Spectrum surveillance
Described herein is scheme to manage a receiver assignment using information from a coarse power spectral density (PSD) obtained at every time and a systems' history of detections. The scheme of detecting or sensing signals in the spectrum applies reinforcement learning and relies on RF communications (e.g., covert communications and anomalous communication) and spectrum management. The scheme provides a low-cost and easy to deploy system of sensing signals in a wide spectrum with limited number of receivers.
1 . A method for frequency spectrum sensing with optimal number of receivers, the method comprising:
obtaining black and white space information of a spectrum and updating a state;
selecting a location for a current receiver, wherein the location represents a time step and a frequency bin in the spectrum for the current receiver to detect one or more signals in the time step and the frequency bin;
assigning the location for the current receiver; and
updating a reinforcement learning model for the location which is assigned.
2 . The method of claim 1 , wherein obtaining the black and white space information comprises performing a coarse band-of-interest detection.
3 . The method of claim 2 , wherein performing the coarse band-of-interest detection comprises determining time steps and frequency bins where a power spectral density is above a threshold.
4 . The method of claim 1 , wherein updating the state comprises updating the black and white space information and a value of a counter, and wherein the value of the counter indicates a number of time steps since the frequency bin of the spectrum was last visited and an activity was detected in the frequency bin.
5 . The method of claim 1 , wherein the spectrum is a time-varying spectrum.
6 . The method of claim 1 , wherein selecting the location comprises:
generating a random bit with a first probability of epsilon for 1 and a second probability of 1-epsilon for 0;
selecting a first position of the current receiver between 1 and N if the random bit is 1; and
selecting a second position of the current receiver based on a Q-function if the random bit is 0, wherein the Q-function is determined by updating the reinforcement learning model.
7 . The method of claim 1 , wherein updating the reinforcement learning model comprises:
collecting a reward and updating a resulting state associated with the location which is assigned.
8 . The method of claim 1 , wherein the current receiver is an IQ receiver.
9 . A system comprising:
a plurality of receivers, wherein an individual receiver of the plurality of receivers is adapted to detect an RF signal from a location in a spectrum;
a database to store signal information from the spectrum and from coarse band-of-interest information; and
a controller coupled to the plurality of receivers and the database,
wherein the controller is to:
obtain black and white space information of the spectrum from the coarse band-of-interest information, wherein the black and white space information is stored in the database;
select the location for the individual receiver, wherein the location represents a time step and a frequency bin in the spectrum for the individual receiver to detect one or more signals in the time step and the frequency bin;
assign the location for the individual receiver; and
update a reinforcement learning model for the location which is assigned.
10 . The system of claim 9 , wherein the coarse band-of-interest information is based on time steps and frequency bins of the spectrum where a power spectral density is above a threshold.
11 . The system of claim 9 , wherein the controller is to update a state via an update to the black and white space information and a value of a counter.
12 . The system of claim 11 , wherein the value of the counter indicates a number of time steps since the frequency bin of the spectrum was last visited and an activity was detected in the frequency bin.
13 . The system of claim 9 , wherein the spectrum is a time-varying spectrum.
14 . The system of claim 9 , wherein the controller is to select the location via:
a random bit generation to generate a random bit with a first probability of epsilon for 1 and a second probability of 1-epsilon for 0;
a first selection of a first position of the individual receiver between 1 and N if the random bit is 1; and
a second selection a second position of the individual receiver based on a Q-function if the random bit is 0, wherein the Q-function is determined by updating the reinforcement learning model.
15 . The system of claim 9 , wherein the controller is to update the reinforcement learning model via collection of a reward and update of a resultant state associated with the location which is assigned.
16 . The system of claim 9 , wherein the individual receiver is an IQ receiver.
17 . A machine-readable storage media having machine-readable instructions stored thereon that when executed cause one or more machines to perform a method for frequency spectrum sensing with optimal number of receivers, the method comprising:
obtaining black and white space information of a spectrum and updating a state;
selecting a location for a current receiver, wherein the location represents a time step and a frequency bin in the spectrum for the current receiver to detect one or more signals in the time step and the frequency bin;
assigning the location for the current receiver; and
updating a reinforcement learning model for the location which is assigned.
18 . The machine-readable storage media of claim 17 , wherein obtaining the black and white space information comprises performing a coarse band-of-interest detection.
19 . The machine-readable storage media of claim 18 , wherein performing the coarse band-of-interest detection comprises determining time steps and frequency bins where a power spectral density is above a threshold.
20 . The machine-readable storage media of claim 17 , wherein updating the state comprises updating the black and white space information and a value of a counter, and wherein the value of the counter indicates a number of time steps since the frequency bin of the spectrum was last visited and an activity was detected in the frequency bin.