Artificial intelligence fault localization in 5G and 6G messages
Upon receiving a corrupted message in 5G or 6G, a receiver generally rejects the message or ignores it entirely, because determining which message elements are faulted is difficult and complex. AI-based procedures are provided for localizing faults in specific message elements, and for determining the corrected values when possible. AI inputs may include the amplitude or phase modulation quality of each message element, the measured SNR of each message element, the modulation quality of a preceding demodulation reference, and current backgrounds, among other factors. After training (adjusting according to measured network data), the AI model may then determine the most likely faulted message elements, and may also direct the search for the most likely corrected values. By recovering the original corrected message without an unnecessary retransmission, the system can save time, reduce transmission energy, and avoid generating backgrounds. Many additional aspects are disclosed.
1. A computer containing software comprising an artificial intelligence (AI) model, the computer configured to:
a) take as input a plurality of input data comprising:
i) a received wireless message, the message comprising message elements, each message element modulated according to a modulation scheme, the modulation scheme comprising a plurality of predetermined amplitude modulation levels or a plurality of predetermined phase modulation levels or both; and
ii) a modulation quality of each message element, the modulation quality based on an amplitude difference between an amplitude of the message element and one of the predetermined amplitude modulation levels of the modulation scheme, or a phase difference between a phase of the message element and one of the predetermined phase modulation levels of the modulation scheme, or both; and
b) provide as output a prediction of which message elements are faulted.
2. The computer of claim 1 , wherein the message complies with 5G or 6G technology.
3. The computer of claim 1 , wherein the input parameters further comprise a signal-to-noise ratio of each message element.
4. The computer of claim 1 , wherein the input parameters further comprise a received demodulation reference, the demodulation reference comprising reference elements, each reference element comprising at least one amplitude level of the modulation scheme or at least one phase level of the modulation scheme.
5. The computer of claim 1 , wherein the input parameters further comprise a measure of a background noise or an interference level related to the message.
6. The computer of claim 5 , wherein the input parameters further comprise an expected format of the message.
7. The computer of claim 5 , wherein the input parameters further comprise:
a) a retransmitted version of the message, the retransmitted version comprising retransmitted elements; and
b) a modulation quality of each retransmitted element.
8. The computer of claim 5 , wherein the message has a particular type, and the input parameters further comprise one or more previously received messages of the particular type.
9. Non-transitory computer-readable media in a base station or a user device of a wireless network, the media containing instructions comprising an algorithm that when implemented by a computing environment causes a method to be performed, the method comprising:
a) taking, as input, a message comprising modulated message elements, a modulation quality of each message element, and an indication of a noise or interference level associated with each of the message elements; and
b) providing, as output, a prediction of which message elements, if any, are faulted; wherein
c) the modulation quality of each message element is based on an amplitude difference between a measured amplitude value of the message element and a predetermined amplitude level of a modulation scheme, or a phase difference between a measured phase value of the message element and a predetermined phase level of the modulation scheme.
10. The media of claim 9 , wherein the algorithm is derived using artificial intelligence.
11. The media of claim 9 , wherein the modulation quality of each message element is further related to a signal-to-noise ratio of the message element.
12. The media of claim 9 , wherein the modulation quality if each message element is inversely related to the amplitude difference or the phase difference or both, such that a larger amplitude or phase difference corresponds to a lower modulation quality, and a smaller amplitude or phase difference corresponds to a higher modulation quality.
13. The media of claim 9 , the method further comprising providing, as output, a corrected amplitude value or a corrected phase value of a message element that is predicted to be faulted.
14. The media of claim 9 , the method further comprising receiving or determining a threshold, and determining that each message element with a modulation quality lower than the threshold is faulted.
15. The media of claim 14 , wherein the algorithm is adapted for use by the base station to predict which message elements of an uplink message are faulted.
16. The media of claim 14 , wherein the algorithm is adapted for use by the user device to predict which message elements of a downlink or sidelink message are faulted.
17. A method for determining which message elements, of a wireless message, are faulted, the method comprising:
a) receiving a message and determining that a hash value associated with the message disagrees with an associated error-detection code;
b) determining a modulation quality of each message element, the modulation quality comprising at least one of:
i) an amplitude difference between a measured amplitude of the message element and a predetermined amplitude level of a modulation scheme; or
ii) a phase difference between a measured phase of the message element and a predetermined phase level of the modulation scheme;
c) providing the message and the modulation quality of each message element as input to an algorithm; and
d) determining, as output from the algorithm, which message element or elements is/are faulted.
18. The method of claim 17 , further comprising:
a) measuring a signal-to-noise ratio of each message element; and
b) providing the signal-to-noise ratio of each message element to the algorithm as additional input.
19. The method of claim 17 , wherein the algorithm is derived from an artificial intelligence model trained using machine learning.
20. The method of claim 17 , wherein the algorithm is adapted for use by a base station or a mobile user device of a wireless network.