CN101710254A - Embedded system energy consumption management method - Google Patents

Abstract

The invention provides an embedded system energy consumption management method which is especially suitable for the energy consumption management of the embedded real-time system based on battery-power and can ensure the real-time and reliability of task execution while reducing the energy consumption of the embedded system. The method comprises the following steps: a task set composed of n real-time periodic tasks is executed for the embedded system, the optimum solution of the execution frequency fi of each real-time periodic task Ti in the task set is solved to ensure the maximum reliability of the embedded system, the energy consumption meeting and limiting conditions required by all the real-time periodic tasks which satisfy timing constrains are executed; when the processor of the embedded system executes each real-time periodic task Ti, the frequency of the processor is set to the solved execution frequency fi of the real-time periodic task; and in the further technical scheme, the artificial fish-swarm algorithm is used to solve the optimal solution of the task execution frequency. The method is characterized by fast convergence rate, high non-inferior solution quality and strong robustness.

Description

The energy consumption management method of embedded system

Technical field

The present invention relates to embedded system power management technique field, be specially adapted to managing power consumption based on battery powered embedded real time system.

Background technology

In Embedded System Design, the energy consumption demand that the high-performance of system is brought is increasing, and particularly for battery powered embedded system, the contradiction between more and more higher energy requirement and the limited battery capacity is just becoming more and more outstanding.Therefore, how to make embedded system adjust self behavior, adapt to power consumption constraint and change, thereby the utilization factor that improves energy consumption has become the severe challenge that Embedded System Design person faces according to the situation that provides of current energy consumption.

Mainly be divided into low static power consumption and dynamic Low-power Technology in existing embedded system low-power consumption realization technology, the low static power consumption technology mainly is to relate to electronic devices and components from hardware point of view; Dynamically Low-power Technology is that angle from software realizes, mainly realizes optimal design to low-power consumption from operating system, compiler and three levels of built-in application program.The dynamic electric voltage adjusting (Dynamic VoltageScaling, DVS) technology provides a kind of operating system grade other managing power consumption ability, and along with the development of large scale integrated circuit, increasing modern processors has adopted the DVS technology.The DVS technology can dynamically change the working voltage and the clock frequency of processor as required, because frequency is directly proportional with voltage, the quadratic power of power consumption and voltage is directly proportional, reduce voltage and can reduce power consumption effectively, also cause simultaneously the task executions time lengthening, to such an extent as to if the task executions time is oversize above its off period, this task executions is counted out so.And, studies show that the execution frequency of processor is low more, the reliability of system is low more.Therefore in managing power consumption, take into account the power consumption limitations of system, real-time and reliability, in the existing power management technique, as publication number is the disclosed a kind of method that reduces power consumption of embedded system of Chinese invention patent Shen Qing Publication instructions of CN101515338A, the energy optimization of taking into account system only, and the power consumption limitations of the system of taking into account is not also arranged, the embedded system energy consumption management method of real-time and reliability in the prior art.

Artificial fish-swarm algorithm (AFSA) is a kind of novel optimizing algorithm that was proposed in 2002 by people such as Li Xiaolei the earliest, and the behavior that this algorithm simulation fish cluster cruises and looks for food makes colony reach optimized purpose by the cooperation of the collective between the fish.

The artificial fish of AFSA elder generation's initialization a group is searched optimum solution by iteration then.In each iteration, artificial fish mainly is to upgrade oneself by behaviors such as looking for food, bunch and knock into the back, thereby realizes optimizing, and concrete behavior is described below:

(1) foraging behavior: refer to that fish is following a kind of behavior that the many directions of food are moved about.Artificial fish X _iIn its visual field, select a state X at random _j, calculate their target function value respectively and also compare, if find Y _jCompare Y _iExcellent, X then _iTo X _jDirection move and move a step; Otherwise, X _iContinuation is at its mobile at random within sweep of the eye selection mode X _j, judge whether to satisfy the condition of advancing, make repeated attempts after trynumber time, the still dissatisfied condition of advancing, then moving at random moves a step makes X _iArrive a new state.

(2) behavior of bunching: refer to that every fish moves and avoid as far as possible a kind of optimizing behavior of overcrowding to the center of closing on the partner in the process of moving about.Artificial fish Xi searches for number of partners and the center in its visual field, if partner center state is more excellent and not too crowded, then Xi moves towards partner's center and moves a step, otherwise carries out foraging behavior.

(3) behavior of knocking into the back: refer to a kind of behavior that the optimal direction of fish in its visible range scope moves.Artificial fish X _iSearch for the partner of functional value optimum among all partners in its visual field, if optimum partner around not too crowded, X then _iMove towards this partner and to move a step, otherwise carry out foraging behavior.

(4) behavior is selected: according to the character that will deal with problems, every artificial fish is estimated current environment of living in, thereby selects a kind of suitable behavior to carry out.Look for food as first Simulation execution, bunch, the behavior of knocking into the back, the value of the objective function after the evaluation action selects optimum behavior to carry out then.

(5) bulletin board: the state of after executing iteration current state and bulletin board being preserved compares, if the state that is better than in the bulletin board then upgrades state in the bulletin board with oneself state, otherwise the state of bulletin board is constant.After the iteration of whole algorithm finished, the value of output bulletin board was the optimal value that we ask.

Summary of the invention

In view of this, in order to address the above problem, for this reason, the present invention proposes a kind of energy consumption management method of embedded system, take into account power consumption limitations, real-time and the reliability of system, when reducing the embedded system energy consumption, real-time and reliability that the assurance task is carried out.

The object of the present invention is achieved like this: the energy consumption management method of embedded system comprises the steps:

1) task-set of being made up of n real-time period task that need carry out for embedded system is found the solution each the real-time period task T in the task-set _iThe execution frequency f _iOptimum solution, make the reliability maximization of embedded system, and the energy consumption of carrying out all real-time period required by task satisfies the power consumption constraint condition, and all real-time period tasks satisfy temporal constraint;

2) processor of embedded system is being carried out real-time period task T _iThe time, be this real-time period task executions frequency f that solves in the step 1) with the frequency configuration of processor _i

Further, the maximization of the reliability of described embedded system is meant the real-time period task executions frequency f in this task-set _iSatisfy following formula:

Max∏ _i＝1 ⁿR _i(f _i)；

In the formula, $R_i\left(f_i\right)=e^{-\mathrm{\λ}\left(f_i\right)*\frac{C_i}{f_i}},$ Wherein, e is the nature index, C _iExpression real-time period task T _iUnder worst case, carry out required clock periodicity, $\mathrm{\λ}\left(f_i\right)={\mathrm{\λ}}_{0}g\left(f_i\right)={\mathrm{\λ}}_0{10}^{\frac{d(1-f_i)}{1-f_{\min }}},$ Wherein, λ ₀System's instantaneous failure rate average of correspondence during for execution frequency maximum, d is a constant, is the susceptibility of system's transient fault to processor frequencies and voltage, f _MinBe the processor frequencies minimum value after normalized;

Further, the energy consumption of all real-time period required by task of described execution satisfies the real-time period task executions frequency f that the power consumption constraint condition is meant that this task-set is interior _iSatisfy following constraint:

$s.t.{\mathrm{\Σ}}_{i=1}^n{E}_i\left(f_i\right)\≤E_{\mathrm{budget}};$

In the formula, E _BudgetBe system energy consumption binding occurrence, E _i(f _i) be the f that executes the task _iRequired energy consumption, $E_i\left(f_i\right)=P_{\mathrm{indi}}\frac{C_i}{f_i}+C_{\mathrm{ef}}{C}_i{f}_i^2,$ P wherein _IndiFor carrying out real-time period task T _iRequired consumption and dynamic power consumption frequency-independent, C _EfEffective switching capacity for flush bonding processor;

Further, described all real-time period tasks satisfy temporal constraint and are meant the execution frequency f _iSatisfy following constraint:

$s.t.d_i-\frac{C_i}{f_i}\≥0;.$

In the formula, d _iBe real-time period task T _iThe execution off period;

Further, in the step 1), use artificial fish-swarm algorithm to real-time period task T _iThe execution frequency f _iFind the solution;

Further, use artificial fish-swarm algorithm to real-time period task T _iThe execution frequency f _iFind the solution specifically and comprise the steps:

11) initialization artificial fish-swarm, the state array X (X of artificial fish ₁, X ₂, X ₃, X ₄X _M) in each state value real-time period task executions frequency f in the corresponding task-set respectively _iFood concentration function F (X) is set is: ∏ _I=1 ⁿR _i(f _i); Calculate the reward value of every artificial fish;

12) find out artificial fish optimum in the shoal of fish, its state value is recorded bulletin board; The initialization algorithm parameter;

13) every artificial fish Simulation execution behavior of looking for food, knock into the back, bunch respectively selects optimum behavior as the final behavior of carrying out of this artificial fish;

14) state with the artificial fish of the artificial fish state after the step 13) act of execution and bulletin board record compares, if the state that is better than writing down in the bulletin board then upgrades state in bulletin board with this artificial fish current state, otherwise the state of bulletin board record is constant;

15) repeating step 13,14) iteration, finish condition up to reaching iteration, then export the state of the record of bulletin board, promptly carry out frequency f _iOptimum solution;

Further, the step of initialization artificial fish-swarm specifically comprises in the step 11):

The initialization of array at random artificial fish-swarm with a N * M dimension, artificial fish of each line display of this array, promptly this artificial fish-swarm has the artificial fish of N bar, and every artificial fish comprises a state array of being made up of M state value, and correspondence comprises the task-set of M task real-time period task; Every artificial fish is carried out initialization, and from first state value of artificial fish, frequency is carried out in one of Random assignment successively, and makes the task-set of this artificial fish correspondence satisfy temporal constraint and power consumption constraint condition;

Further, in the step 12), the continuous not max-thresholds of change frequency of default bulletin board optimal value during the initialization algorithm parameter; In the step 15), iteration is finished condition and is meant the continuous not max-thresholds of change frequency of the bulletin board optimal value that reaches default.

The present invention has following advantage with respect to prior art: take into account power consumption limitations, real-time and the reliability of system, and when reducing the embedded system energy consumption, real-time and reliability that the assurance task is carried out; Select suitable energy consumption, real-time and reliability constraint condition respectively, make resulting task carry out frequency and reach optimum; Owing to have three constraint conditions, make task carry out finding the solution of frequency and become a nonlinear combination optimization problem with complicated constraint, and the optimum solution of using artificial fish-swarm algorithm that task is carried out frequency is found the solution, characteristics with fast convergence rate, noninferior solution quality height, strong robustness, can try to achieve optimum solution exactly fast.

Other advantage of the present invention, target, to set forth in the following description to a certain extent with feature, and to a certain extent,, perhaps can obtain instruction from the practice of the present invention based on being conspicuous to those skilled in the art to investigating hereinafter.The objectives and other advantages of the present invention can be passed through following instructions, claims, and the specifically noted structure realizes and obtains in the accompanying drawing.

Description of drawings

In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in further detail below in conjunction with accompanying drawing:

Fig. 1 shows the schematic flow sheet of the energy consumption management method of embedded system;

Fig. 2 shows and uses fish-swarm algorithm to find the solution the optimum frequency synoptic diagram.

Embodiment

Hereinafter with reference to accompanying drawing, the preferred embodiments of the present invention are described in detail.

At first the defined power consumption of embedded system model of present embodiment, transient fault model and task model are described in detail.

The power consumption of embedded system model:

Power consumption of embedded system is defined as follows:

Wherein, P _SBe quiescent dissipation, promptly embedded system is in the power consumption under the dormant state; P _IndBe constant, expression and the irrelevant dynamic power consumption of processor frequencies are as the power consumption of processor peripheral hardwares such as external memory storage power consumption; P _dBe the dynamic power consumption relevant, comprise the dynamic power consumption of processor with processor frequencies, and other any power consumptions relevant with system voltage or processor frequencies, be defined as follows:

$P_d=\mathrm{\α}{C}_L{V}_{\mathrm{dd}}^{2}f---\left(2\right)$

α wherein, C _LBe the switching coefficient and the load capacitance of flush bonding processor, V _DdBe respectively the operating voltage and the frequency of flush bonding processor with f.Because V _DdLinear with f, so P _dCan equivalence be:

P _d＝C _eff ³ (3)

C wherein _EfBe effective switching capacity.In the formula (1),

Expression embedded system current working state, if

Then current embedded system is moved, if

Then current embedded system has been closed or has been in dormant state, and this moment, the embedded system dynamic power consumption was zero.This shows that the frequency of operation that reduces processor can reduce the dynamic energy consumption of frequency dependence, but because the prolongation of task execution time, the dynamic energy consumption with frequency-independent in the embedded system also can increase thereupon.Therefore, in order to reduce the energy consumption of embedded system, the frequency of operation of processor is not low more good more, certainly exists the frequency of operation f an of the best _EeMake embedded system energy consumption minimum:

$f_{\mathrm{ee}}=\sqrt[3]{\frac{P_{\mathrm{ind}}}{2C_{\mathrm{ef}}}}---\left(4\right)$

If processor working frequency is lower than f _Ee, the energy consumption of system will increase so, even the system energy consumption when moving with upper frequency than processor is also big, so the optimum frequency that should carry out greater than task of the frequency of operation of processor.

The transient fault model:

Because the influence (as hardware failure, electromagnetic interference (EMI), cosmic rays etc.) of various factors, task may break down and cause task to carry out failure in the process of implementation, and the probability of transient fault is far above the probability of permanent fault.The transient fault that studies show that system is obeyed Poisson distribution, and the parameter lambda of Poisson distribution is the function of processor frequencies:

$\mathrm{\λ}\left(f\right)={\mathrm{\λ}}_{0}g\left(f\right)={\mathrm{\λ}}_0{10}^{\frac{d(1-f)}{1-f_{\min }}}---\left(5\right)$

Suppose that processor has M discrete operating frequencies { f ₁, f ₂... f _M, f wherein _Min=f ₁, f _Max=f _M(normalization makes f _Max=1); λ then ₀Be frequency f _MaxThe average of=1 o'clock correspondence, so g (f _Max)=1.D is a constant, and the expression transient fault is to the susceptibility of processor frequencies and voltage.By (5) as can be known, reducing the voltage of processor and index that frequency can be brought system's transient fault probability rises.

Task model:

The task-set of n real-time period task composition is Ψ={ T ₁, T ₂..., T _n, separate between each task, all be in ready state, task T constantly 0 _iAvailable five-tuple is expressed as { C _i, P _Indi, P _i, d _i, R _i(f _i), wherein

(1) C _iExpression T _iUnder worst case, carry out required clock periodicity (WCC).

(2) P _IndiRepresent to execute the task T _iThat consume and dynamic power consumption frequency-independent.

Carry out individual task T _iRequired energy consumption is:

$E_i\left(f_i\right)=P_{\mathrm{indi}}\frac{C_i}{f_i}+C_{\mathrm{ef}}{C}_i{f}_i^2---\left(6\right)$

When frequency is f _EeThe time, task T _iThe energy consumption minimum.

(3) P _iBe task T _iPerformance period, d _iBe task T _iThe off period of carrying out, establish off period d _iAnd period T _iEquate, so, surpass then this task execution failure of off period when the task executions time.

(4) R _i(f _i) be task T _iThe probability of finishing is defined as:

$R_i\left(f_i\right)=e^{-\mathrm{\λ}\left(f_i\right)*\frac{C_i}{f_i}}---\left(7\right)$

Wherein, λ (f _i) by (4) definition, f _iBe T _iThe execution frequency, R _i(f _i) along with f _iIncrease and increase.The reliability of system is:

$R={\mathrm{\Π}}_{i=1}^n{R}_i\left(f_i\right)---\left(8\right)$

All tasks are all carried out on a variable voltage processor, and this processor can dynamically change the frequency (ignore frequency and adjust expense) of processor in the process of implementation, supposes that processor has M discrete operating frequencies { f ₁, f ₂... f _M, f wherein _Min=f ₁, f _Max=f _M(normalization makes f _Max=1).

The frequency of this processor can be at f _MinTo f _Max(normalization makes f _Max=1) regulates (ignoring the frequency adjustment expense), task T between arbitrarily _iIn frequency f _iUnder execution time t _iBe C _i/ f _i, required energy consumption is E _i(f _i).The total energy consumption constraint of system is designated as E _Budget, the total energy consumption of carrying out all tasks must not surpass E _Budget

To sum up, referring to Fig. 1, the energy consumption management method of the embedded system of present embodiment comprises the steps:

1) task-set of being made up of n real-time period task that need carry out for embedded system is found the solution each the real-time period task T in the task-set _iThe execution frequency f _iOptimum solution, make the reliability maximization of embedded system, that is, the maximization of the reliability of described embedded system is meant the real-time period task executions frequency f in this task-set _iSatisfy following formula:

Max∏ _i＝1 ⁿR(f _i)；

In the formula, $R_i\left(f_i\right)=e^{-\mathrm{\λ}\left(f_i\right)*\frac{C_i}{f_i}},$ Wherein, e is the nature index, C _iExpression real-time period task T _iUnder worst case, carry out required clock periodicity, $\mathrm{\λ}\left(f_i\right)={\mathrm{\λ}}_{0}g\left(f_i\right)={\mathrm{\λ}}_0{10}^{\frac{d(1-f_i)}{1-f_{\min }}},$ Wherein, λ ₀System's instantaneous failure rate average of correspondence during for execution frequency maximum, d is a constant, is the susceptibility of system's transient fault to processor frequencies and voltage, f _MinBe the processor frequencies minimum value after normalized;

And the energy consumption of carrying out all real-time period required by task satisfies the power consumption constraint condition, that is, the energy consumption of all real-time period required by task of described execution satisfies the real-time period task executions frequency f that the power consumption constraint condition is meant that this task-set is interior _iSatisfy following constraint:

$s.t.{\mathrm{\Σ}}_{i=1}^n{E}_i\left(f_i\right)\≤E_{\mathrm{budget}};$

In the formula, E _BudgetBe the system energy consumption binding occurrence, i.e. the battery energy that can provide, E _i(f _i) be the T that executes the task _iRequired energy consumption, $E_i\left(f_i\right)=P_{\mathrm{indi}}\frac{C_i}{f_i}+C_{\mathrm{ef}}{C}_i{f}_i^2,$ P wherein _IndiFor carrying out real-time period task T _iRequired consumption and dynamic power consumption frequency-independent, this value is the build-in attribute of each task, C _EfEffective switching capacity for flush bonding processor;

And all real-time period tasks satisfy temporal constraint, that is, described all real-time period tasks satisfy temporal constraint and are meant the execution frequency f _iSatisfy following constraint:

$s.t.d_i-\frac{C_i}{f_i}\≥0;.$

In the formula, d _iBe real-time period task T _iThe execution off period;

2) processor of embedded system is being carried out real-time period task T _iThe time, be this real-time period task executions frequency f that solves in the step 1) with the frequency configuration of processor _i

In the step 1), to real-time period task T _iThe execution frequency f _iOptimum solution find the solution and can use the whole bag of tricks of the prior art, but best because this is solved to a nonlinear combination optimization problem with complicated constraint, therefore utilize artificial fish-swarm algorithm to real-time period task T _iThe execution frequency f _iOptimum solution find the solution the most suitablely, referring to Fig. 2, utilize artificial fish-swarm algorithm to real-time period task T _iThe execution frequency f _iOptimum solution find the solution specifically and comprise the steps:

11) initialization of the array at random artificial fish-swarm of tieing up with a N * M, artificial fish of each line display of this array, promptly this artificial fish-swarm has the artificial fish of N bar, every artificial fish comprises a state array of being made up of M state value, corresponding task-set with M real-time period task, the state array X (X of artificial fish ₁, X ₂, X ₃, X ₄X _M) in each state value real-time period task executions frequency f in the corresponding task-set respectively _iFirst state value from every artificial fish, frequency is carried out in one of Random assignment successively, judge simultaneously whether the current task collection satisfies temporal constraint and power consumption constraint condition, do not adjust accordingly carrying out frequency if do not satisfy then, if still can not satisfy, then temporarily abandon current real-time period task and real-time period task afterwards thereof, the execution frequency assignment that is about in this artificial fish since then is 0; Food concentration (income degree) function F (X) is set is: ∏ _I=1 ⁿR _i(f _i); Calculate the reward value of every artificial fish;

12) find out artificial fish optimum in the shoal of fish, its state value is recorded bulletin board; The initialization algorithm parameter comprises the field range Visual of artificial fish, the crowding factor delta, and step-length step, the maximum number of times trynumber that sounds out, the bulletin board optimal value is the max-thresholds Maxbetter of change frequency not continuously; Define artificial fish X _iWith X _jBetween distance be Distance (X _i, X _j)=| X _i-X _j|+| X _i-X _j|, expression does not belong to X simultaneously _iWith X _jThe number of element.Define artificial fish X _iArtificial fish center in the visual field is $X_{\mathrm{center}}=\frac{1}{n_f}{\mathrm{\Σ}}_{j=1}^{n_f}{X}_j,$ X _jBe X _iArtificial fish in the visual field (comprises X _i).

13) every artificial fish Simulation execution behavior of looking for food, knock into the back, bunch respectively selects optimum behavior as the final behavior of carrying out of this artificial fish; Check whether this artificial fish satisfies temporal constraint and power consumption constraint condition when each artificial fish state changes, then it is not modified to a feasible state value the most close with current state if do not satisfy;

Foraging behavior: artificial fish X _iIn its visual field, select a state X at random _j, i.e. Distance (X _i, X _jIf)＜Visual is F (X _j)＞F (X _i), then move and move a step, that is: to this direction $X_i=\mathrm{step}\·\frac{X_j-X_i}{|X_j-X_i|}+X_i,$ Otherwise select other states X again at random _j, judge whether to satisfy the condition of advancing, then do not move at random and move a step if still do not satisfy the progress bar part after attempting trynumber time, then make new state of Xi arrival, X _i=X _i+ rand, rand are random number;

The behavior of knocking into the back: calculate artificial fish X _iIf the income degree of all partner fishes in the visual field is the optimum partner X of income degree _jAround not too crowded, X then _iMove towards this partner and to move a step, $X_i=\mathrm{step}\·\frac{X_j-X_i}{|X_j-X_i|}+X_i,$ Otherwise execution foraging behavior;

The behavior of bunching: artificial fish X _iSearch for the number of partners n in its visual field _f, and center X _Center, if F (X _Center)＞F (X _i) and n _f＜δ, then X _iMove towards partner's center and to move a step, $X_i=\mathrm{step}\·\frac{X_{\mathrm{center}}-X_i}{|X_{\mathrm{center}}-X_i|}+X_i,$ Otherwise execution foraging behavior;

14) with the step 13) act of execution and the state of the artificial fish state after checking and the bulletin board artificial fish of writing down compare, if the state that is better than writing down in the bulletin board is then with the state in this artificial fish current state renewal bulletin board, otherwise the state of bulletin board record is constant;

15) repeating step 13,14) iteration, finish condition up to reaching iteration, the iteration of present embodiment finish condition be the bulletin board optimal value continuously not change frequency reach max-thresholds Maxbetter, then export the state of the record of bulletin board, promptly carry out frequency f _iOptimum solution.

The above is the preferred embodiments of the present invention only, is not limited to the present invention, and obviously, those skilled in the art can carry out various changes and modification and not break away from the spirit and scope of the present invention the present invention.Like this, if of the present invention these are revised and modification belongs within the scope of claim of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these changes and modification interior.

Claims (8)

1. the energy consumption management method of embedded system is characterized in that, comprises the steps:

1) task-set of being made up of n real-time period task that need carry out for embedded system is found the solution each the real-time period task T in the task-set _iThe execution frequency f _iOptimum solution, make the reliability maximization of embedded system, and the energy consumption of carrying out all real-time period required by task satisfies the power consumption constraint condition, and all real-time period tasks satisfy temporal constraint;

2) processor of embedded system is being carried out real-time period task T _iThe time, be this real-time period task executions frequency f that solves in the step 1) with the frequency configuration of processor _i

2. the energy consumption management method of embedded system according to claim 1 is characterized in that: the reliability maximization of described embedded system is meant the real-time period task executions frequency f in this task-set _iSatisfy following formula:

Max∏ _i=1 ⁿR _i(f _i)；

In the formula, $R_i\left(f_i\right)=e^{-\mathrm{\λ}\left(f_i\right)*\frac{C_i}{f_i}},$ Wherein, e is the nature index, C _iExpression real-time period task T _iUnder worst case, carry out required clock periodicity, $\mathrm{\λ}\left(f_i\right)={\mathrm{\λ}}_{o}g\left(f_i\right)={\mathrm{\λ}}_o{10}^{\frac{d(1-f_i)}{1-f_{\min }}},$ Wherein, λ ₀System's instantaneous failure rate average of correspondence during for execution frequency maximum, d is a constant, is the susceptibility of system's transient fault to processor frequencies and voltage, f _MinBe the processor frequencies minimum value after normalized.

3. the energy consumption management method of embedded system according to claim 1 is characterized in that: the energy consumption of all real-time period required by task of described execution satisfies the power consumption constraint condition and is meant real-time period task executions frequency f in this task-set _iSatisfy following constraint:

$s.t.{\mathrm{\Σ}}_{i=1}^n{E}_i\left(f_i\right)\≤E_{\mathrm{budget}};$

In the formula, E _BudgetBe system energy consumption binding occurrence, E _i(f _i) be the T that executes the task _iRequired energy consumption, $E_i\left(f_i\right)=P_{\mathrm{indi}}\frac{C_i}{f_i}+C_{\mathrm{ef}}{C}_i{f}_i^2,$ P wherein _IndiFor carrying out real-time period task T _iRequired consumption and dynamic power consumption frequency-independent, C _EfEffective switching capacity for flush bonding processor.

4. the energy consumption management method of embedded system according to claim 1, it is characterized in that: described all real-time period tasks satisfy temporal constraint and are meant the execution frequency f _iSatisfy following constraint:

$s.t.d_i-\frac{C_i}{f_i}\≥0;$

In the formula, d _iBe real-time period task T _iThe execution off period.

5. according to the energy consumption management method of each described embedded system in the claim 1 to 4, it is characterized in that: in the step 1), use artificial fish-swarm algorithm real-time period task T _iThe execution frequency f _iFind the solution.

6. the energy consumption management method of embedded system according to claim 5 is characterized in that: use artificial fish-swarm algorithm to real-time period task T _iThe execution frequency f _iFind the solution specifically and comprise the steps:

11) initialization artificial fish-swarm, the state array X (X of artificial fish ₁, X ₂, X ₃, X ₄X _M) in each state value real-time period task executions frequency f in the corresponding task-set respectively _iFood concentration function F (X) is set is: ∏ _I=1 ⁿR _i(f _i); Calculate the reward value of every artificial fish;

12) find out artificial fish optimum in the shoal of fish, its state value is recorded bulletin board; The initialization algorithm parameter;

13) every artificial fish Simulation execution behavior of looking for food, knock into the back, bunch respectively selects optimum behavior as the final behavior of carrying out of this artificial fish;

14) state with the artificial fish of the artificial fish state after the step 13) act of execution and bulletin board record compares, if the state that is better than writing down in the bulletin board then upgrades state in bulletin board with this artificial fish current state, otherwise the state of bulletin board record is constant;

15) repeating step 13,14) iteration, finish condition up to reaching iteration, then export the state of the record of bulletin board, promptly carry out frequency f _iOptimum solution.

7. the energy consumption management method of embedded system according to claim 6, it is characterized in that: the step of initialization artificial fish-swarm specifically comprises in the step 11):

The initialization of array at random artificial fish-swarm with a N * M dimension, artificial fish of each line display of this array, promptly this artificial fish-swarm has the artificial fish of N bar, and every artificial fish comprises a state array of being made up of M state value, and correspondence comprises the task-set of M task real-time period task; Every artificial fish is carried out initialization, and from first state value of artificial fish, frequency is carried out in one of Random assignment successively, and makes the task-set of this artificial fish correspondence satisfy temporal constraint and power consumption constraint condition.

8. the energy consumption management method of embedded system according to claim 6 is characterized in that: in the step 12), preset the continuous not max-thresholds of change frequency of bulletin board optimal value during the initialization algorithm parameter; In the step 15), iteration is finished condition and is meant the continuous not max-thresholds of change frequency of the bulletin board optimal value that reaches default.

Images