Data analyses using compressive sensing for internet of things (IoT) networks

Patent metadata

Field	Value
Publication number	US-10560530-B2
Application number	US-201816008351-A
Country	US
Kind code	B2
Filing date	Jun 14, 2018
Priority date	Feb 16, 2018
Publication date	Feb 11, 2020
Grant date	Feb 11, 2020

Abstract

Internet of Things (IoT) devices ( 101 A) continuously capture raw data over a regular interval of time. The captured raw data is transmitted to gateway devices ( 101 B) deployed in an environment, for example, a warehouse. Continuous transmission of such data leads to data redundancy, continuous channel utilization and bandwidth usage, etc. To overcome this problem, present disclosure implements a Compressive Sensing based Data Prediction (CS-DP) model that predicts data at the gateway devices by learning the data pattern received from IoT devices, estimates and computes, using a Compressive Sensing based Data Estimation (CS-DE) model, optimal data instead of considering the overall data captured at the gateway devices and reconstructs, using a Compressive Sensing based Data Reconstruction (CS-DR) model, missing data and/or corrupted data using the partial information received at the gateway devices.

First claim

What is claimed is:

1. A processor implemented method, comprising:
receiving, by one or more gateway devices, using a Compressive Sensing based Data Prediction (CS-DP) model, observation data from one or more Internet of Things (IoT) devices at one or more time intervals, wherein the observation data is received when the one or more gateway devices are in an adaptation phase, and wherein the observation data comprises a plurality of data packets ( 202 );
capturing by the one or more gateway devices, a sequence of raw data samples from the received observation data in a data vector representation ( 204 );
computing at each of the one or more gateway devices, a first prediction error based on a residual weighted vector and a prediction vector, wherein the prediction vector is computed using the sequence of raw data samples captured in the data vector representation, and wherein the residual weighted vector is computed using the prediction vector ( 206 );
switching the one or more gateway devices from the adaptation phase to an estimating phase and computing at each of the one or more gateway devices, using a Compressive Sensing based Data Estimation (CS-DE) model, bind information for each of the one or more IoT devices, and transmitting thereof, wherein the bind information is computed when the first prediction error is less than a first pre-defined threshold ( 208 );
obtaining a second prediction error from each of the one or more IoT devices, wherein the second prediction error is computed based on the bind information received from the one or more gateway devices ( 210 ); and
switching the one or more gateway devices from the estimating phase to a prediction phase and identifying and re-constructing at least one of missing data and corrupted data in the plurality of data packets, using a Compressive Sensing based Data Reconstruction (CS-DR) model, wherein the at least one of missing data and corrupted data in the plurality of data packets are reconstructed when the second prediction error is less than a second pre-defined threshold ( 212 ).

2. The processor implemented method of claim 1 , wherein when the first prediction error is greater than the first pre-defined threshold, the method comprises switching the one or more gateway devices to an adaptation phase and enabling training of the switched one or more gateway devices.

3. The processor implemented method of claim 1 , wherein when the second prediction error is greater than the second pre-defined threshold, the method comprises receiving by the one or more gateway devices, the observation data from the one or more IoT devices.

4. The processor implemented method of claim 1 , further comprising computing a prediction value by each of the one or more IoT devices based on the bind information.

5. The processor implemented method of claim 4 , further comprising computing a residual weighted value by each of the one or more IoT devices based on the prediction value and the bind information.

6. The processor implemented method of claim 5 , further comprising computing an updated weighted value by each of the one or more IoT devices based on the bind information, and an updated predicted value by each of the one or more IoT devices based on the updated weighted value.

7. The processor implemented method of claim 6 , wherein the second prediction error is computed based on the updated predicted value.

8. An Internet of Things (IoT) network system ( 100 ) comprising:
one or more IoT devices ( 101 A);
one or more gateway devices ( 101 B) in communication with the one or more IoT devices ( 101 A), wherein each of the one or more IoT devices ( 101 A) and the one or more gateway devices ( 101 B) comprises:
a memory ( 102 ) storing instructions;
one or more communication interfaces ( 106 ); and

one or more hardware processors ( 104 ) coupled to the memory ( 102 ) via the one or more communication interfaces ( 106 ), wherein the one or more hardware processors ( 104 ) are configured by the instructions, and wherein when the hardware processors ( 104 ) executes the instructions, the one or more gateway devices ( 101 B) are configured to:
receive during an adaptation phase, using a Compressive Sensing based Data Prediction (CS-DP) model, observation data from one or more Internet of Things (IoT) devices at one or more time intervals, wherein the observation data comprises a plurality of data packets;
capture, a sequence of raw data samples from the received observation data in a data vector representation;
compute at each of the one or more gateway devices ( 101 B), a first prediction error based on a residual weighted vector and a prediction vector, wherein the prediction vector is computed using the sequence of raw data samples captured in the data vector representation, and wherein the residual weighted vector is computed using the prediction vector;
switch from the adaptation phase to an estimating phase and compute at each of the one or more gateway devices ( 101 B), using a Compressive Sensing based Data Estimation (CS-DE) model, bind information for each of the one or more IoT devices, and transmitting thereof, wherein the bind information is computed when the first prediction error is less than a first pre-defined threshold;
obtain a second prediction error from each of the one or more IoT devices ( 101 A), wherein the second prediction error is computed based on the bind information received from the one or more gateway devices ( 101 B); and
switch the one or more gateway devices ( 101 B) from the estimating phase to a prediction phase and identify and re-construct at least one of missing data and corrupted data in the plurality of data packets, using a Compressive Sensing based Data Reconstruction (CS-DR) model, wherein the at least one of missing data and corrupted data in the plurality of data packets are reconstructed when the second prediction error is less than a second pre-defined threshold.

9. The IoT network system of claim 8 , wherein when the first prediction error is greater than the first pre-defined threshold, the one or more gateway devices are switched to an adaptation phase and training is enabled for the switched one or more gateway devices.

10. The IoT network system of claim 8 , wherein when the second prediction error is greater than the second pre-defined threshold, the method comprises receiving by the one or more gateway devices, observation data from the one or more IoT devices.

11. The IoT network system of claim 8 , wherein each of the one or more IoT devices are configured by the instructions to compute a prediction value based on the bind information.

12. The IoT network system of claim 11 , wherein each of the one or more IoT devices are further configured by the instructions to compute a residual weighted value based on the prediction value and the bind information.

13. The IoT network system of claim 12 , wherein each of the one or more IoT devices are further configured by the instructions to compute: an updated weighted value based on the bind information, and an updated predicted value based on the updated weighted value.

14. The IoT network system of claim 13 , wherein the second prediction error is computed based on the updated predicted value.

15. One or more non-transitory machine readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors causes:
receiving, by one or more gateway devices, using a Compressive Sensing based Data Prediction (CS-DP) model, observation data from one or more Internet of Things (IoT) devices at one or more time intervals, where

Assignees

• Tata Consultancy Services Ltd

Inventors

• Karjee Jyotirmoy
• Rath Hemant Kumar
• Pal Arpan

Classifications

• H04L43/16

  Threshold monitoring · CPC title

• H04L43/0829

  Packet loss · CPC title

• H04L43/028

  by filtering · CPC title

• H04L43/045

  for graphical visualisation of monitoring data · CPC title

• H04L69/04

  Protocols for data compression, e.g. ROHC · CPC title