Predictive spectral allocation in mobile networks
View Patent ↗Technologies are generally described for discerning patterns in the “goodness” or “badness” of time-frequency slots to allow predictive allocation of spectral resources that may be appropriate for a wireless user. According to some examples, information on device location, time slots, sub-carrier(s) allotted for each time slot, and quality indicators may be received from mobile devices. The time slots may be grouped by location to form analysis intervals. A time-frequency vector may then be identified for each analysis interval and a unit of geographic grid. A “goodness” indicator may be computed for each time-frequency vector. Clusters of time-frequency vectors may be categorized for each analysis interval and associated unit of geographic grid such that mobile devices can be assigned “good” clusters through sub-carrier allocation.
1. A method to employ predictive spectral allocation in wireless networks, the method comprising:
receiving a request for sub-carrier allocation from a mobile device, the request including a timestamp and a location of the mobile device;
identifying a particular cluster based on the timestamp and the location;
selecting a time-frequency vector from the particular cluster;
transmitting information pertaining to the sub-carrier allocation to the mobile device, wherein the information is based on the time-frequency vector; and
in response to receiving an indication from the mobile device that the sub-carrier allocation is unacceptable, switching to a default sub-carrier allocation and forwarding a bad quality indication associated with the unacceptable sub-carrier allocation to an analysis server.
2. The method according to claim 1 , wherein transmitting the information comprises transmitting a sequence of frequencies to be used in consecutive time slots.
3. The method according to claim 1 , wherein identifying the particular cluster based on the timestamp and the location comprises:
receiving information on device location, time slots, sub-carrier(s) allotted for each time slot, and quality indicators from a plurality of mobile devices;
grouping the time slots by device location to form analysis intervals;
identifying a time-frequency vector for each analysis interval, wherein the time-frequency vector associates the analysis interval with the device location;
computing a goodness indicator for each time-frequency vector; and
identifying one or more clusters of time-frequency vectors for each analysis interval and associated location as good clusters and bad clusters.
4. The method according to claim 1 , further comprising:
employing a machine learning technique to identify clusters of time-frequency vectors with substantially similar “goodness” indicator value.
5. The method according to claim 4 , wherein employing the machine learning technique comprises using one or more of a neural network, a support vector machine (SVM), and a Bayes classifier.
6. The method according to claim 1 , wherein receiving the request comprises receiving the request over a wireless communication technology that utilizes a multiple access technology.
7. The method according to claim 1 , further comprising:
determining the time-frequency vector for one of a single sub-carrier frequency or a group of sub-carrier frequencies.