1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
|
This is an excerpt of the Project Report
"Power Management For Laptops Running Linux:
Using Linear Regression To Predict Battery Usage"
of Mitrabhanu Mohanty at Texas A&M University.
He kindly gave me his permission to publish the part of his report covering
the theory behind ibam. In his project he also developed another battery
monitor, called monx, which uses a different and also promising battery
model. I would like to thank Mitrabhanu Mohanty for the interesting
discussions we had and of course for mentioning ibam in the chapter about
"Related Work".
Sebastian Ritterbusch.
--------------------------------------------------------------------------
[...]
2. Related Work
Although there are several APM based battery monitoring utilities available
[2, 3, 4, 5], most of them simply read the estimates provided by the apmd
daemon and display it. Only IBAM uses statistical modeling techniques to
improve the accuracy of the values provided by the apmd daemon.
The basic idea behind IBAM is to record the time it takes for the charge
percentage to move from one value to the other. From this, it calculates the
delta changes and the remaining time estimate is determined by the summation
of the delta changes. To be a little bit more specific, the underlying
assumption in IBAM is that the time needed for a percentage step (T_c) is a
continuous random variable with mean t_c and standard deviation d_c. It then
uses the fact that the mean of all measured data for a percentage step
change (say from 70% to 69%) is converging to the average of the normal
distribution (i.e. T_70). Thus, to determine the time taken for the charge
percentages to drop from say 70% to 65%, IBAM uses the random variable T =
T_70 + T_69 + T_68 + T_67 + T_66 to determine the mean time needed to go
from 70% charged to 65% charged.
The accuracy of the estimates provided by IBAM depends on several factors.
It assumes that the power supplied remains more or less constant with only
some random fluctuations in the BIOS measurements. These days almost every
laptop has power saving features available, and this introduces
complications with the above assumption. In order to counter this, IBAM uses
an additional linear model to improve the accuracy.
The accuracy of IBAM improves as more data is collected from charging and
discharging cycles of the battery. It was found that although the initial
predictions of IBAM were almost the same as that provided by the BIOS, they
improved considerably after a few charging/discharging cycles.
[...]
8. References
[...]
[1] IBAM, S. Ritterbusch, Available at
http://sourceforge.net/projects/ibam/, August 2001.
[2] KBatt, J. Orosz, Available at
http://sourceforge.net/projects/kbatt/, October 2001.
[3] Battstat, J. Pehrson, Available at
http://freshmeat.net/projects/battstatapplet/, July 2000.
[4] wmapm, M.G. Henderson, Available at
http://nis-www.lanl.gov/~mgh/, January 2000.
[5] asapm, Tigr, Available at
http://freshmeat.net/projects/asapm/, July 1998.
[...]
|
