and everyone is welcome to pay attention~
