Device and method for updating power schedule

1 . A device for wireless communications, comprising:

at least one memory comprising instructions;

a wireless communications chipset; and

at least one processor coupled to wireless communications chipset and the at least one memory and configured to:

receive an indication of one or more parameters for input to a reinforcement learning machine learning (ML) model;

determine a power schedule for the wireless communications chipset based on the one or more parameters using the reinforcement learning ML model;

measure a value for a performance metric based on a human perceptible delay or human perceptible performance change resulting in image quality degradation;

update the power schedule based on the value, wherein, to update the power schedule, the at least one processor is configured to:

determine an amount of power used based on the power schedule;

determine a penalty value based on the amount of power used and the performance metric; and

compare the determined penalty value to a stored penalty value associated with the power schedule to determine whether the determined penalty value diverges from the stored penalty value; and

determine to switch the wireless communications chipset into a low power state or a higher power state based on the updated power schedule.

2 . The device of claim 1 , wherein the one or more parameters include at least one of a primary frequency of operations, a latency budget, or a power budget.

3 . The device of claim 2 , wherein the indication of the primary frequency of operations is based on a rate at which images are received for display.

4 . The device of claim 2 , wherein the latency budget is based on a maximum allowed time for motion to be displayed.

5 . The device of claim 1 , wherein the performance metric is based on an amount of time used by the device to receive an input and output an image for display based on the received input.

6 . The device of claim 1 , wherein the reinforcement learning ML model includes a set of predetermined penalty scores.

7 . The device of claim 6 , wherein the set of predetermined penalty scores are determined based on a training process.

8 . The device of claim 7 , wherein the power schedule comprises a target wake time (TWT), and wherein the set of predetermined penalty scores are predetermined for a range of TWT service period start offsets and a range of lengths of time for the higher power state, and wherein the training process determines penalty scores across the range of TWT service period start offsets and range of lengths of time.

9 . The device of claim 1 , wherein the at least one processor is further configured to:

transmit an indication to block requests for the wireless communications chipset while the wireless communications chipset is in the low power state; and

transmit an indication to unblock requests for the wireless communications chipset while the wireless communications chipset is in the higher power state.

10 . The device of claim 9 , wherein the indication to block requests and the indication to unblock requests are transmitted to a driver for the wireless communications chipset.

11 . The device of claim 1 , wherein a driver for the wireless communications chipset is configured to block requests for the wireless communications chipset and unblock requests for the wireless communications chipset based on the power schedule.

12 . The device of claim 1 , wherein the at least one processor is further configured to apply an adjustable balancing factor to the performance metric.

13 . A method for wireless communications, comprising:

receiving an indication of one or more parameters for input to a reinforcement learning machine learning (ML) model;

determining a power schedule for a wireless communications chipset of a device based on the one or more parameters using the reinforcement learning ML model;

measuring, by the device, a value for a performance metric based on a human perceptible delay or human perceptible performance change resulting in image quality degradation;

updating the power schedule based on the value, wherein updating the power schedule comprises:

determining an amount of power used based on the power schedule;

determining a penalty value based on the amount of power used and the performance metric; and

comparing the determined penalty value to a stored penalty value associated with the power schedule to determine whether the determined penalty value diverges from the stored penalty value; and

determining to switch the wireless communications chipset into a low power state or a higher power state based on the updated power schedule.

14 . The method of claim 13 , wherein the one or more parameters include at least one of a primary frequency of operations, a latency budget, or a power budget.

15 . The method of claim 14 , wherein the indication of the primary frequency of operations is based on a rate at which images are received for display.

16 . The method of claim 14 , wherein the latency budget is based on a maximum allowed time for motion to be displayed.

17 . The method of claim 13 , wherein the performance metric is based on an amount of time used by a device to receive an input and output an image for display based on the received input.

18 . The method of claim 13 , wherein the reinforcement learning ML model includes a set of predetermined penalty scores.

19 . The method of claim 13 , further comprising:

transmitting an indication to block requests for the wireless communications chipset while the wireless communications chipset is in the low power state; and

transmitting an indication to unblock requests for the wireless communications chipset while the wireless communications chipset is in the higher power state.