Wearable device-aware supervised power management for mobile platforms
View Patent ↗Methods, systems, and computer program products are provided for supervised power management between a primary platform and a secondary platform. Communication between a primary platform and a secondary platform is established. An application running on the secondary platform is captured. Input features and output measures are collected to build a training set for the application, wherein the input features are collected through direct measurement and the output measures reflect characteristics of the application. Based on the training set, power consumption of the secondary platform with an expected performance level is predicted for a new application running on the secondary platform. Accordingly, an optimal power management policy is derived that minimizes the total power consumption of the primary and secondary platforms.
1. A power management system on a primary platform, comprising:
a memory;
a processor coupled to the memory, the processor being configured, based on instructions stored in the memory, to:
establish communication between the primary platform and a secondary platform;
capture an application running on the secondary platform;
collect an input feature of the application and an output measure of the application reflecting a characteristic of the application as a training set for the application, wherein the input feature includes a utilization rate, wherein the utilization rate is a ratio of an active period over a total period, and wherein the total period is a sum of the active period and an idle period, and wherein the input feature and the output measure are categorized into a low, medium, or high category;
predict power consumption of the secondary platform with an expected performance level for a second application running on the secondary platform based on the training set; and
adjust a power management policy with a clock frequency of the secondary platform that reduces a total power consumption of the primary platform and the secondary platform.
2. The system of claim 1 , wherein the processor is further configured to build the training set based on a machine learning technique.
3. The system of claim 1 , wherein the primary platform comprises a mobile platform.
4. The system of claim 1 , wherein the secondary platform comprises a wearable mobile device communicating with the primary platform.
5. The system of claim 1 , wherein the output measure comprises the power consumption of the secondary platform or a data rate of the secondary platform.
6. The system of claim 1 , wherein the power consumption of the secondary platform is proportional to a data rate of the secondary platform.
7. The system of claim 1 , wherein the processor is further configured to:
categorize a feature y of the secondary platform into a class 1 based on:
y
MAP
=
arg
max
l
Prob
(
y
=
l
❘
x
1
,
x
2
,
…
,
x
n
)
=
arg
max
l
Prob
(
x
1
,
x
2
,
…
,
x
n
❘
y
=
l
)
·
Prob
(
y
=
l
)
Prob
(
x
1
,
x
2
,
…
,
x
n
)
=
arg
max
l
Prob
(
y
=
l
)
·
∏
j
=
1
l
prob
(
x
j
❘
y
=
l
)
wherein x 1 is a utilization rate, x 2 is a frame frequency of the secondary platform, and y map is a maximal probability of assigning the class 1 to the feature y.
8. The system of claim 1 , wherein the power management policy is adjusted based on:
DFS
new
=
arg
min
freq
1
,
freq
2
(
Energy
primary
+
Energy
secondary
)
=
arg
min
freq
1
,
freq
2
[
(
p
-
·
t
p
+
p
-
·
t
p
2
)
+
(
func
(
d
-
)
·
t
s
+
func
(
d
-
)
·
t
s
2
)
]
such that FPS secondary >user_defined
wherein freq1 is a clock frequency of the primary platform; freq2 is a clock frequency of the secondary platform; t p is an active period of the primary platform; t s is an active period of the secondary platform, assuming prediction of a power consumption range Px=[p−, p+] and a data rate range Dy=[d−, d+] and the power consumption of the secondary platform is estimated as P_2 nd =[func(d−), func(d+)].
9. A non-transitory computer-readable storage device having instructions stored thereon, execution of which by a computing device, causes the computing device to perform operations, the operations comprising:
establishing communication between a primary platform and a secondary platform;
capturing an application running on the secondary platform;
collecting an input feature and an output measure reflecting a characteristic of the application as a training set for the application, wherein the input feature includes a utilization rate, wherein the utilization rate is a ratio of an active period over a total period, and wherein the total period is a sum of the active period and an idle period, and wherein the input feature and the output measure are categorized into a low, medium, or high category;
predicting power consumption of the secondary platform with an expected performance level for a second application running on the secondary platform based on the training set; and
adjusting a power management policy with a clock frequency of the secondary platform that reduces a total power consumption of the primary platform and the secondary platform.
10. The non-transitory computer-readable storage device of claim 9 , the operations further comprising:
building the training set based on a machine learning technique.
11. The non-transitory computer-readable storage device of claim 9 , wherein the primary platform comprises a mobile platform.
12. The non-transitory computer-readable storage device of claim 9 , wherein the secondary platform comprises a wearable mobile device communicating with the primary platform.
13. The non-transitory computer-readable storage device of claim 9 , wherein the output measure comprises power consumption of the primary platform or a data rate of the secondary platform.
14. The non-transitory computer-readable storage device of claim 9 , wherein the power consumption of the secondary platform is proportional to a data rate of the secondary platform.
15. A method for supervised power management, comprising:
establishing communication between a primary platform and a secondary platform;
capturing an application running on the secondary platform;
collecting an input feature and an output measure as a training set for the application, wherein the output measure reflects a characteristic of the application, wherein the input feature includes a utilization rate, wherein the utilization rate is a ratio of an active period over a total period, and wherein the total period is a sum of the active period and an idle period, and wherein the input feature and the output measure are categorized into a low, medium, or high category;
predicting power consumption of the secondary platform with an expected performance level for a new application running on the secondary platform based on the training set; and
adjusting a power management policy with a clock frequency of the secondary platform that reduces a total power consumption of the primary platform and the secondary platform.
16. The system of claim 1 , wherein the processor is further configured to:
adjust Dynamic Voltage Frequency Scaling (DVFS) parameters for the second application to reduce the total power consumption of the primary platform and the secondary platform.
17. The non-transitory computer-readable storage device of claim 9 , the operations further comprising:
adjusting Dynamic Voltage Frequency Scaling (DVFS) parameters for the second application to reduce the total power consumption of the primary platform and the secondary platform.
18. The method of claim 15 , further comprising:
adjusting Dynamic Voltage Frequency Scaling (DVFS) parameters for the second application to reduce the total power consumption of the primary platform and the secondary platform.
19. The system of claim 1 , wherein the output measure comprises a data rate between the primary platform and the secondary platform.
20. The non-transitory computer-readable storage device of claim 9 , wherein the output measure comprises a data rate between the primary platform and the secondary platform.
21. The system of claim 1 , wherein the input feature includes a frame frequency categorized into a plurality of categories.