AI enabled sensor data acquisition
Example implementations described herein can dynamically adapt to changing nature of sensor data traffic and through artificial intelligence (AI, strike a good tradeoff between reducing volume of sensed data, and retain enough data fidelity so that subsequent analytics applications perform well. The example implementations eliminate heuristic methods of setting sensing parameters (such as DAQ sampling rate, resolution etc.) and replaces them with an automated, AI driven edge solution core that can be readily ported on any Internet of Things (IoT) edge gateway that is connected to the DAQ.
1. A method for a system comprising sensors providing sensor data, the method comprising:
a) sampling the sensor data under a maximum DAQ parameters configuration for the system;
b) learning an optimized DAQ parameters configuration for the system through a reinforcement learning process comprising:
sampling a DAQ parameters configuration from a probability distribution of DAQ parameters for the system;
comparing a first fast Fourier transform (FFT) of the sampled sensor data under the maximum DAQ parameters configuration with a second FFT of the sampled sensor data under the sampled DAQ parameters configuration; and
updating the probability distribution over DAQ parameters for the system based on the comparing;
c) setting the optimized DAQ parameters configuration for the sensors of the system based on the probability distribution over DAQ parameters;
d) sampling the sensor data with the optimized DAQ parameters configuration; and
e) executing a machine learning model configured to detect a trend or a shift based on features extracted from a spectrogram derived from the sampling of the sensor data with the optimized DAQ parameters configuration.
2. The method of claim 1 , further comprising:
for the machine learning model detecting a trend or shift, re-executing steps a), b), and c) to relearn the optimized DAQ parameters configuration.
3. The method of claim 1 , wherein the comparing the first fast Fourier transform (FFT) of the sampled sensor data under the maximum DAQ parameters configuration with the second FFT of the sampled sensor data under the sampled DAQ parameters configuration comprises:
computing a distance between the first FFT and the second FFT; and
for the distance being below a threshold, assigning a reward to the sampled DAQ parameters configuration that is inversely related to bits used under the sampled DAQ parameters configuration.
4. The method of claim 3 , wherein updating the probability distribution of DAQ parameters for the system based on the comparing is conducted based on updating probabilities for the sampled DAQ parameters configuration based on the reward.
5. A non-transitory computer readable medium, storing instructions for a process executed on a system comprising sensors providing sensor data, the instructions comprising:
a) sampling the sensor data under a maximum DAQ parameters configuration for the system;
b) learning an optimized DAQ parameters configuration for the system through a reinforcement learning process comprising:
sampling a DAQ parameters configuration from a probability distribution of DAQ parameters for the system;
comparing a first fast Fourier transform (FFT) of the sampled sensor data under the maximum DAQ parameters configuration with a second FFT of the sampled sensor data under the sampled DAQ parameters configuration; and
updating the probability distribution over DAQ parameters for the system based on the comparing;
c) setting the optimized DAQ parameters configuration for the sensors of the system based on the probability distribution over DAQ parameters;
d) sampling the sensor data with the optimized DAQ parameters configuration; and
e) executing a machine learning model configured to detect a trend or a shift based on features extracted from a spectrogram derived from the sampling of the sensor data with the optimized DAQ parameters configuration.
6. The non-transitory computer readable medium of claim 5 , the instructions further comprising:
for the machine learning model detecting a trend or shift, re-executing steps a), b), and c) to relearn the optimized DAQ parameters configuration.
7. The non-transitory computer readable medium of claim 5 , wherein the comparing the first fast Fourier transform (FFT) of the sampled sensor data under the maximum DAQ parameters configuration with the second FFT of the sampled sensor data under the sampled DAQ parameters configuration comprises:
computing a distance between the first FFT and the second FFT; and
for the distance being below a threshold, assigning a reward to the sampled DAQ parameters configuration that is inversely related to bits used under the sampled DAQ parameters configuration.
8. The non-transitory computer readable medium of claim 7 , wherein updating the probability distribution of DAQ parameters for the system based on the comparing is conducted based on updating probabilities for the sampled DAQ parameters configuration based on the reward.
9. A management apparatus for a system comprising sensors providing sensor data, the management apparatus comprising:
a processor, configured to:
a) sample the sensor data under a maximum DAQ parameters configuration for the system;
b) learn an optimized DAQ parameters configuration for the system through a reinforcement learning process comprising:
sampling a DAQ parameters configuration from a probability distribution of DAQ parameters for the system;
comparing a first fast Fourier transform (FFT) of the sampled sensor data under the maximum DAQ parameters configuration with a second FFT of the sampled sensor data under the sampled DAQ parameters configuration; and
updating the probability distribution over DAQ parameters for the system based on the comparing;
c) set the optimized DAQ parameters configuration for the sensors of the system based on the probability distribution over DAQ parameters;
d) sample the sensor data with the optimized DAQ parameters configuration; and
e) execute a machine learning model configured to detect a trend or a shift based on features extracted from a spectrogram derived from the sampling of the sensor data with the optimized DAQ parameters configuration.
10. The management apparatus of claim 9 , wherein the processor is further configured to:
for the machine learning model detecting a trend or shift, re-execute steps a), b), and c) to relearn the optimized DAQ parameters configuration.
11. The management apparatus of claim 9 , wherein the processor is configured to compare the first fast Fourier transform (FFT) of the sampled sensor data under the maximum DAQ parameters configuration with the second FFT of the sampled sensor data under the sampled DAQ parameters configuration by:
computing a distance between the first FFT and the second FFT; and
for the distance being below a threshold, assigning a reward to the sampled DAQ parameters configuration that is inversely related to bits used under the sampled DAQ parameters configuration.
12. The management apparatus of claim 11 , wherein the processor is configured to update the probability distribution of DAQ parameters for the system based on the comparing is conducted based on updating probabilities for the sampled DAQ parameters configuration based on the reward.