AI-based correction of corrupted 5G/6G messages
A central challenge in next-generation 5G/6G networks is achieving high message reliability despite very dense usage and unavoidable signal fading at high frequencies. To provide enhanced fault detection, localization, and mitigation, the disclosed procedures can enable an AI model (or an algorithm derived from it) to discriminate between faulted and unfaulted message elements according to signal quality, modulation parameters, and other inputs. The AI model can estimate the likelihood that each message element is faulted, and predict the most probable corrected value, among other outputs. The AI model can also consider the quality of a demodulation reference used to demodulate the message, and the quality of the associated error-detection code. The AI model can also consider previously received messages to the same receiver, or messages of a similar type. Fault mitigation by the receiver can save substantial time and resources by avoiding a retransmission. Many other aspects are disclosed.
1. A method for a wireless receiver to mitigate message faults, the method comprising:
a) receiving a message comprising message elements, each message element comprising one resource element of a resource grid;
b) demodulating each message element according to a modulation scheme, the demodulating based on a demodulation reference received proximate to the message;
c) for each message element, determining a signal quality according to a signal comprising the message element;
d) for each message element, determining a modulation quality based on the demodulation reference;
e) providing, as input to an AI (artificial intelligence) model, the signal quality and the modulation quality of each message element; f) operating the AI model; and
g) determining, as output from the AI model, an indication of which message elements are faulted.
2. The method of claim 1 , wherein the message is received according to 5G or 6G technology.
3. The method of claim 1 , wherein:
a) the signal quality of each message element comprises a SNR (signal-to-noise ratio) of the signal comprising the message element: and
b) the modulation quality of each message element comprises a difference between a modulation of the message element and a closest modulation level of the modulation scheme.
4. The method of claim 1 , wherein the operating the AI model comprises powering a processor containing the AI model while providing the input or inputs to the AI model, and while recording the output or outputs of the AI model.
5. The method of claim 1 , wherein the output further comprises, for each message element, a likelihood that the message element is faulted.
6. The method of claim 1 , further comprising determining, as further output from the AI model, for each faulted message element, a substitute value of the faulted message element.
7. The method of claim 6 , wherein the output further comprises, for each faulted message element, a likelihood that the substitute value of the faulted message element is correct.
8. The method of claim 1 , further comprising providing, as further input to the AI model, a signal quality of each resource element of the demodulation reference.
9. The method of claim 1 , further comprising providing, as further input to the AI model, data about a background comprising interference and noise occurring proximate to the message.
10. The method of claim 1 , further comprising providing, as further input to the AI model, data about previous messages previously received by the wireless receiver.
11. The method of claim 1 , further comprising:
a) receiving a second copy of the message;
b) determining, for each message element of the second copy, a signal quality and a modulation quality;
c) providing, as further input to the AI model, the signal quality and the modulation quality of each message element of the second copy; and
d) determining, as further output from the AI model, an indication of a corrected version of the message.
12. An artificial intelligence (AI) model comprising instructions in a processor, the AI model further comprising:
a) a first plurality of inputs, each input comprising a parameter of a wireless message, wherein the wireless message comprises message elements, each message element comprising one resource element of a resource grid;
b) a second plurality of internal functions, each internal function configured to perform calculations based on the inputs of the first plurality or on other internal functions of the second plurality; and c) one or more outputs, each output functionally related to the internal functions, wherein each output comprises an indication of a message element that is predicted to be faulted.
13. The AI model of claim 12 , wherein the AI model is a neural net or a hidden Markov model.
14. The AI model of claim 12 , wherein the first plurality of inputs comprise a SNR (signal-to-noise ratio) of each message element.
15. The AI model of claim 12 , wherein the first plurality of inputs further comprise a modulation quality of each message element, wherein the modulation quality comprises a difference between a modulation of the message element and a closest predetermined modulation level of a modulation scheme.
16. The AI model of claim 12 , wherein the first plurality of inputs further comprise an amplitude and a phase of a raw waveform of each message element, wherein the raw waveform comprises a received signal of each message element.
17. Non-transitory computer-readable media in a wireless receiver, the non-transitory computer-readable media containing instructions that, when executed by a computing environment, cause a method to be performed, the method comprising:
a) providing, as input to an algorithm contained in a processor of the wireless receiver, parameters of a message comprising message elements, each message element comprising one resource element of a resource grid; and
b) determining, as output from the algorithm, an indication of which message elements are faulted;
c) wherein the parameters of the message comprise a signal quality of each message element, a modulation quality of each message element, and a noise or interference level associated with the message.
18. The non-transitory computer-readable media of claim 17 , wherein the algorithm comprises an AI (artificial intelligence) model trained, using machine learning, to recognize faulted message elements based on the parameters of each message element.
19. The non-transitory computer-readable media of claim 17 , wherein the algorithm is derived from an artificial intelligence model.
20. The non-transitory computer-readable media of claim 17 , further configured to provide, as output, for each message element that is predicted to be faulted, a substitute message element, wherein the substitute message element is predicted to be not faulted.