Description:An energy-optimized system for tracking and localization of objects in wireless sensor networks

FIELD OF INVENTION
[0001] The present invention primarily relates to the field of tracking and localization of sensor nodes in the wireless sensor networks (WSN). The field of the invention is to provide a method for energy-efficient localization of sensor nodes or smart objects in the wireless sensor networks (WSN).
[0002] More particularly, this present invention relates to the field of process and system for energy efficient tracking and localization of sensor nodes which may be smart objects in wireless sensor networks (WSN) using hybrid approach based on machine learning models (artificial neural network) and parameters-based calculation.

BACKGROUND & PRIOR ART

[0003] The subject specification in substance discussed in the background section of the present application should not be understand to be the prior art merely because of its discussion in this background section. Similarly, the technical problem described in the background section or problem-solution regarding the subject-matter of this background section should not be understand to have been previously identified in the prior art.

[0004] Wireless sensor networks are basically the cluster/group organization of sensor nodes. Wireless sensor networks comprise large number of small, low-powered, low cost and having limited processing capabilities sensor nodes. The wireless sensor networks (WSN) are a combination of spatially or geographically distributed smart sensor objects or Internet of things (IoT) devices which have data sensing capabilities and embedded identification through RFID technology. Particularly, the integration of sensor nodes, RFID tags/devices, and transferring/receiving technologies forms the network of Wireless sensor Networks. The wireless sensor networks address the traceability and controllability of physical objects or smart devices. The wireless sensor networks together with the smart objects via communicating network and other computing devices transforms real world environment into smart objects which can sense the environmental physical objects and communicate the usable data accordingly. There is a wide list of usability or applications of wireless sensor networks like environmental monitoring, healthcare services, food supply chain, transportation and logistics, smart homes, driving automation, forest animals tracking, social networking etc.

[0005] These sensor nodes in the wireless sensor networks (WSN) are powered through battery and the life of the sensor nodes is basically dependent on the life of the battery power. Radio frequency identification is used in the sensor nodes for tracking of objects, people and animals. Electronic Product Codes are encoded in radio frequency identification tags which can be used to track/locate objects through wireless sensor networks. There are number of smart sensor objects in the wireless sensor networks that have limited battery and computing capabilities. Further, the usability of the data collected through these sensor nodes is location dependent i.e., the geographic location of the sensor nodes plays a vital role in the usability of the sensed data. Thus, the life of the sensor nodes in the wireless sensor networks is dependent on the low battery usage which indirectly refers to very limited data processing.

[0006] Since, the use of these wireless sensor nodes involves data dissemination periodically to the server or other computing devices by the communication network, the usability of the disseminated data depends on the location of the sensor nodes from where the said data is disseminated. The need of the precision in the location of the sensor nodes depends on the applicability of the wireless sensor nodes i.e., whether there is a requirement of highly precise location or somewhat lesser. Determining the location of the sensor nodes in the wireless sensor network is called the tracking and localization of the sensor nodes in the wireless sensor networks. Hence, tracking and localization of the sensor nodes or objects in the wireless sensor networks plays predominant role. The sensor nodes or objects used in the wireless sensor networks is made up of large number of tiny sensor nodes or objects which does not have capability to determine their location.

[0007] The wireless sensor networks are basically a combination of two types of nodes i.e., anchor nodes and non-anchor nodes. The sensor nodes having the capability of determining their location are called anchor nodes or beacon nodes while the other nodes which do not know their location information are called non-anchor nodes. The beacon nodes are able to know their location through global positioning system (GPS) or manual setup arrangement. These anchor nodes are used to determine the location of other non-anchor nodes in the wireless sensor networks (WSN). Hence, more the anchor nodes or beacon nodes in the wireless sensor networks, the precise is the location of the sensor nodes or objects in the wireless sensor networks but the cost of these anchor nodes is much higher than that of the non-anchor nodes. Thus, usage of higher number of anchor nodes in the wireless sensor networks increases the cost of the wireless sensor networks. Further, the localization of the non-anchor nodes in the wireless sensor networks is respective of the anchor nodes in the network.

[0008] From the above paragraph, it is clear that the more the number of beacon nodes in the wireless sensor network, the precise is the location which in turn increases the cost of the wireless sensor network. Inversely, by decreasing the number of beacon nodes in the network reduces the cost but in turn provides the less precise location. Hence, there is always the unsaid need of the method which can efficiently determines the precise location of the sensor nodes or objects in the wireless sensor networks while using the lesser number of beacon nodes or anchor nodes.

[0009] Second important aspect of the sensor nodes is the battery life. The battery capacity of the sensor nodes used in the wireless sensor network is very less and having limited processing capability too. The life of the sensor nodes is mainly depending on the battery usage. Battery drainage in turn finishes the life of the sensor nodes and increasing cost of the wireless sensor networks due to replacement of these sensor nodes. Further, the usage of the processing capability of the sensor nodes consumes more battery power which in turn reduces the life of said nodes. Hence, the more the usage of the processing of the data at nodes, the less the life of the sensor nodes. Thus, to make the wireless sensor networks more robust and energy efficient, there is a need of such kind of process and system which can provide tracking and localization of the objects in efficient way without using much processing capabilities and lesser beacon nodes while consuming less amount of battery.

[0010] Thus, there is a need of the system which can automatically determines the location of the sensor nodes without repetitive calculation and less data processing i.e., to automate the localization and tracking process. One such emerging technology to automate such kind of process is machine learning models. Machine learning is a kind of artificial intelligence in which machines are trained to predict the outcome values using historical data or training dataset. Further, machine learning model also trained themselves through experiences and provides precise outcome values by time.

[0011] There are various prior arts that provides methods for tracking and localization of objects in wireless sensor networks which uses genetic algorithms. Further, the known method for localization of the sensor nodes always calculates the location of the sensor nodes which in turns reduces the life of the sensor nodes due to increased battery usage. Before going forward to the subject-matter of the present invention, let’s have a look some of the prior art of tracking and localization techniques of sensor nodes in wireless sensor networks.

[0012] US8849926 B2 – This document discloses a method for tracking and localization of sensor nodes in wireless sensor networks whereby sensor nodes are configured to broadcast consecutive messages in a predefined and incremental power levels. The master node detects the reception of broadcast messages and their power level to detect the two neighbourhood beacon nodes. Further the location of said sensor node is identified based on relative location of the two identified beacon nodes. The drawback of the said method is that there is continuous broadcast of messages to identify the location of sensor nodes which utilizes more battery of the sensor nodes and the location is determined every time broadcast messages are received.

[0013] EP3189310 B1 – The said document discloses a method for localization of sensor nodes in the wireless sensor network. The said process includes the steps of receiving the captured data from first sensor node indicating start and stop time of the event, receiving the captured data from second sensor node indicating start and stop time of the event, comparing the first sensor data with the second sensor data to determine the cross-correlation and identifying the location of the first sensor node relative to the second sensor nodes based on comparison data.

[0014] Besides this, there are a number of prior arts in the said field of art that claims to determine the location of the sensor nodes in the wireless sensor networks but all of the prior arts calculate the location of the sensor nodes every time during the reception of data. Further, the process mentioned in all of the prior art uses more battery power for localization of sensor nodes in the wireless sensor networks. Hence, there is a need to have an energy efficient method and system that provides the tracking and localization of sensor nodes with a smaller number of beacon nodes and using less battery power. Further, the aim of the present subject matter is to develop such a system that can learn itself and populates the database to automate the process of localization without calculating and reducing the battery power usage.

SUMMARY OF THE INVENTION

[0015] Before describing the present subject matter in detail, it is to be assumed or understood that this invention is not restricted to the particular systems, and process as described, as there can be multiple possible technical embodiments of the present subject matter which are not expressly indicated in the present description. It is also cleared that the terminology used in the present description is for the purpose of describing the special or particular embodiments only and is not intended to restrict the present application. The summary as provided to introduce concepts related to energy efficient systems and methods for tracking and localization of objects in the wireless sensor networks and the concepts are further described in detail in the below description.

[0016] The present claimed subject matter mainly solves the technical problems existing in the prior arts in this particular field as detailed above. In solution to these problems, the present invention discloses an energy efficient method and system for tracking and localization of objects or sensor nodes in wireless sensor networks using hybrid technology i.e., machine learning models and calculation-based method. The present invention makes a kind of system which is adaptive in nature and makes the system self-learned and populates its database to quick and efficient localization and tracking of sensor nodes in wireless sensor networks without using much battery and develop a robust system which will enhances the efficiency of the system in future using transfer learning.

[0017] In the proposed invention, the wireless sensor network uses hybrid technology to determine the location of the sensor nodes or objects in the network i.e., both machine learning model and parameter-based calculation approach to determine the location of sensor nodes. The wireless sensor network is organized in a predetermined network topology comprising anchor nodes or beacon nodes having GPS capability to determine the location of that node. One of the preferred network topologies which may be used in the wireless sensor network is mesh based network architecture. The anchor nodes are arranged in the predefined network topology as per the size of the wireless sensor network and according to the area coverage per beacon node. These beacon nodes comprise the capability to determine their location through Global positioning system or manual setup methodology. The other sensor nodes which are called as non-anchor nodes are arranged in the range of the beacon nodes as per the adopted network topology.

[0018] The proposed model i.e., adopted machine learning model of the present invention comprises two phases i.e., training phase and location determination phase. In the training phase, the proposed model uses training dataset to determine the location of the sensor nodes or objects. The training dataset or historical data representing the mapping between the parameters required in determining the location of the sensor nodes and their corresponding location based on the parameters set. The training dataset is a parameter vector or mapping associated with each anchor node and their corresponding location. The parameter vector which is used to train the adopted model of present invention consists mapping among various parameters like signal strength, hop count, signal flight time and their corresponding location coordinates of the sensor node. The network is first trained using the training dataset or historical data. Whenever, the sensor nodes disseminate date to the master node, the parameters required to determine the location of the sensor nodes are computed. These parameters required in the proposed model of the present invention are signal strength, hop count and signal flight time. These parameter values i.e., signal strength, hop count and signal flight time is matched in the dataset of the machine learning model and if the match is found in the parameter vector with respect to the values of the parameters of the sensor nodes, the location of the sensor node is predicated automatically using the adopted machine learning model and the location coordinates corresponding to the match in the dataset will be the location of the said sensor nodes. Further, if the network does not find any matched parameter values in the parameter vector, then the network calculates the location coordinates using parameters values of signal strength, hop count and signal flight time, in the location determination phase, using various methodologies already predetermined in the state of the art. The determined location coordinates and the parameter values are entered in the parameter vector for future use or to learn the machine learning model which can be used by the network in future i.e., transfer learning is used in the proposed invention to train the model in more robust way.

[0019] An aspect of the present disclosure relates to a method for energy efficient tracking and localization of sensor nodes in wireless sensor network, the method comprising: setting up the sensor nodes in the predefined network topology architecture; installing beacon nodes and fixed nodes in the predefined network architecture having self-location determination facility; training the predefined network nodes on the training dataset i.e., parameter vector according to the adopted machine learning model; receiving the disseminated data from the non-anchor nodes along with data related to the signal strength, hop count and signal flight time; comparing the received values of signal strength, hop count and signal flight time with the parameter vector data; if there is a match with the values of received signal strength, hop count and signal flight time then predict the location coordinated corresponding to the match found; else determining the location of the sensor nodes based on the calculation approach adopted by the system using parameter values signal strength, hop count and signal flight time.

[0020] An aspect of the present disclosure relates to system for energy efficient tracking and localization of sensor nodes in wireless sensor network, the method comprising: communication network for sending or receiving the data from the sensor nodes in the wireless sensor network; connecting the wireless sensor network to the outer network through the communication network; setting up the sensor nodes in the predefined network topology architecture; installing beacon nodes and fixed nodes in the predefined network architecture having self-location determination facility; training the predefined network nodes on the training dataset i.e., parameter vector according to the adopted machine learning model; receiving the disseminated data from the non-anchor nodes along with data related to the signal strength, hop count and signal flight time; comparing the received values of signal strength, hop count and signal flight time with the parameter vector data; if there is a match with the values of received signal strength, hop count and signal flight time then predict the location coordinated corresponding to the match found; else determining the location of the sensor nodes based on the calculation approach adopted by the system using parameter values signal strength, hop count and signal flight time.

OBJECTIVE OF THE INVENTION

[0021] A primary objective of the present invention is to provide a method and system for providing energy efficient localization and tracking of sensor nodes in wireless sensor network using hybrid technology based on machine learning model and calculation-based approach. Yet another objective of the present invention is to provide a methodology to provide a method that can identity the location of sensor nodes in the network by using less battery power of sensor nodes with limited processing capability. Yet another objective of the present invention is to make a robust system that can train and enrich itself to determine or predict the location of sensor nodes using machine learning model more frequently using transfer learning by using calculation data related to the location coordinates.

BRIEF DESCRIPTION OF DRAWINGS

[0022] For better clarification and various embodiments of the present invention, the invention will be detailed with the help of the drawings which will illustrate the various technical features or embodiment of the present invention. It should be appreciated further that the drawings presented in the complete specification depict only specific embodiments or technical features of the invention and are therefore not to be considered as the limiting scope of the present invention. The invention has been described using the reference numerals presented in the drawings for better intelligibility and understanding of the sought subject matter. In order to define the advantages of the present invention clearly and easily, a detail description of the invention is discussed below in conjunction with the drawings, which, however, should not be considered to restrict the scope of the claimed subject matter to the accompanying drawings, in which:

[0023] Figure 1 shows flow diagram of the method for installing and learning the nodes as per adopted machine learning model. Figure 2 shows the flow diagram representing the process involved in determining the location of sensor nodes in accordance with the hybrid approach of machine learning model. Figure 3 shows a block-diagram of proposed system.

DETAIL DESCRIPTION

[0024] The present invention relates to a method and system for energy efficient tracking and localization of sensor nodes using hybrid approach of machine learning model in accordance with the proposed invention. Although the present disclosure has been described with the purpose of providing a localization and tracking of sensor nodes in wireless sensor networks using hybrid technology, it should be appreciated that the same has been done merely to illustrate the different embodiments of the subject invention. Further, it should also be noted that the present disclosure fully and particularly describes the present invention and discloses the best method of performing the present invention.

[0025] Figure 1 show a flow diagram representing the installing and learning of nodes in accordance with the adopted machine learning model. The wireless sensor network is organized in a predetermined network topology comprising anchor nodes or beacon nodes having GPS capability to determine the location of that node. One of the preferred network topologies which may be used in the wireless sensor network is mesh based network architecture. The anchor nodes and fixed nodes are arranged in the predefined network topology as per the size of the wireless sensor network and according to the area coverage per beacon node. These beacon nodes comprise the capability to determine their location through Global positioning system or manual setup methodology. Further, the fixed nodes are static in nature and having fixed location coordinates and do not change over time. The other sensor nodes which are called as non-anchor nodes are arranged in the range of the beacon nodes as per the adopted network topology. In the training phase, the proposed model uses training dataset to determine the location of the sensor nodes or objects. The training dataset or historical data representing the mapping between the parameters required in determining the location of the sensor nodes and their corresponding location based on the parameters set. The training dataset is a parameter vector or mapping associated with each anchor node and their corresponding location. The parameter vector which is used to train the adopted model of present invention consists mapping among various parameters like signal strength, hop count, signal flight time and their corresponding location coordinates of the sensor node. The network is first trained using the training dataset or historical data. Further, the data related to the parameter values and the calculated location coordinates in the location determination phase are also saved in the parameter vector for future learning through experiences which is called transfer learning.

[0026] Figure 2 shows the flow-diagram representing the process steps involved in localization and tracking of sensor nodes in the wireless sensor networks. Whenever the sensor nodes disseminate the data over the wireless sensor network, the data is received by the beacon nodes or nodes having the capability to process the data. The parameter values i.e., signal strength, hop count and signal flight time are calculated and the same values are compared with the values in the parameter vector and determines the match. If the match is found then the corresponding location coordinates are the location of the sensor nodes else the location coordinates are determined using the location determination technique that are already known in the state of the art using the parameter values signal strength, hop count and signal flight time.

[0027] Figure 3 shows the block-diagram of proposed system comprising a communication network (201) which is used to connect the wireless sensor network with the outer network or other computing devices. The system comprises the wireless sensor network comprises of beacon nodes and fixed nodes having capability to determine their location. Further, these nodes are trained using the adopted machine learning model using training dataset having parameter vector which provides mapping between parameter values i.e., signal strength, hop count and signal flight time with the location coordinates of the sensor nodes. The non-anchor nodes disseminate the data to the network and the as and when the data is received the location coordinates of the sensor nodes are determined using the process as mentioned above.
[0028] In this way, the present invention or the machine learning model is self-learned model and the dataset i.e., parameter vector is continuously enriched by the adopted machine learning model. Further, the proposed system uses less energy of the sensor nodes as the network only calculates the location coordinates when the parameter vector does not find any match with the parameters. The proposed invention sufficiently optimizes the energy utilization while determining or tracking the location of the sensor nodes or objects in the wireless sensor networks.

[0029] The proposed invention is different with the prior art in that, in the prior art, location of objects is calculated every time or continuously while changing or receiving the parameters to find the location of wireless sensors in the WSN. The Proposed model is self-learned model which utilizes the parameter vector which is a mapping of various parameters or features used to determine the location of the objects and map the said parameters with the location of the objects or sensor nodes. The proposed invention calculates the location of the sensor nodes only when the system does not find any match in the parameter vector related to the sensor nodes. The advantage of this proposed invention is that, since the sensor nodes have very limited processing capability and limited power source, the system calculates the location of the sensor nodes only when suitable match is not find in the parameter vector populated by the adopted machine learning model.

, Claims:CLAIMS
We claim:
1. A computer implemented method for energy efficient localization and tracking of sensor nodes in the wireless sensor network, the method comprising:
setting up the sensor nodes in the predefined networking topology architecture;
installing beacon nodes and fixed nodes in the predefined network architecture having self-location determination facility;
training the predefined network nodes on the training dataset i.e., parameter vector according to the adopted machine learning model; receiving the disseminated data from the non-anchor nodes along with data related to the signal strength, hop count and signal flight time; comparing the received signal strength values, hop count and signal flight time with the parameter vector data;
if there is a match with the received signal strength values, hop count and signal flight time then predict the location coordinated corresponding to the match found;
else determine the location of the smart objects or sensor nodes based on the calculation approach adopted by the system using parameter values signal strength, hop count and signal flight time.

2. The method as claimed in claim 1, wherein beacon nodes having the global positioning system functionality and fixed nodes have manual setup coordinates.

3. The method as claimed in claim 1, wherein the network architecture adopted for wireless sensor network may be mesh based network topology.

4. A system for energy efficient localization and tracking of sensor nodes in the wireless sensor network, the system comprising:

a communication network for connecting and transmitting/receiving data from the wireless sensor network to the outer network or computing devices;
parameter vector or training dataset comprising mapping between parameter values and location coordinates;
transfer learning system for enriching or populating the training dataset according to experience or calculated data values;
wherein the beacon node or fixed node configured to perform the steps of;
setting up the sensor nodes in the predefined network topology architecture;
installing beacon nodes and fixed nodes in the predefined network architecture having self-location determination facility;
training the predefined network nodes on the training dataset i.e., parameter vector according to the adopted machine learning model; receiving the disseminated parameter values from the non-anchor nodes along with data related to the signal strength, hop count and signal flight time; comparing the received signal strength values, hop count and signal flight time with the parameter vector data;
if there is a match with the values of received signal strength, hop count and signal flight time then predict the location coordinated corresponding to the match found;
else determine the location of the sensor nodes based on the calculation approach adopted by the system using parameter values signal strength, hop count and signal flight time.