AI-based algorithm for optimizing modulation in 5G/6G
Artificial Intelligence (AI) means are disclosed for enabling network operators to optimize 5G and 6G messaging performance, in real-time. AI models, or fieldable algorithms derived therefrom, can select an appropriate modulation scheme according to network conditions. Modulation variables can then be adjusted to optimize performance, such as throughput or failure rates, for low or high traffic densities. Three development phases are described: network data acquisition including faults experienced under various network conditions, AI structure tuning for accurate prediction of performance, and implementation of a fieldable algorithm based on the AI structure. Network operators can use the fieldable algorithm to compare predicted performance metrics in real-time, according to various operating conditions (such as available modulation schemes), and thereby adjust particular modulation parameters (such as amplitude or phase levels).
1. A computer comprising a memory, the memory containing an artificial intelligence (AI) mathematical structure comprising:
a. a plurality of input parameters comprising one or more wireless message fault rates and one or more parameters of a modulation table;
b. a plurality of intermediate functions configured to perform calculations based at least in part on the input parameters; and
c. one or more output parameters functionally related to the intermediate functions, the output parameters comprising a predicted network performance parameter;
d. wherein the modulation table comprises integer Namp amplitude levels and integer Nphase phase levels, Namp being different from Nphase;
e. wherein the wireless message fault rates comprise at least one of:
i. an adjacent-amplitude fault wherein a faulted message symbol and a corresponding unfaulted message symbol differ by one amplitude level;
ii. an adjacent-phase fault wherein a faulted message symbol and a corresponding unfaulted message symbol differ by one phase level;
iii. a non-adjacent fault wherein a faulted message symbol and a corresponding unfaulted message symbol differ by more than one amplitude level or phase level.
2. An algorithm, derived from an artificial intelligence model, the algorithm configured to:
a. take, as input, a particular modulation table of a set of modulation tables, and a current network performance parameter comprising at least one of:
i. a current traffic density;
ii. a current message throughput
iii. a current message failure rate;
iv. a current average delay per message; and
b. provide, as output, at least one of:
i. a predicted traffic density;
ii. a predicted message throughput
iii. a predicted message failure rate;
iv. a predicted average delay per message; and
v. combinations thereof;
c. wherein the particular modulation table comprises Namp amplitude modulation levels and Nphase phase modulation levels, and the current message failure rate comprises at least one of:
i. a rate of adj acent-amplitude faults, each adj acent-amplitude fault comprising a message symbol altered from an initial amplitude level to an adjacent amplitude level;
ii. a rate of adjacent-phase faults, each adjacent-phase fault comprising a message symbol altered from an initial phase level to an adjacent phase level; or
iii. a rate of non-adjacent faults, each non-adjacent fault comprising a message symbol altered by at least two phase or amplitude levels relative to initial phase and amplitude levels.
3. The algorithm of claim 2 , further configured to:
a. calculate a predicted network performance parameter according to each modulation table of the set of modulation tables; and
b. provide, as further output, a suggested modulation table, of the set of modulation tables, the suggested modulation table having the largest predicted network performance parameter.
4. An algorithm derived from an artificial intelligence model for selecting a selected modulation table, the algorithm comprising:
a. one or more input parameters comprising at least a current message fault rate; and
b. one or more output parameters comprising at least a predicted network performance parameter;
c. wherein the algorithm is in non-transitory computer-readable memory of a base station or a core network of a wireless network; and
d. wherein the algorithm is configured for use by a processor in the base station or core network of the wireless network;
e. wherein the current message fault rate is determined by comparing a measured amplitude or phase modulation value of a message symbol, to a predetermined amplitude or phase level of a current modulation table, the current modulation table comprising modulation states suitable for modulating symbols of a wireless message.
5. An algorithm derived from an artificial intelligence model for selecting a selected modulation table, the algorithm comprising:
a. one or more input parameters comprising at least a current message fault rate; and
b. one or more output parameters comprising at least a predicted network performance parameter;
c. wherein the algorithm is in non-transitory computer-readable memory of a base station or a core network of a wireless network: and
d. wherein the algorithm is configured for use by a processor in the base station or core network of the wireless network;
e. wherein the current message fault rate is a number per unit time of messages which, after being demodulated at a receiver, differ in at least one symbol from the corresponding messages as modulated at a transmitter.
6. The algorithm of claim 5 , wherein the input parameters further include a rate of adjacent-amplitude faults, each adjacent-amplitude fault comprising a message symbol modulated according to a first amplitude level and demodulated according to a second amplitude level, the first and second amplitude levels being adjacent levels in a current modulation table.
7. The algorithm of claim 6 , wherein the input parameters further include a rate of adjacent-phase faults, each adjacent-phase fault comprising a message symbol modulated according to a first phase level and demodulated according to a second phase level adjacent to the first phase level in the current modulation table.
8. The algorithm of claim 6 , wherein the input parameters further include a rate of non-adjacent faults, each non-adjacent fault comprising a message symbol wherein the demodulated value of the symbol differs from the modulated value of the symbol by at least two amplitude levels or at least two phase levels of the current modulation table.
9. An algorithm derived from an artificial intelligence model for selecting a selected modulation table, the algorithm comprising:
a. one or more input parameters comprising at least a current message fault rate; and
b. one or more output parameters comprising at least a predicted network performance parameter;
c. wherein the algorithm is in non-transitory computer-readable memory of a base station or a core network of a wireless network; and
d. wherein the algorithm is configured for use by a processor in the base station or core network of the wireless network;
e. wherein the algorithm is further configured to select, as the selected modulation table, a particular member of the set of available modulation tables, wherein the selected modulation table is predicted, by the algorithm, to provide at least one of:
i. a predicted message throughput higher than a current message throughput;
ii. a predicted message failure rate lower than a current message failure rate; and
iii. a predicted average delay time per message lower than a current message delay time per message;
f. wherein the current message throughput, the current message failure rate, and the current delay time per message are values measured while a current modulation table is in use.