AI-based message interpretation and fault mitigation in 5G and 6G
View Patent ↗Message faulting is expected to be a major challenge in 5G-Advanced and especially 6G, due to increased pathloss and phase noise at FR2 frequencies, and exponential crowding of networks. Legacy methods for forward-correction or automatic retransmissions are unsuitable to the fast-paced demands of next-generation users. Therefore, disclosed herein is an AI-based receiver that interprets a corrupted message to determine the most likely meaning or intent, and thereby provides one or more candidate corrected messages along with a likelihood that each of the candidate corrected messages is indeed correct. The AI model may also be provided with data on the context or current activity of the receiver, data on the waveform of each message element, and other data available to the receiver, so that the AI model can further refine the likelihood values. By recovering corrupted messages in the receiver, a costly retransmission may be avoided, saving time and resource usage.
1. A method for a wireless receiver to demodulate a corrupted message, the method comprising:
a) providing the corrupted message as input to an artificial intelligence (AI) model;
b) determining, as output from the AI model, one or more candidate corrected messages; and
c) determining, as further output from the AI model, for each candidate corrected message, a likelihood value that the candidate corrected message is correct.
2. The method of claim 1 , wherein the corrupted message is received according to 5G or 6G technology.
3. The method of claim 1 , wherein the output from the AI model comprises, associated with each of the one or more candidate corrected messages, a likelihood value that the candidate corrected message is correct.
4. The method of claim 1 , wherein the output from the AI model comprises, associated with each of the one or more candidate corrected messages, an indication of a meaning or intent of the candidate corrected message.
5. The method of claim 4 , wherein the output from the AI model further comprises, associated with each of the one or more candidate corrected messages, a likelihood value that the candidate corrected message is correct, the likelihood value based at least in part on the meaning or intent of the candidate corrected message.
6. The method of claim 1 , further comprising, as further input to the AI model, a plurality of prior messages received by the wireless receiver or by another wireless receiver, and associated with each of the prior messages, an indication of the meaning or intent of the prior message.
7. The method of claim 1 , wherein the AI model is trained according to a plurality of prior messages and, associated with each of the prior messages of the plurality, a meaning or intent of the prior message.
8. The method of claim 1 , wherein the output of the AI model further comprises a most likely corrected message, and an indication of a meaning or intent of the most likely corrected message.
9. A wireless receiver comprising a processor that includes an artificial intelligence (AI) model, the wireless receiver configured to:
a) receive a message comprising message elements, each message element comprising a modulated resource element of a resource grid;
b) determine, according to an error-detection code associated with the message, that the message or the error-detection code, or both, is or are corrupted;
c) provide, as input to the AI model, a modulation of each message element of the message and of the error-detection code; and
d) determine, as output from the AI model, a plurality of candidate corrected messages.
10. The wireless receiver of claim 9 , wherein the output from the AI model further comprises, for each of the candidate corrected messages, a likelihood that the candidate corrected message is correct.
11. The wireless receiver of claim 9 , wherein the output from the AI model further comprises, for each of the candidate corrected messages, a meaning or intent of the candidate corrected message.
12. The wireless receiver of claim 9 , wherein the input to the AI model further comprises data about a context of the message or a current activity of the wireless receiver.
13. The wireless receiver of claim 12 , wherein the output of the AI model further comprises, for each of the candidate corrected messages, a likelihood that the candidate corrected message is correct, the likelihood based at least in part on the context of the message or the current activity of the wireless receiver.
14. The wireless receiver of claim 12 , wherein the output of the AI model further comprises, for each of the candidate corrected messages, an intent or meaning of the candidate corrected message, the intent or meaning based at least in part on the context of the message or the current activity of the wireless receiver.
15. Non-transitory computer-readable media in a wireless receiver, the non-transitory computer-readable media comprising an artificial intelligence (AI) model and instructions that, when implemented in a computing environment, cause a method to be performed, the method comprising:
a) receiving a wireless message and determining that the wireless message is corrupted;
b) providing the wireless message as input to the AI model; and
c) determining, as output from the AI model, one or more candidate corrected messages;
d) wherein the output from the AI model further comprises, for each candidate corrected message, a meaning or intent of the candidate corrected message.
16. The non-transitory computer-readable media of claim 15 , wherein the output from the AI model further comprises, for each candidate corrected message, a likelihood that the candidate corrected message is correct.
17. The non-transitory computer-readable media of claim 15 , the method further comprising:
a) providing, as further input to the AI model, a context or current activity of the wireless receiver; and
b) determining, as further output from the AI model, for each candidate corrected message, a likelihood, based at least in part on the context or current activity of the wireless receiver, that the candidate corrected message is correct.
18. The non-transitory computer-readable media of claim 15 , the method further comprising
a) providing, as further input to the AI model, a context or current activity of the wireless receiver; and
b) determining, as further output from the AI model, for each candidate corrected message, a meaning or intent of the candidate corrected message, based at least in part on the context or current activity of the wireless receiver.