KR100753469B1 - Power management method based on chemical characteristics of portable battery - Google Patents

Abstract

A power management method based on chemical characteristics of a portable battery is provided to extend a lifespan of the battery by applying an optimized DVS(Dynamic Voltage Scaling) scheme after dividing a state of the battery into a recovery region, a power maintenance region, and an alarm region. A power management system(210) monitors a state of a processor, collects information required for a DVS(Dynamic Voltage Scaling) process, and determines a proper strategy for the processor. The power management system includes a monitoring module(211), a policy module(214), and a processor control module(218). A battery remaining quantity which is monitored on a smart battery(201) is inputted to the monitoring module. The remaining quantity is analyzed in a battery state monitor(212). A current task quantity is measured by a task monitor(213) and the measured result is inputted to the policy module. The policy module determines a proper processor state by using a selective DVS policy(216). A current state of the processor(221) is inputted to an algorithm selector(215) through an API(Application Program Interfaces)(219).

Description

Power management method based on chemical characteristics of battery {Power management method based on chemical characteristics of portable battery}

1 is a diagram showing the power consumption characteristics of a lithium-ion battery over time as data underlying the present invention;

2 is a view showing generalized power consumption characteristics of the lithium ion battery shown in FIG. 1;

3 is a diagram illustrating a combination of a workload operating in a kernel and a battery consumption in the present invention, and illustrating a change in battery instantaneous consumption through DVS,

4 is a diagram showing sudo code for actually programming a kernel structure proposed by the present invention;

5 is a block diagram showing a basic hardware configuration to which the power management method according to the present invention is applied;

6 is a block diagram of a low power kernel structure using lithium ion battery characteristics according to the present invention;

7 to 11 are diagrams showing experimental results of a power management method according to the present invention;

<Explanation of symbols for main parts of the drawings>

201: smart battery 210: power management system

211: monitoring module 212: battery status monitor

213: Task Monitor 214: Policy Modules

215: algorithm selector 216: optional DVS policy

217: API control function unit 218: processor control module

219: Processor Status API 220: Processor Control API

221: processor

The present invention relates to a power management method of a mobile device using the chemical characteristics of the battery, and based on the chemical characteristics according to the remaining battery capacity of the battery recovery period, the three stages of the charge recovery effect area, power consumption ratio maintenance effect area, warning area The present invention relates to a power management method of a mobile device using the chemical characteristics of a battery to extend the battery life using the DVS technique optimized for each section.

Recently, as the use of mobile devices such as PDA, smart phone, HPC (Handheld PC), notebook, portable multimedia player (PMP), DMB, wireless sensor node, etc. increases, the performance of the mobile device and battery life This is recognized as an important factor. Therefore, a lot of research is being done in terms of hardware and software to extend the battery life.

On the hardware side, active researches are being conducted in terms of saving energy consumption of devices. On the software side, techniques for reducing power consumption by efficiently operating low power designed hardware are being studied. A representative technique on the software side is known as Dynamic Voltage Scaling (DVS), which changes the operating frequency and supply voltage of the processor. Processor DVS is considered to be one of the most efficient ways to reduce power consumption because of its high power consumption and continual power consumption. Processors supporting DVS must dynamically adjust the operating frequency of the processor to meet the required performance. Therefore, it is necessary to study algorithms in terms of an operating system that manages mobile device resources.

The existing DVS algorithm has been performed assuming that a battery, a power supply, can continuously supply a certain amount of power, and that the use time of the battery is only related to the power consumption. From this point of view, the minimum operating frequency was determined according to the processor usage and the supply voltage was determined accordingly. Representative examples of this method include Wiser and Govil's method of predicting future processor usage using an interval prediction based method and determining the CPU supply voltage and operating frequency according to the predicted usage. Along with power consumption, an important consideration in DVS is the QoS of the task. Wiser and Govil's method determines the state of the processor using only the expected operating frequency, so it is difficult to guarantee proper QoS in each task.

Therefore, Pering and Shin developed DVS algorithm that does not degrade QoS by performing DVS in consideration of task deadline, but it is not suitable for general-purpose operating system. To improve this, Flautner and Lorch recommend DVS mechanism and algorithm suitable for general-purpose operating system. Suggested. However, contrary to the assumptions of the previous studies, since batteries have different chemical characteristics depending on the remaining battery power, the power supply cannot be constantly maintained, and the use time of the battery is changed according to the power consumption and idle time.

Therefore, considering the characteristics of the battery in the DVS policy is effective in extending the battery life. The DVS policy considering batteries was presented by Luo, Martin and Rakhmatov. Luo proposed a way to extend battery life by optimizing the battery's power consumption and avoiding peak power, and Martin proposed how to change memory bandwidth and processor speed based on battery characteristics. Policy to extend battery life. However, the proposed techniques were simulation results based on battery modeling, and because of the complexity of the model, it is difficult to use the online policy because the state of the battery cannot be monitored in real time. Rakhmatov also proposed a method to increase battery life by calculating power consumption for real-time tasks and arranging tasks within the range that guarantees deadlines according to the consumption and inserting idle time. Rakhmatov presented a model that can monitor the state of the battery in real time and derived the results using the model. However, since the task itself is a real-time task with deadline and energy consumption, it is difficult to apply to a general-purpose operating system.

In addition to the battery-considered DVS policy, Benini also proposed a DPM policy, where Benini proposed a method to increase battery life by monitoring battery status in real time and adjusting QoS according to battery level. The MP3 player adds hardware to measure the battery and monitors the battery's condition in real time to implement the policy, but this method extends life by simply reducing QoS through the optimal value between QoS and battery consumption. This method helps to prolong the battery life but has the disadvantage of reducing QoS.

Therefore, the present invention was invented to solve the above problems, and divided the state section of the battery into three stages: a charge recovery effect region, a power consumption ratio maintenance effect region, and a warning region based on the chemical characteristics according to the remaining battery capacity. After that, the purpose of the present invention is to provide a power management method of a mobile device using the chemical characteristics of a battery to extend the battery life using the DVS technique optimized for each section.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

The present invention provides a power management method of a mobile device that is powered from a smart battery,

(a) monitoring, by the monitoring module of the mobile device, current battery level information from the smart battery, and monitoring and transmitting task information currently performed in the kernel and scheduler of the operating system to a policy module;

(b) the policy module uses the information received from the monitoring module and the current processor state information input from the processor control module to maintain the current state of the battery characteristics according to the remaining battery charge recovery area and power consumption ratio. Determining one of an area and a warning area, and then selecting and applying a DVS algorithm determined according to each area to determine a processor state of a next step;

(c) receiving a processor state determined by the policy module from the processor control module to control the state of the processor, and feeding back the current processor state information to the policy module by the processor controller module;

Characterized in that comprises a.

As a preferred embodiment, in the step (b), the step of determining the current battery state to any one of the charge recovery effect area, the power consumption ratio maintenance effect area and the warning area,

The value of the change amount ΔB (Q _i ) (= B (Q _{i -1} )-B (Q _i )) of the battery consumption B (Q _i ) of the task calculated from the received information and defined by the following formula (A): It is characterized by judging from.

Formula (A):

B (Q _i ): How much battery is consumed during the Q _i time the task is performed,

L _max : Remaining charge when the battery is fully charged,

L _i : Current battery level.

Herein, when the value of ΔB (Q _i ) is maintained as a positive value, it is determined as a charge recovery effect region, and when the value of ΔB (Q _i ) is equal to or greater than a predetermined negative value, the power consumption ratio maintaining effect region is determined. If it is smaller than a predetermined negative value, it is determined that the warning area.

In the step (b), when the current battery characteristic state is determined as the charge recovery effect region, the task is performed by setting the power state P _i of the processor to the power state P _max at which the processor performance is the highest (P _i = P _max ). The processor state of the next step is determined as much as possible.

In the step (b), when the current battery characteristic state is determined as the power consumption ratio maintenance effect region, the task is performed by setting the power state P _i of the processor as the power state P _{optimal (i)} defined by Equation (B) below. It is characterized by determining the processor state of the next step to be performed.

Formula (B):

W _{real (i-1)} : execution time of the i-1th task obtained by the task monitor,

Q _{real (i-1)} : period of the i-1th task obtained by the task monitor,

W _{pred (i)} : execution time of the i th task predicted by the previously obtained value,

Q _{real (i)} : Period of the i th task predicted by the previously obtained value.

In the step (b), when the current battery characteristic state is determined as the charge recovery effect region, the task is performed by setting the power state P _i of the processor to the power state P _max at which the processor performance is the highest (P _i = P _max ). The processor state of the next step is determined as much as possible.

Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings.

The present invention relates to a power management method of a mobile device using a chemical characteristic of a battery, and monitors the battery level using a smart battery standard and monitors the task online in the operating system online, and then a power management method suitable for the battery level, that is, Provides an optional technique for applying the DVS technique.

Conventional low-power techniques have been performed assuming that the battery, the power source of the mobile device, maintains a constant power supply, but as is known, the battery exhibits different chemical characteristics depending on the remaining amount, which is effective for prolonging the battery life. Can not do it.

Therefore, the present invention proposes a DVS technique based on battery characteristics not considered in the conventional DVS technique, and based on the chemical characteristics according to the remaining battery capacity, the state section of the battery is based on the charge recovery effect region, power consumption ratio maintenance effect region, The main focus is to extend the battery life using the DVS technique optimized for each section after dividing into three levels of warning area.

Batteries vary their power consumption by various factors. The most influential characteristics are the power consumption rate maintenance rate and the recovery effect. The effect of maintaining the power consumption ratio means that the battery usage time changes according to the discharge amount of the battery. For example, a battery with 10 hours of battery life discharged at 1A increases the discharge time to 2A, which decreases the battery life time at a rate greater than the rate at which the amount of discharge increases. . A battery is a device that converts chemical energy of an active material into electrical energy through an electrochemical oxidation / reduction reaction. That is, an electrolyte reacts with an anode and a cathode, which are electrodes, and electrical energy is generated by the movement of electrons. At this time, the reaction occurs from the electrolyte close to the electrode, and the electrolyte far from the electrode replenishes by transferring electrons to the already reacted electrolyte. Therefore, when the consumption of the battery at the same time increases, the electrolyte close to the electrode decreases rapidly, and the electron transfer rate is slowed. This is why the power consumption rate maintenance effect occurs.

In addition, when an idle time occurs while the battery is consumed, a charge recovery effect occurs because the electrolyte far from the electrode can stably transfer electrons toward the close electrolyte. Considering the battery's remaining capacity and chemical characteristics, first, in the case of the charge recovery effect, since the electrolyte is sufficiently present when the battery level is large, the charging recovery effect may occur at a short time even if the idle time is short, but the battery level decreases. If not enough idle time is given, there is no sufficient recovery effect due to lack of electrolyte. Therefore, in the case of the charge recovery effect, the supply of sufficient electrons can be made in the electrolyte only when the battery is consumed at a slow rate as the remaining amount of the battery decreases.

Considering such a chemical model, if the battery consumption and idle time is repeated, it is expected that the charging recovery effect and the consumption rate maintenance effect may be differently affected according to the remaining battery capacity. Was run. For the experiment, I installed Familar Linux v7.2 and GPE on iPAQ 3600 PDA, and modified the kernel to read the result obtained by voltage drop in hardware to measure the remaining battery. In the experiment, the task performance and idle time of the system were repeated for 30 seconds, and then the remaining battery was monitored. The reason for repeating every 30 seconds was that the battery could be fully charged and maintained its power consumption ratio within the range that minimized the experiment time.

1 shows an experimental result. The x-axis shows the battery usage time (s), and the y-axis shows the remaining battery capacity (mV) to show the change in remaining capacity over time. As can be seen in Figure 1, the amount of consumption and recovery is almost constant in the interval (region A) until the remaining battery reaches about 4150mV, it can be seen that the battery remaining between about 4200 ~ 4300mV. However, when the remaining capacity falls to the section below 4150mV (Area B), the battery level decreases at a constant rate. The remaining battery capacity, which decreased at a constant rate, decreases while drawing a sharp curve when it reaches an area of about 3650mV or less (area C). The phenomenon that the consumption pattern is different depending on the remaining battery charge is due to the battery's chemical properties. In the region A, the battery consumption is restored by the characteristics of the charge recovery effect of the battery, and in the case of the region B, the power consumption is more important because the charge recovery effect is not significantly affected. In the case of region C, the consumption curve of the battery drops drastically regardless of the charging recovery effect.

Based on the experimental results, the present invention proposes a DVS technique that maximizes the execution time of a mobile device by selectively applying a DVS algorithm according to the remaining battery capacity in consideration of battery characteristics.

The basic concept is as follows. As shown in FIG. 2, the algorithm application interval is monitored through a real-time battery monitoring according to the discharge characteristics of the battery, a recovery effect region, a power consumption ratio maintenance region, and a warning region ( Divide into three battery status sections of the alarm region and apply the appropriate DVS algorithm to each section. If the remaining charge when the battery is fully charged is L _full , and the remaining cut-off is L _low , the battery is discharged from L _full to L _low , and this is the battery life. At this time, if the battery remaining capacity until the effect of the charge recovery effect on the battery level is lower than the power consumption ratio maintenance effect is L _R , the DVS algorithm using the charge recovery effect is applied until the battery level is lowered from L _full to L _R. Can be applied. In the power consumption ratio maintenance effect region, the remaining amount of battery is an interval from L _R to L _C , and the consumption curve of the battery is lowered at a constant ratio as shown in FIG. 2. In this section, DVS algorithm should be used to minimize battery consumption, keep it constant, and guarantee QoS. L _C and L _R can be obtained experimentally by a battery monitor. The battery consumption curve drops sharply in the warning area at the point where the remaining battery level falls below L _C and reaches L _low . In this case, it is possible to maintain the maximum battery life by applying an algorithm that can maximize the battery life by ignoring QoS.

Hereinafter, a specific power management method of the present invention will be described so that the above-described basic concepts can be applied to a task of a general-purpose operating system. First, the current battery state is applied to a charge recovery effect area, a power consumption ratio maintenance effect area, and a warning area. The process of determining by any of the following is as follows.

The basic task model is described as follows. In general, tasks are divided into interactive tasks and periodic tasks. In addition, the periodic tasks require continuous battery consumption and can affect QoS when DVS is used. In the present invention, it is possible to prolong the battery life by selectively applying a suitable DVS algorithm for each of the battery status intervals which are dynamically obtained while guaranteeing the QoS of the periodic task.

3 shows a task model having periodicity. Assuming that the processor has n power states, the current power state that the processor is performing is defined as P _i , and the deadline time at which the task should be completed is Q _i . In addition, if the power state at which the performance of the processor is the highest is defined as P _max , and the power state is defined as P _min at the lowest performance, the task is P _max and P _min. At any time in between, the task execution time W _i depends on P _i . Idle time I _i is Q _i Is defined as W _i . The values of Q _i , W _i and I _i are obtained by the task monitor, respectively. This relationship is expressed by the following equation (1).

(1)

(One)

In order to perform the selective DVS technique, a battery interval must first be set. In the present invention, three battery sections are presented as a charge recovery effect region, a power consumption ratio maintenance region, and a warning region according to the discharge characteristics of the battery. The battery interval should reflect the change in the battery consumption pattern of the task according to the remaining amount, and can be obtained by monitoring the change in battery amount used by the task. Specifically, a method of obtaining a battery consumption amount is as follows. First, the amount of battery consumed during Q _i time is called B (Q _i ), and the remaining amount when the battery is fully charged is L _max and the current remaining amount is indicated by L _i . L _max and L _i can be obtained through the battery monitor, and when B (Q _i ) is obtained through this, Equation (2) is obtained.

(2)

(2)

The relationship between the characteristics of the battery and W _i , I _i can be seen in FIG. 3. That is, W _i is related to the power consumption ratio maintenance effect, and I _i is related to the charging recovery effect. Therefore, when the characteristics of the battery are displayed as the battery consumption, it may be displayed as B (W _i ) and B (I _i ), respectively. In the case of B (W _i ), since the battery is consumed by the task, it is always positive, and the ratio of power consumption varies depending on the length of W _i . On the other hand, in the case of B (I _i ), negative numbers appear as the battery consumption decreases and the charge recovery effect occurs. Based on this relationship and the equations (1) and (2), the current battery consumption B (Q _i ) of the task can be expressed as in the following equation (3).

(3)

(3)

The current battery state is determined according to the value of B (Q _i ). In other words, when the value of B (Q _i ) according to Equation (3) is negative, the charge recovery effect priority area becomes a positive value. In addition, if the value of B (Q _i ) according to Equation (3) is positive and larger than a value in the power consumption ratio maintaining effect priority region by a predetermined level or more, that is, larger than a predetermined amount, it is determined as a warning region. In the priority area, there is a certain amount of battery consumption, but in the warning area, the change in B (Q _i ) value is much larger than in the power consumption rate maintenance effect area.

In real systems, B (W _i) and B (I _i) variation _{ΔB (Q i) = B (} Q i -1) of the B (Q _i) in order to separate the battery section in the present invention because they do not seek a - B (Q _i ) can be used. In this case, if ΔB (Q _i ) maintains a positive value, the interval may be regarded as maintaining the state (remaining amount) of the battery by the charge recovery effect, and thus the interval may be designated as the charge recovery effect priority region. . On the other hand, if ΔB (Q _i ) is turned to a negative value, the battery recovery due to the charge recovery effect does not occur and may be designated as a power consumption ratio maintenance priority region. When ΔB (Q _i ) is measured as a negative value with a predetermined level or more, the section may be designated as a warning area. That is, if the value of ΔB (Q _i ) is equal to or greater than a predetermined negative value, the power consumption ratio maintenance effect priority area is designated.

In the present invention, after the battery state is determined as described above, a power management technique, that is, a DVS technique, is selected and applied to the determined region. Hereinafter, the DVS algorithm applied to each region will be described.

As described above, when the battery sections are determined, a DVS algorithm suitable for each section should be applied. Charge recovery effect The priority area should use an algorithm that maximizes the charge recovery effect. The longer the recharge effect is, the greater the effect of I _i is. Therefore, the task should be performed by increasing I _i . To increase I _i , set the processor state with P _i = P _max in Equation (1), where W _i = W _max and I _i becomes the longest.

The task should be performed in a direction of reducing the load of the battery in the power consumption ratio maintenance effect priority region. In the present invention, Luo [J. Luo, JK Niraj, "Battery-aware static scheduling for distributed real-time embedded systems Design," Proceedings of ACM / IEEE conference on Design automation, Las vegas, Nevada, United States, pp. 444-449, December 2001.] and Flautner [TL Martin, Balancing Batteries, Power and Performance: System Issues in CPU Speed-Setting for Mobile Computing, PhD thesis, Department of Electrical and Computer Engineering] , Carnegie Mellon University. 1999.] proposes an algorithm that can guarantee QoS while reducing peak power.

The proposed algorithm is as follows. When the processor power state that can reduce peak power and maintain the maximum power consumption ratio is P _optimal , P _optimal is determined by processor usage, that is, the ratio of task execution time to the total cycle, that is, W _i / Q _i . Can be done. The ideal processor usage for P _optimal is W _i / Q _i Since = 1, we need to find P _optimal to be close to 1. The monitored task execution time and period are called W _real and Q _real , respectively, and the predicted performance values are W _pred and Q _pred , respectively. When the state of the _i- 1th monitored processor is P _i-1 , i P _{optimal (i)} of the th period can be expressed by the following equation (4).

(4)

(4)

Where W _{real (i-1)} and Q _{real (i-1)} are the execution time and period of the i-1th task measured by the task monitor, and W _{pred (i)} and Q _{pred (i)} previously The execution time and period of the i th task predicted by the measured value, which is obtained by using the latest average value of the past processor usage and the period. That is, the equation (4) predicting the processor utilization in the future using the history processor utilization and, in view of the processor usage and power status of the task W _i / Q _i It is a method to find P _optimal close to = 1. For processors that actually support DVS, performance needs to be changed in n discrete steps (4 steps: 100 MHz, 200 MHz, 300 MHz, and 400 MHz for XScale processors). In the example of the present invention, based on the XScale processor, the processor's performance change, that is, Wi / Q _i is fixed to 0.25, 0.5, 0.75, and 1, respectively, and this range is designated as the performance change threshold to change the processor's stage.

The policy in the warning area is that the consumption curve of the battery is drastically reduced, so when the power consumption is high, the remaining battery capacity is drastically reduced. In this area, if QoS is considered, the same algorithm as in power consumption ratio maintenance policy should be applied, but in view of the efficiency of battery life, it is better to prevent sudden battery consumption even if there is a decrease in QoS. Therefore, the present invention uses a policy that increases the battery life even if QoS is reduced by applying the policy proposed by Benini [L. Benini, G. Castelli, A. Macii, E. Macii, R. Scarsi, "Battery-Driven Dynamic Power Management of Portable Systems," International Symposium on System Synthesis ( ISSS ), Madrid, Spain, pp. 25-33, September 2000.]. To do this, set the power state in this zone to the minimum power state P _min .

4 shows a selective DVS scheme based on the remaining battery capacity described above.

Hereinafter, a DVS system in consideration of a battery for implementing the above-described DVS technique of the present invention into an actual system will be described.

5 shows a basic hardware configuration to which the power management method according to the present invention is applied. In order to apply the present invention, the mobile device must have a structure as shown in FIG. 5. That is, the DVS capable processor 101 is basically mounted, the programmable regulator 102 is used for the voltage supplied to the processor 101, and the lithium ion battery 104 is used as the main power supply source. In addition, the power of the lithium ion battery 104 is supplied through a DC-DC converter 103 that supports the smart battery standard (smart battery) standard, and the current battery level is measured through the DC-DC converter 103. . The components shown in FIG. 5 are those that can be followed without difficulty in most mobile devices.

6 is a configuration diagram of a low power kernel structure using lithium ion battery characteristics according to the present invention. Referring to this, a power management system according to the present invention will be described below.

The power management system (DVS system) 210 of the present invention considering a battery monitors the state of the battery, the task being performed, and the processor to efficiently collect information necessary for the DVS policy, and appropriately collect the state of the processor determined by the policy. Will be set. 6 shows a configuration of the DVS system 210 according to the present invention, which is largely divided into a monitoring module 211, a policy module 214, and a processor control module 218. It is composed. Referring to FIG. 6, the remaining amount of the battery monitored through the smart battery 201 is input through the monitoring module 211 which is a device driver in the kernel. The input remaining amount is analyzed by the battery status monitor 212, and the amount of tasks currently being performed is measured by the task monitor 213 and transmitted to the policy module 214. The policy module 214 then determines the appropriate processor state using the optional DVS policy 216. The current state of the processor 221 is input to the algorithm selector 215 via the processor state API 219 and then compared. The processor control API 220 changes the state of the processor 221.

In more detail, the monitoring module 211 collects battery 201 and task information, and includes a battery status monitor 212 and a task monitor 213. . The policy module 214 selects an appropriate DVS policy using information from the monitoring module 211, determines a processor state of a next step according to the policy, and informs the processor control module 218.

The policy module 214 is composed of an algorithm selector 215, a selective DVS policy 216, and an API control function 217. The processor control module 218 is a module that applies DVS according to performance through Dynamic Frequencey Management (DFM) and Frequencey Change Sequence (FCS), and feeds back the current processor status to the policy module 214.

The battery status monitor 212 collects information of the battery 201 from a smart battery monitor (not shown) and converts the battery 201 information into a form required for the policy module 214 [http://www.sbs-forum.org/ ]. The smart battery 201 is a standard for monitoring the state of the battery in a mobile system, and various vendors provide various products. Smart batteries are mostly used in notebooks, and are being installed in small mobile devices such as iPAQ. Wen presented a way to obtain battery information from the Proc file system by modifying APM (Advanced Power Management) in Linux to predict the battery life by monitoring the smart battery in real time on iPAQ [W. Ye, R. Wolski, C. Krintz, History-based, Online, Battery Lifetime Prediction for Embedded and Mobile Devices , University of California Santa Babara, Technical Report # 2003-17, April 2003.]. Recently, ACPI defines a standardized API for smart batteries [http://www.acpi.info/].

However, since it is difficult to obtain a variety of information provided by the smart battery with the provided API, it is necessary to define and implement a new API from the standpoint of the smart battery and the operating system. In the present invention, a variety of information required can be obtained by implementing a real-time battery monitor in the kernel through a newly defined API. In this way, the existing SPICE [S. Hageman, "Simple PSpice Models Let You Simulate Common Battery Types," Electronic Design News ( EDN ) , pp. 117-129, October 1993.] or Doyle [M. Doyle, T. Fuller, J. Newman, "Modeling of Galvanostatic Charge and Discharge of a Lithium / Polymer / Insertion Cell," Journal of the Electrochemical Society , Vol. 140, No. 6, pp. 1526-1533, June 1993.] can obtain the battery information needed by the policy module in real-time than the battery monitor using the model of.

On the other hand, the task monitor 213 is Lee [J. Lee, H. Kim, H. Cha, "A Process Monitoring Strategy towards DVS Applications," Korea Information Science Society , Seoul, Korea, October 2003. By monitoring tasks inside the kernel and scheduler, the task monitor can monitor tasks with less overhead and provide an accurate picture of processor usage and periodicity. Accurate task usage can be grasped by using the method that calculates the usage whenever the status of the task changes in the scheduler, rather than using the interval method that checks the status of the task at regular intervals to understand the task status and processor usage. Do. In addition, the scheduler tracks the information of the monitored task and the call information of system calls inside the kernel to accurately find the periodicity of the task. Using the monitored task's periodicity and processor usage, the deadline time of the task can be predicted by a general-purpose operating system, enabling accurate DVS application.

The algorithm selector 215 determines the state of the battery based on the information input through the monitoring module 211 to determine which battery section is present, and then selects an appropriate DVS policy and passes it to the optional DVS policy 216. The optional DVS policy 216 performs the algorithm selected by the algorithm selector 215 to determine the appropriate processor state according to the state of the task.

The state of the processor 221 of the next step passes the information received through the task monitor 213 and the current processor state obtained from the processor control module 218 to the DVS policy 216 selected by the algorithm selector 215, Determined by the results obtained. The determined processor state is passed to the processor control module 218 through the API control function 217. The processor control module 218 includes a processor control API 220 and a processor state API 219 capable of monitoring a current processor state. The processor control API 220 may determine a determined processor state by using an API control function unit ( 217) to appropriately control the processor state of the next step (controlling the clock of the processor appropriately).

Experimental Example

The experiment was carried out by implementing a battery monitor and task monitor on a Compaq iPAQ 5550 PDA and Sharp Zaurus SL-C860 with Linux operating system. The battery monitor uses the DS2760 standard smart battery monitor chip information for iPAQ to modify the Linux kernel's Hardware Abstration Layer (HAL) area to determine the battery level and current consumption. In the case of Zaurus, the MAX1111 AD converter provides Battery remaining information was obtained in the same manner. The task monitor has modified the kernel to collect the task execution information from the scheduler and system calls to measure the QoS, execution time, and idle time. In the case of iPAQ, the MPEG1 data having 200 kbps and 10 frames is performed through the Berkeley MPEG1 player [http://bmrc.berkeley.edu/frame/research/mpeg/mpeg_play.html] and the task monitor is used. Deadline Q _i was monitored in real time. In the case of Zaurus, MPEG4 data having 200 kbps and 20 frames was performed and monitored through MPlayer [http://www.mplayerhq.hu]. To compare the performance, Vertigo [K. Flautner, T. Mudge, Vertigo: automatic performance-setting for linux, Proceedings of the Fifth USENIX Symposium on Operating Systems Design and Implementation, 2002 pp. 105-116.] Method using the DVS technique based on recent usage (denoted 'Plain DVS'), method without DVS (denoted 'non-DVS'), and the method according to the present invention. (Marked as 'Proposed') each performed the same task. Also, to measure the QoS, the number of frames reproduced and the QoS drop were measured. Vertigo is a DVS that is actually implemented in a general-purpose operating system among the DVS without a battery, and was used as a comparison model because it consumes less power without degrading QoS.

For the contents of the experiment, first, the thresholds for domain classification were determined, and the effects of the policy presented in the present invention in each domain were examined, and finally, the experimental results in the entire domain were obtained. In order to obtain the threshold for determining the battery interval, the DVS policy is performed from L _full to L _low so that the charge recovery effect can occur. Change in battery consumption (ΔB (Q _i ) = B (Q _{i -1} )-B ( Q _i )) was obtained. As shown in FIG. 7, the power range consumed by the battery is maintained at 0 or more from the initial full charge to the point of 3810 mV (4100 mV in the Zaurus). This is because there is little change in power consumption because the remaining battery power is recovered by the charge recovery effect. However, when it falls below 3810mV (Zaurus is 4100mV), the change in power consumption falls below zero. This is because the amount of power consumed is larger than the amount of power restored. In this area, the charge recovery effect occurs, but it is lowered to 0 or less because the characteristic shown by the power consumption ratio maintenance effect is larger than that. When it reaches 3600mV (Zaurus is 3880mV) or more, it drops rapidly to 0 or less. At this time, the area can be designated as a warning area. Therefore, in this experiment, the recovery charge effect area threshold L _R was measured as 3810mV (Zaurus is 4100mV), and the warning area threshold L _C was 3600mV (Zaurus is 3880mV).

8, 9 and 10 are the results for each section in which the task of executing the MPEG player is performed by applying an algorithm suitable for each section according to the battery level according to the obtained threshold value.

FIG. 8 is a graph measuring a time for reaching the section by designating an area having a battery level of 3810 mV (Zaurus: 4100 mV) or more as a charge recovery effect section. When using the policy for the charging recovery effect according to the algorithm proposed in the present invention, it took the longest time to reach L _R. The method without DVS shows similar results, and the policy using DVS reaches L _R as soon as possible. This is because, with DVS, the charge recovery effect is less than that of tasks without DVS. In the charging recovery effect section according to the present invention, the battery life of about 18% (11% for the Zaurus) was extended than the state without the policy.

FIG. 9 shows a task execution time until a battery level drops below a cut-off voltage in an area of 3810 mV (Zaurus is 4100 mV) or less in the power consumption ratio maintenance effect period. When the algorithm of the present invention is used, the task execution time is longest, and when DVS is not used, the execution time is shortest. In the case of using DVS, the execution time was longer than in the case of not using DVS due to the effect of maintaining the power consumption ratio. However, in the case of the algorithm according to the present invention, the overall use time could be extended by appropriately performing tasks in accordance with the power consumption ratio. This extended battery life by about 12% (24% for the Zaurus) over the non-DVS method. What is unusual about the result of the power consumption ratio maintenance interval is that the results in Zaurus are better than those in the iPAQ, because the power consumption ratio maintenance interval is longer because the battery capacity of Zaurus is more than that of the iPAQ. .

FIG. 10 shows a task execution time until a battery level drops below a cut-off voltage by setting an area below 3600 mV (Zaurus is 3880 mV) as a warning period. In this area, we used a policy to abandon QoS and perform tasks with the processor's power state at its lowest. As can be seen in FIG. 10, both the policy using DVS and the policy not using consume power at a similar rate. When using DVS, power consumption increases according to tasks to guarantee QoS. In the warning area, even a small increase in power consumption can drastically increase battery consumption, so using DVS can dramatically reduce battery run time. Therefore, even if abandon QoS, performing the task with the lowest power state can extend the battery life. In the warning zone, battery life has been extended by about 10% (16% for the Zorus).

The results of FIGS. 8 to 10 are the results of setting the remaining amount corresponding to the threshold value of the section as a result for each section, and performing the task until the set remaining amount is reached. In the previous experiments, the battery time for each region was measured without considering QoS, and it was determined whether the DVS algorithm for each section was effective. Therefore, dynamic section discrimination and QoS experiments were needed. 11 and Table 1 below are the results of experiments performed by dynamically monitoring the battery to identify the battery interval and applying the appropriate DVS algorithm until the battery is fully discharged. In the experiment, each section was identified by the battery monitor in real time, and the battery monitor time for discriminating the section was set to 30 seconds according to the smart battery monitor specification. Since DVS can seriously affect the QoS of the task as well as the battery life, we also performed QoS measurements in this experiment. In order to measure the QoS of the task, the MPEG player was modified to log the number of frames, and then the average number of frames was calculated.

As shown in Table 1, by using the technique of the present invention, it is possible to obtain an extended battery life of about 13% to 21% while reducing QoS by only 0.9% to 1.6% than in a system not using DVS. In addition, 4% to 8% longer battery life can be achieved than with conventional DVS. In the experiments for each section performed in FIGS. 8, 9, and 10, the use time was extended by 10 to 24% according to each section, but in this case, each area was fixed and performed, which was somewhat better than the result in FIG. The results could come out. Since the experiment of FIG. 11 allows the DVS algorithm to be dynamically selected over the entire interval, a lower result is obtained because the algorithm may be incorrectly applied in each interval. As smart battery monitors become more accurate, this difference is expected to decrease.

Low-power research considering conventional batteries was based on battery models. Model-based methods are difficult to apply in real-time general-purpose operating systems because they are simulated in limited task situations and are difficult to use in real-time because of their complexity.

In order to solve this problem, the present invention proposes a method applicable to an actual general operating system using a battery monitor. In addition to the chemical properties of the battery, depending on the remaining amount of the characteristics that can be applied to DVS can extend the battery life more efficiently.

In the present invention, a selective DVS technique according to the battery level reflecting such characteristics is proposed. The technique proposed in the present invention is a method of using the DVS algorithm to set the battery interval according to the remaining battery capacity and to utilize the characteristics of the battery reflected in the interval. In the present invention, the battery section is divided into three sections and then applied using a suitable DVS algorithm. The result is a 4% to 8% improvement in battery time over the previous DVS without QoS degradation, and a 13% to 21% improvement in battery time over the time without DVS.

As described above, according to the power management method of the mobile device using the chemical characteristics of the battery according to the present invention, the state of the battery based on the chemical characteristics according to the remaining battery charge recovery effect area, power consumption ratio maintenance effect area In addition, by using the DVS technique optimized for each section after dividing into three stages of the warning area, the battery life can be extended.

Claims (6)

In the power management method of a mobile device that is powered from a smart battery, (a) monitoring, by the monitoring module of the mobile device, current battery level information from the smart battery, and monitoring and transmitting task information currently performed in the kernel and scheduler of the operating system to a policy module; (b) the policy module uses the information received from the monitoring module and the current processor state information input from the processor control module to maintain the current state of the battery characteristics according to the remaining battery charge recovery area and power consumption ratio. Determining one of an area and a warning area, and then selecting and applying a DVS algorithm determined according to each area to determine a processor state of a next step; (c) receiving a processor state determined by the policy module from the processor control module to control the state of the processor, and feeding back the current processor state information to the policy module by the processor controller module; Power management method of a mobile device using the chemical characteristics of the battery, characterized in that comprises a. The method according to claim 1, In the step (b), the step of determining the current battery state to any one of the charge recovery effect area, the power consumption rate maintenance effect area and the warning area, The value of the change amount ΔB (Q _i ) (= B (Q _{i -1} )-B (Q _i )) of the battery consumption B (Q _i ) of the task calculated from the received information and defined by the following formula (A): Power management method of a mobile device using the chemical characteristics of the battery, characterized in that determined from. Formula (A):

B (Q _i ): How much battery is consumed during the Q _i time the task is performed, L _max : Remaining charge when the battery is fully charged, L _i : Current battery level. The method according to claim 2, If the value of ΔB (Q _i ) is maintained at a positive value, it is determined as a charge recovery effect region. If the value of ΔB (Q _i ) is equal to or greater than a predetermined negative value, the power consumption ratio maintenance effect area is determined. If the negative value is less than the power management method of the mobile device using the chemical characteristics of the battery, characterized in that determined as a warning area. The method according to claim 1, In the step (b), when the current battery characteristic state is determined as the charge recovery effect region, the task is performed by setting the power state P _i of the processor to the power state P _max at which the processor performance is the highest (P _i = P _max ). A method for managing power of a mobile device using a chemical characteristic of a battery, characterized by determining a processor state of a next step. The method according to claim 1, In the step (b), when the current battery characteristic state is determined as the power consumption ratio maintenance effect region, the task is performed by setting the power state P _i of the processor as the power state P _{optimal (i)} defined by Equation (B) below. A method of power management of a mobile device using a chemical characteristic of a battery, characterized in that for determining the processor state of the next step to be performed. Formula (B):

W _{real (i-1)} : execution time of the i-1th task obtained by the task monitor, Q _{real (i-1)} : period of the i-1th task obtained by the task monitor, W _{pred (i)} : execution time of the i th task predicted by the previously obtained value, Q _{real (i)} : Period of the i th task predicted by the previously obtained value. The method according to claim 1, In the step (b), when the current battery characteristic state is determined as the charge recovery effect region, the task is performed by setting the power state P _i of the processor to the power state P _max at which the processor performance is the highest (P _i = P _max ). A method for managing power of a mobile device using a chemical characteristic of a battery, characterized by determining a processor state of a next step.

Images