A method for indoor passive positioning using LoRa signal fingerprint

Abstract

A method for indoor passive positioning using LoRa signal fingerprinting is proposed. First, in the offline phase: an indoor positioning environment is established using LoRa node devices and a LoRa gateway device, enabling communication between the LoRa gateway and node devices; a proxy server is used as a data relay station to receive data containing the LoRa node device ID and RSSI sent from the LoRa gateway; a listening program written in Java stores the received data locally from the proxy server; the collected data is first processed using an interpolation-amplitude limiting filtering algorithm to remove abnormal data while retaining the overall characteristics of the data; then, a Gaussian-Naive Bayes algorithm is used to train the model to obtain a fingerprint database based on the LoRa signal reception strength. In the online positioning phase: the RSSI data of the LoRa node devices is collected for the device or person to be located, and the location is estimated based on the trained model. This method can monitor intrusion into a specific indoor area and further infer the specific location information of the object or person within the room, even without the person carrying any positioning device.

Description

Technical Field

[0001] The present invention relates to a new positioning technology, specifically a method for using LoRa signal fingerprint to determine whether there is a person in a certain indoor area and to determine the specific location of the person if the person exists. Background Art

[0002] From the ancient invention of the compass technology to the modern radio-based positioning technology, every advancement in positioning technology has brought great convenience to people's daily production and life. Most of the commonly used indoor positioning technologies currently require the deployment of certain infrastructure, such as Bluetooth, RFID, and cellular networks. These positioning technologies have different advantages and disadvantages, and each has its own limitations in positioning accuracy and applicable scenarios. In an indoor area, determining whether there are people or determining their specific location when there are people is a hot issue. For example, in a special area in a large museum or prison, it is necessary to monitor the personnel in this area (whether there is human intrusion), so that corresponding processing can be made in a timely manner.

[0003] According to the different positioning methods, positioning technology can be divided into two categories: active positioning and passive positioning.

[0004] Active positioning: This positioning method requires the object or person to be positioned to carry a positioning device and actively participate in the positioning process when positioning is needed. The more common wireless signals used for active positioning include Wi-Fi, Bluetooth, RFID, and infrared. There are mainly the following methods for active positioning:

[0005] 1) Based on signal fingerprint. This method is mainly divided into two stages: offline stage and online stage. In the offline stage, the experimenter divides the area to be located into small units, and collects the RSSI of the signal transmitted by each transmitter at the center point of the unit area, and stores it as a signal fingerprint. In the online positioning stage, according to the real-time collected data to be located, the machine learning algorithm is used to optimally match the data in the fingerprint library to obtain the predicted position.

[0006] 2) Based on arrival time. This method requires strict synchronization of the transmitter and receiver clocks, and uses the signal propagation speed and the signal propagation time in space to calculate the distance between the transmitter and the receiver. Then, other methods, such as trilateral positioning, are used to solve the position.

[0007] 3) Based on arrival time difference. Similar to the arrival time-based method, but without the need for clock synchronization between the transmitter and the receiver, it is only necessary to calculate the time difference between the transmitter and the two different receivers and the propagation speed of the signal in space to obtain the distance difference between the transmitter and the two receivers, and further solve the position coordinates of the object according to the definition of the hyperbola.

[0008] 4) Based on the angle of arrival. This method requires a special antenna array, and by calculating the signal propagation angle between different receivers, the position of the object can be determined based on two straight lines.

[0009] Passive positioning: This positioning method does not require the object or person to be located to carry any positioning equipment. It only requires the person or object to enter a certain monitoring area, and the positioning system will automatically identify whether there are people or objects entering or leaving. The main idea is: in the area to be monitored, first deploy LoRa node devices and LoRa gateway devices at fixed positions. The LoRa signal for communication between the two will exist in the indoor space. When people enter and leave the area, it will affect the propagation of the LoRa signal in the indoor space. The RSSI of each LoRa node can be collected and analyzed to determine whether there is a human intrusion; and in the case of the presence of people, the position of the person in the indoor area can be further inferred by analyzing the change pattern of the LoRa signal RSSI when there are people in different areas of the room.

[0010] In recent years, with the vigorous development of the Internet of Things industry, LPWAN (Low-Power Wide-Area Network) technology has gradually entered people's field of vision, and LoRa technology, as one of the representatives, has been widely studied by scholars since its launch due to its long transmission distance, low power consumption, high receiving sensitivity, etc. The present invention uses LoRa related equipment to build an indoor passive positioning system, and infers the location information of personnel by analyzing the collected data. Summary of the invention

[0011] The problem to be solved by the present invention is how to track and monitor the location information of an object or person to be located in a certain area indoors when the object or person to be located does not carry any device to be located. This method only requires the installation of simple LoRa-related devices in the monitoring area to perform intrusion detection and position judgment on people or objects who do not carry any device.

[0012] The method for realizing indoor passive positioning by using LoRa signal fingerprint of the present invention comprises the following steps:

[0013] 1) In the area to be monitored, use LoRa related devices (mainly including LoRa gateway devices, LoRa node devices and network servers) to build an indoor signal acquisition system. The system can send uplink signal frames to the LoRa gateway device through the LoRa node device. After receiving the signal, the LoRa gateway device can parse the RSSI value of the node device; the LoRa gateway device then binds the RSSI value and its corresponding LoRa node device ID into a data frame and sends it to the network server; write a Java program in the server to receive the collected RSSI value and store it on the server.

[0014] 2) In the data processing stage, the collected RSSI data is filtered using the difference-limiting filtering algorithm to remove abnormal data and retain the overall characteristics of the data.

[0015] 3) The Gaussian-Naive Bayes algorithm is used to train the model on the filtered data to obtain an RSSI fingerprint library. For the data to be located, the best matching position is obtained according to the model, that is, the predicted location information of the object or person.

[0016] The beneficial effects of the present invention are as follows: the present invention only needs to deploy a small number of LoRa node devices and gateway devices at fixed locations. During the positioning stage, the object to be positioned or the person to be positioned does not need to carry any device to be positioned, such as a mobile phone, a receiver or a transmitter, etc., which can greatly reduce the cost overhead; and the present invention is suitable for certain scenarios where personnel need to be forcibly positioned, such as personnel intrusion detection in a special area of ​​a large museum or prison, without the need for active cooperation of the person to be positioned. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a network architecture diagram for collecting data in an indoor passive positioning system based on LoRa devices;

[0018] Figure 2 It is the division of the indoor area to be located;

[0019] Figure 3 and Figure 4 It is the spatial distribution diagram of RSSI values ​​of different LoRa nodes at different locations;

[0020] Figure 5 and Figure 6 They are the distribution diagrams of RSSI values ​​before and after the collected data is processed using the difference-limiting filter algorithm;

[0021] Figure 7 This is the overall architecture diagram for indoor passive positioning based on LoRa signal fingerprint. DETAILED DESCRIPTION

[0022] The present invention is further described below with reference to the accompanying drawings and examples.

[0023] To realize the present invention, the following problems need to be solved: 1) realize the communication between LoRa nodes and LoRa gateway devices; 2) send and receive data to the network server in the form of message queues; 3) retain the overall characteristics of the data as much as possible and eliminate abnormal data caused by accidental factors; 4) generate a signal fingerprint library according to the machine learning algorithm.

[0024] The indoor passive positioning system based on LoRa signal fingerprint mainly consists of three components: LoRa node device, LoRa gateway device and network server. The overall network architecture diagram is shown in the attached figure. Figure 1 shown.

[0025] 1)LoRa node device.

[0026] The LoRa node device is fixed in advance at a fixed position in the area to be positioned. During its operation, the LoRa node device will send an uplink signal frame to the LoRa gateway device. The data frame structure is shown in Table 1 below. Since there are multiple node devices, it is necessary to set an ID for each node device for distinction, where id1 represents the ID name of the node device, and time1 is the time when the node sends the signal frame structure. All LoRa nodes use polling to send signals to the same LoRa gateway.

[0027] Table 1 Signal frame structure sent by LoRa node device

[0028] LoRa node device ID: id1 The time when the node sends the signal frame: time1

[0029] 2) LoRa gateway device.

[0030] The LoRa gateway device also needs to be fixed at a certain position indoors in advance. After receiving the uplink signal frame sent by the LoRa node, it parses it to obtain the RSSI value of the node device and then sends it to the network server. The signal frame structure sent in this process is shown in Table 2 below, where index is the index value of the signal frame sent, id is the identifier of the LoRa node device, time is the time when the gateway sends the signal frame, and rssi is the RSSI value of the LoRa node corresponding to ID id.

[0031] Table 2 Signal frame structure sent by LoRa gateway to network server

[0032] Data frame index: index LoRa node ID: id Gateway sending time: time LoRa node RSSI value: rssi

[0033] 3) Network server.

[0034] The present invention uses an Apache-Apollo proxy server as a network server, which can receive signal frames sent by the MQTT protocol and store them under a unique topic (subscription number), waiting for the subscriber of the subscription number to take out the message from the message queue.

[0035] During the data collection phase, in indoor areas such as Figure 2 The LoRa related devices are deployed as shown. According to the above steps, data can be sent from the LoRa node device to the LoRa gateway device, and then the data can be sent from the LoRa gateway device to the network server. Finally, the message in the proxy server can be retrieved and stored locally.

[0036] Attached Figure 3 and 4 This is the spatial distribution diagram of the data collected using this system, where dots of different colors represent different LoRa node devices. Figure 2 The RSSI values ​​of the three LoRa node devices at the six locations identified in Figure 3 is the spatial distribution diagram of RSSI of different LoRa nodes at positions 1, 3 and 6, Figure 4 This is the spatial distribution diagram of RSSI at all 6 locations. It can be clearly seen from the figure that at different locations, the spatial distribution of RSSI of LoRa nodes is hierarchical. We also learn and train the data model based on this law, and then judge the location information of personnel through real-time data collection.

[0037] For the collected data, the difference-limited filtering algorithm is first used to process the abnormal or mutation data to ensure the stability of the data, while also retaining the overall characteristics of the data for subsequent data training. Figure 5 and 6 They are the RSSI distribution diagrams at a certain location before and after using the difference-limiting filtering algorithm.

[0038] After preprocessing the data using the filtering algorithm, the Gaussian-Naive Bayes algorithm is used to divide the data set and train the model, and finally a fingerprint library model based on the LoRa signal is obtained. In the online positioning stage, according to the real-time collected data, it is optimally matched with the data in the fingerprint library according to the trained model, and the result is the inferred location information.

[0039] By dividing the indoor area to be located in a real environment, collecting data at different locations and further processing and analyzing it, the experiment proves that the solution can achieve an accuracy of 98.5% within a range of 3 meters to determine whether there are intruders or objects in the area, and according to further experiments, it has achieved an average accuracy of 95% within a range of 3 meters to determine the specific location information of people or objects in the case of the presence of people or objects.

[0040] The present invention has many specific application scenarios, such as certain special areas of a large museum or a prison. The above is only a preferred implementation scheme of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principle of the present invention, and these improvements should also be regarded as the scope of protection of the present invention.

Claims

1. A method for indoor passive positioning using LoRa signal fingerprint, characterized in that: The following steps are involved: 1) In the indoor area to be located, a passive positioning system based on LoRa signal fingerprint is built to send signals from the LoRa node device to the LoRa gateway device, the RSSI of the node device is calculated on the gateway device, and the collected data is transmitted to the network server, and a Java program is written to obtain data information on the server; This step uses a passive method to collect data, that is, for people who invade a certain area, they are not expected to carry any positioning equipment that can emit signals. Instead, relevant equipment is deployed in the area in advance, and the impact of people or objects on the signal strength value of spatial distribution is used to determine the intrusion detection of people and their location information; 2) The RSSI data collected in step 1) is used to remove abnormal data through the difference-limiting filtering algorithm, and the data set is divided into different lists according to the node ID information. Then, according to different situations, the RSSI data is re-divided into a new list set according to the fixed node ID order; In the data filtering stage, the difference-limiting filtering algorithm is used: first, the difference Δ between the collected adjacent data is calculated i , and then calculate all the differences Δ i The average value Δ, and then Δ and Δ i Compare them. If the difference between the two is within the given threshold δ, keep the current RSSI data. If the difference between the two is greater than the threshold δ, continue to compare RSSI i and RSSI i+1 If RSSI i Greater than RSSI i+1 , then RSSI i -δ as a replacement value for the current data value, otherwise RSSI i +δ as a replacement value for the current data value; 3) The newly generated list set is locally used for model learning and training, and the Gaussian-Naive Bayes algorithm is used to train the data set to generate a LoRa signal fingerprint library. The algorithm can achieve binary classification and multi-classification; The Gaussian-Naive Bayesian algorithm is used to train the model: this algorithm is used to solve the maximum value of argmaxP(yk|x) under a given prediction set x. According to Bayesian theorem and conditional independence hypothesis theory, it is finally simplified to: argmaxP(yk|x)=argmaxP(yk)ПP(xi|yk), where P(yk) is the prior probability, and P(xi|yk) can be solved according to the Gaussian normal density distribution function. 4) For the data to be located collected in the real-time positioning stage, it is input into the trained model and optimally matched with the data in the fingerprint library. The result obtained is the final predicted location information.

Images