CN104168661A - Transmission scheduling method for network lifetime maximization that satisfies fairness condition - Google Patents

Abstract

The invention relates to a transmission scheduling method for network lifetime maximization that satisfies a fairness condition, and belongs to the field of wireless communication. The method includes modeling the problem of sensor node selection into a constraint Markov decision process, using a Bellman equation in dynamic programming to solve a CMDP problem, reducing computation complexity to obtain a final constraint optimal lifetime and an optimal strategy, and determining a sensor node selected at each time slot. Aiming at a circumstance that a wireless body area network needs to satisfy certain fairness in practical application, the method in the invention overcomes the defect of a conventional transmission scheduling method which only considers lifetime maximization but neglects fairness, maximizes the lifetime of the network while ensuring satisfaction of a fairness constraint condition, and thus is more suitable for a circumstance of monitoring multiple physiological parameters of a patient.

Description

A kind of maximization network transmission dispatching method life cycle that meets fairness condition

Technical field

The present invention relates to wireless communication field, is a kind of signal transmission dispatching method towards wireless body area network.

Background technology

Wireless body area network (WBAN, Wireless Body Area Network) in the application of the aspects such as medical monitoring, becomes the focus that people pay close attention to as the important component part of Internet of Things and wireless sensor network gradually.Because it is battery-powered to form the sensor node of wireless body area network, energy is subject to strict restriction, so design efficient, fair transmission dispatching method, extend research emphasis and the focus that network lifetime is body area network, not only there is theoretical significance, also there is important actual application value.

The people such as Dagher have proposed a kind of maximization network iterative algorithm of life cycle, and the problem of maximization network life cycle is summed up as to PO (Pareto-Optimal) problem and has drawn optimal solution.Chen etc. have proposed the vague generalization formula of life cycle in wireless sensor network, and have proved that channel condition information and dump energy information are the principal elements that affects network lifetime.Chen etc. proposes again a network lifetime that is called as the dynamic transmission scheduling protocol of DPLM (Dynamic Protocol for Lifetime Maximization) and has obtained asymptotic optimization subsequently.Cohen etc. have proposed to become when a kind of chance agreement, and simulation result shows to be better than life cycle that this agreement obtains other several agreements with its contrast.Madan etc. have proposed a kind of combined optimization algorithm that combines physical layer, MAC layer and route layer, and this algorithm is generalized into a Nonlinear Optimization Problem by calculating optimal path stream, link scheduling and through-put power, has greatly extended network lifetime.The method of the utilization collaboration communications such as Zhai has designed a kind of new cooperative MAC protocol, effectively reduces transmitting power, has improved the life cycle of wireless sensor network.2012, Movassaghi etc. propose a self-induced body area network efficiency route protocol of heat and energy, according to the temperature of node and energy level, be the selection that each path computing cost function decides route, effectively utilize available resources, extended the life cycle of body area network.Ortiz etc. have proposed a kind of adaptive routing protocol based on multi-hop tree, have built a spanning tree of simultaneously considering node battery allowance, received signal strength and jumping figure, and the energy consumption of agreement between can balance node extends network useful life.Liu Hanchun etc. be take and extended body area network network lifetime and proposed a kind of high energy efficiency routing algorithm based on adding interim node as target.The new situation that this algorithm occurs in actual applications for body area network, consumes by making full use of in the dump energy balancing network of interim node energy between each sensor node, and simulation result shows, this algorithm has extended the life cycle of network greatly.

The resource of network is answered fair each sensor node of distributing to, and fairness is one of key issue in wireless body area network dispatching method.Fair dispatching method should be according to the corresponding resource of characteristic demand assignment of each sensor node, prevents that certain node from taking too much channel resource and having influence on other users' transmission.Wireless body area network is very high to the requirement of real-time, causes other nodes to can not get for a long time chance transmission if certain sensor node takies too much channel resource, to important physiological parameter early warning, can cause very serious consequence not in time.Good transmission dispatching method not only will maximize the life cycle of whole network, also should guarantee the fairness of each sensor node simultaneously.The method of above-mentioned various maximization network life cycle is not all considered the requirement of fairness, can not meet the multiple sensors node situation different to resource requirement in practical application completely.Retrieval through to prior art document, there is not yet and combined to the relevant report of the method for consideration life cycle with fairness.

Summary of the invention

The invention provides a kind of maximization network transmission dispatching method life cycle that meets fairness condition, to solve the method for current various maximization network life cycle, all do not consider the requirement of fairness, can not meet the problem of multiple sensors node to the different situation of resource requirement in practical application completely.

The technical scheme that the present invention takes is to comprise

(1) wireless body area network, this wireless body area network is one and take the network that human body is monitoring target, the different sensor node of several functions is placed in to the relevant position in body surface or body, monitor periodically and record various physiologic informations, by the data of collection, the human body communication technology by bluetooth, Zigbee, ultra broadband or other modes passes to aggregation node AP (Access Point) to each sensor node, and AP crosses external network by these information exchanges again and is sent to remote control center;

In the mac-layer protocol of IEEE802.15.6 communication standard, time has been divided into isometric superframe structure, superframe structure comprises following four parts: the control stage, contention access stage CAP, competition idle phase CFP, non-activation phase, the CFP stage continues to be again divided into some time slots, in wireless body area network, time is the resource to each sensor node to be allocated, in wireless body area network, all the sensors node shares a channel, at each time slot, can only there is a sensor node that the data of collection are sent to AP node, the selected required energy of sensor node consumption transmission data,

(2) problem of sensor node being selected is modeled as constraint Markovian decision process, and a tactful u in a lucky corresponding CMDP problem of transmitting and scheduling agreement, is L by start maximum network lifetime from state i ^*(i), be expressed as follows:

$L^{*}\left(i\right)=\underset{u}{\max }{L}_u\left(i\right)$

If the remuneration maximum (being that network lifetime is the longest) that network is used certain tactful u to obtain before reaching halted state, we claim that this strategy is optimal policy so, use u ^*represent,

$L_{u^{*}}\left(i\right)=L^{*}\left(i\right),\∀i\∈S\backslash S_t$

Definition U _a={ u ∈ U:f>=f _nit is possible strategy collection.If tactful u ^*∈ U _al (u satisfies condition ^*)>=L (u), (u ∈ U _a), u ^*be called as constraint optimal policies;

The target of CMDP is to find an optimum tactful u ^*∈ U _acarry out maximization network L life cycle, optimum transmitting and scheduling agreement is by optimal policy u ^*provide clear and definite optimal policy u ^*just can know at which sensor node of each Slot selection and send data;

Parameters in CMDP model is described below:

(a) state space

At each time slot, the state i of network is by dump energy e, transmits needed energy w and fair coefficient f forms jointly.We define grid state space S is as follows:

S＝{i＝(e,w,f)}

When network lifetime exhausts, network reaches a special state of termination S _tas follows:

$S_t=\{(e,w,f):\&ForAll;\{e_n:e_n<{\mathrm{\&epsiv;}}_1\}\mathrm{ore}<w\}$

E _nfor the dump energy of sensor node n, ε ₁be the energy that a time slot sensor node adopts minimum emissive power transmission data to need, e _n: e _n< ε ₁the dump energy that represents sensor node n all can not complete once transmission under any channel conditions, and e<w means a bust this;

(b) actionable space

With A, represent the set of action, when network is during in state i=(e, w, f) ∈ S, actionable space can be expressed as:

A(i)＝A[(e,w,f)]＝{n:e _n≥w _n}

According to the definition of state of termination, can draw, the actionable space of any nonfinal state is non-NULL;

(c) transition probability

When the state of network is i, after the effect of action a, the probability that next state is j is

$P_{\mathrm{iaj}}=p\left(w^{\&prime;}\right)1_{[e^{\&prime;}=e-I_n{w}_n]}$

Wherein p (w')=Pr{W=w'} is the probability density function of W, by channel fading, is determined;

(d) transmission remuneration

After every data transfer, network all can obtain the return of a unit, until network enters halted state, that is to say, if the state of network is the return immediately of this time slot is

$R\left(i\right)=1_{[i\&Element;S\backslash S_T]}$

(e) constraints

Suppose that network is in state i, if the fair coefficient f of network is more than or equal to given threshold value f after the effect of action a _n, we a that claims to take action is feasible action, formula can be expressed as follows:

f(i,a)≥f _n

F (i, a) represents the fair coefficient of network when state i arrives next state after the effect of a, network at the feasible action collection of state i by A _a(i) represent;

(f) strategy

Tactful u in set of strategies U is a sequence u={u ₀, u ₁... }, u wherein _n: S → and 1 ..., N} is illustrated in the sensor node of n Slot selection, u _n(i ₁, a ₁, i ₂, a ₂..., i _n-1, a _n-1, i _n) be the conditional probability measure of A, the moment exhausting with first node energy tolerance life cycle, so, L life cycle of whole network can arrive halted state St by network before all return sums describe, definition L _u(i) for network is from state i, the life cycle of usage policy u, i.e. all return sums;

(3) solve optimal policy

From state i, meet the maximum network lifetime L of constraints ^*(i) be the unique solution of the optimum equation of following Bellman,

$L^{*}\left(i\right)=L^{*}\left[(e,w,f)\right]=R\left(i\right)\underset{n\&Element;A\left(i\right)}{\max }\left\{\underset{j\&Element;S}{\mathrm{\&Sigma;}}{p}_{\mathrm{iaj}}{L}^{*}\left(j\right)\right\},\&ForAll;i\&Element;S$

s.t.f _j≥f _n

F wherein _jthe fair coefficient of network under state j, f _nit is the fair threshold value providing according to different application scenarioss;

In fact, formula above can also be write

$L^{*}\left(i\right)=L^{*}\left[(e,w,f)\right]=R\left(i\right)\underset{n\&Element;A_a\left(i\right)}{\max }\left\{\underset{j\&Element;S}{\mathrm{\&Sigma;}}{p}_{\mathrm{iaj}}{L}^{*}\left(j\right)\right\},\&ForAll;i\&Element;S$

The optimal policy u of transmitting and scheduling scheme can be drawn by following formula

$u\left(i\right)=\underset{n\&Element;A_a\left(i\right)}{\arg \max }\left\{{\underset{j\&Element;S}{\mathrm{\&Sigma;}}{p}_{\mathrm{iaj}}L}^{*}\left(j\right)\right\},\&ForAll;i\&Element;S\backslash S_t$

(4) reduce computation complexity

An optimum equation of equivalent Bellman is write as following form:

$L[e,f]=\underset{w}{\mathrm{\&Sigma;}}p\left(w\right)\left\{R\right[e,w]+\underset{n\&Element;A_a}{\max }L[e-I_n{w}_n\left]\right\}$

So constraint optimal policies also can be obtained by following formula

$u\left[(e,w,f)\right]=\arg \underset{n\&Element;A_a}{\max }L[e-I_n{w}_n]$

In implementation procedure, each time slot at first, beacon signal of AP node broadcasts is waken all the sensors node up, in order to make AP node obtain global channel information, all sensor nodes all need to send the beacon signal that a pilot signal is replied AP, the signal that the utilization of AP node receives is estimated the channel conditions of all nodes and is obtained the energy w that transmission needs, then according to state (e, w, f) calculate optimal policy u, clear and definite optimal policy, known the sensor node that should select at each time slot, finally, No. ID of the selected sensor node of AP node broadcasts, this sensor node starts to transmit the data that it collects, because AP node is known the channel conditions of all the sensors node, so can continue to upgrade network state and be the arrival of next time slot prepares.

Fair coefficient f of the present invention is defined as follows: definition (T ₁, T ₂..., T _n) representing the number of transmissions of each sensor node reality, O represents the total degree having transmitted, (b ₁, b ₂..., b _n) mean the weights of each sensor node importance.For sensor node n, define standardized the number of transmissions and be

x _n＝T _n/b _nO

Suppose to have N user's competition network resource, the resource that nth user obtains is x _n, the fair coefficient of define grid is so

$f=\frac{{\left(\mathrm{\&Sigma;}{x}_n\right)}^2}{\mathrm{N\&Sigma;}{x_n}^2}$

F is value between 0 and 1.If the fair coefficient f=1 of network, represents that network is completely fair.The fair coefficient of network is larger, and fair degree is higher, and vice versa.

The problem that the present invention selects sensor node is carried out modeling and is summed up as a constraint Markovian decision process (CMDP, Constrained Markov Decision Processes) problem, proposed a kind of under fairness constraints maximization network life cycle transmission dispatching method and solve optimum life cycle and the optimal policy under fairness constraint.

The present invention is directed to wireless body area network needs to meet the situation of certain fairness in actual applications, overcome that conventional transmission dispatching method is only considered to maximize life cycle and the shortcoming of having ignored fairness, guaranteeing to meet the life cycle that adopts optimal policy to maximize network under fairness constraints, be more applicable for the situation to a multinomial physiological parameter of patient monitoring.

Accompanying drawing explanation

Fig. 1 is WBAN network simulation model of the present invention;

Fig. 2 is that the present invention requires lower optimal policy and other agreement comparison diagram life cycle without fairness;

Fig. 3 be transducer weighted value of the present invention when identical network lifetime with fair index variation figure;

When Fig. 4 is transducer weighted value difference of the present invention, network lifetime is with fair index variation figure.

Embodiment

Comprise:

(1) wireless body area network, this wireless body area network is one and take the network that human body is monitoring target, the different sensor node of several functions is placed in to the relevant position in body surface or body, monitor periodically and record various physiologic informations, by the data of collection, the human body communication technology by bluetooth, Zigbee, ultra broadband or other modes passes to aggregation node AP (Access Point) to each sensor node, and AP crosses external network by these information exchanges again and is sent to remote control center;

Remote medical monitoring is the most typical application of wireless body area network, particularly be used for monitoring continuously and recording some chronic diseases (as heart disease, diabetes and asthma etc.) patient's physiological parameter, by being placed in the various physiological parameters of the sensor node Real-time Collection human body of human body, as blood oxygen, pulse, body temperature, electrocardio etc. and be transferred to the object that remote medical center is monitoring real-time health guidance is provided.Patient can not be subject to the constraint of instrument cable like this, has expanded activity space.

In the mac-layer protocol of IEEE802.15.6 communication standard, time has been divided into isometric superframe structure, superframe structure comprises following four parts: the control stage, the contention access stage (CAP), competition idle phase (CFP), non-activation phase, the CFP stage continues to be again divided into some time slots, we pay close attention to the agreement based on TDMA that packet mainly transmits in the CFP stage, so in wireless body area network, time is the resource to each sensor node to be allocated, because all the sensors node shares a channel in wireless body area network, at each time slot, can only there is a sensor node that the data of collection are sent to AP node, the selected required energy of sensor node consumption transmission data, the transmission of which sensor node of each Slot selection can meet node fairness require under life cycle of maximization network be the key issue in research,

(2) problem of sensor node being selected is modeled as constraint Markovian decision process, and a tactful u in a lucky corresponding CMDP problem of transmitting and scheduling agreement, is L by start maximum network lifetime from state i ^*(i), be expressed as follows:

$L^{*}\left(i\right)=\underset{u}{\max }{L}_u\left(i\right)$

If the remuneration maximum (being that network lifetime is the longest) that network is used certain tactful u to obtain before reaching halted state, we claim that this strategy is optimal policy so, use u ^*represent,

$L_{u^{*}}\left(i\right)=L^{*}\left(i\right),\&ForAll;i\&Element;S\backslash S_t$

Definition U _a={ u ∈ U:f>=f _nit is possible strategy collection.If tactful u ^*∈ U _al (u satisfies condition ^*)>=L (u), (u ∈ U _a), u ^*be called as constraint optimal policies;

The target of CMDP is to find an optimum tactful u ^*∈ U _acarry out maximization network L life cycle, optimum transmitting and scheduling agreement is by optimal policy u ^*provide clear and definite optimal policy u ^*just can know at which sensor node of each Slot selection and send data;

Parameters in CMDP model is described below:

(a) state space

At each time slot, the state i of network is by dump energy e, transmits needed energy w and fair coefficient f forms jointly.We define grid state space S is as follows:

S＝{i＝(e,w,f)}

When network lifetime exhausts, network reaches a special state of termination S _tas follows:

$S_t=\{(e,w,f):\&ForAll;\{e_n:e_n<{\mathrm{\&epsiv;}}_1\}\mathrm{ore}<w\}$

E _nfor the dump energy of sensor node n, ε ₁be the energy that a time slot sensor node adopts minimum emissive power transmission data to need, e _n: e _n< ε ₁the dump energy that represents sensor node n all can not complete once transmission under any channel conditions, and e<w means a bust this;

(b) actionable space

With A, represent the set of action, when network is during in state i=(e, w, f) ∈ S, actionable space can be expressed as:

A(i)＝A[(e,w,f)]＝{n:e _n≥w _n}

According to the definition of state of termination, can draw, the actionable space of any nonfinal state is non-NULL;

(c) transition probability

When the state of network is i, after the effect of action a, the probability that next state is j is

$P_{\mathrm{iaj}}=p\left(w^{\&prime;}\right)1_{[e^{\&prime;}=e-I_n{w}_n]}$

Wherein p (w')=Pr{W=w'} is the probability density function of W, by channel fading, is determined;

(d) transmission remuneration

After every data transfer, network all can obtain the return of a unit, until network enters halted state, that is to say, if the state of network is the return immediately of this time slot is

$R\left(i\right)=1_{[i\&Element;S\backslash S_T]}$

(e) constraints

Suppose that network is in state i, if the fair coefficient f of network is more than or equal to given threshold value f after the effect of action a _n, we a that claims to take action is feasible action, formula can be expressed as follows:

f(i,a)≥f _n

F (i, a) represents the fair coefficient of network when state i arrives next state after the effect of a, network at the feasible action collection of state i by A _a(i) represent;

(f) strategy

Tactful u in set of strategies U is a sequence u={u ₀, u ₁... }, u wherein _n: S → and 1 ..., N} is illustrated in the sensor node of n Slot selection, u _n(i ₁, a ₁, i ₂, a ₂..., i _n-1, a _n-1, i _n) be the conditional probability measure of A, the moment exhausting with first node energy is measured life cycle, so L life cycle of whole network can arrive halted state S by network _tall return sums are described before, definition L _u(i) for network is from state i, the life cycle of usage policy u, i.e. all return sums;

(3) solve optimal policy

From state i, meet the maximum network lifetime L of constraints ^*(i) be the unique solution of the optimum equation of following Bellman,

$L^{*}\left(i\right)=L^{*}\left[(e,w,f)\right]=R\left(i\right)\underset{n\&Element;A\left(i\right)}{\max }\left\{\underset{j\&Element;S}{\mathrm{\&Sigma;}}{p}_{\mathrm{iaj}}{L}^{*}\left(j\right)\right\},\&ForAll;i\&Element;S$

s.t.f _j≥f _n

F wherein _jthe fair coefficient of network under state j, f _nit is the fair threshold value providing according to different application scenarioss;

In fact, formula above can also be write

$L^{*}\left(i\right)=L^{*}\left[(e,w,f)\right]=R\left(i\right)\underset{n\&Element;A_a\left(i\right)}{\max }\left\{\underset{j\&Element;S}{\mathrm{\&Sigma;}}{p}_{\mathrm{iaj}}{L}^{*}\left(j\right)\right\},\&ForAll;i\&Element;S$

The optimal policy u of transmitting and scheduling scheme can be drawn by following formula

$u\left(i\right)=\underset{n\&Element;A_a\left(i\right)}{\arg \max }\left\{{\underset{j\&Element;S}{\mathrm{\&Sigma;}}{p}_{\mathrm{iaj}}L}^{*}\left(j\right)\right\},\&ForAll;i\&Element;S\backslash S_t$

(4) reduce computation complexity

An optimum equation of equivalent Bellman is write as following form:

$L[e,f]=\underset{w}{\mathrm{\&Sigma;}}p\left(w\right)\left\{R\right[e,w]+\underset{n\&Element;A_a}{\max }L[e-I_n{w}_n\left]\right\}$

So constraint optimal policies also can be obtained by following formula

$u\left[(e,w,f)\right]=\arg \underset{n\&Element;A_a}{\max }L[e-I_n{w}_n]$

In implementation procedure, each time slot at first, beacon signal of AP node broadcasts is waken all the sensors node up, in order to make AP node obtain global channel information, all sensor nodes all need to send the beacon signal that a pilot signal is replied AP, the signal that the utilization of AP node receives is estimated the channel conditions of all nodes and is obtained the energy w that transmission needs, then according to state (e, w, f) calculate optimal policy u, clear and definite optimal policy, known the sensor node that should select at each time slot, finally, No. ID of the selected sensor node of AP node broadcasts, this sensor node starts to transmit the data that it collects, because AP node is known the channel conditions of all the sensors node, so can continue to upgrade network state and be the arrival of next time slot prepares.

Fair coefficient f of the present invention is defined as follows: definition (T ₁, T ₂..., T _n) representing the number of transmissions of each sensor node reality, O represents the total degree having transmitted, (b ₁, b ₂..., b _n) mean the weights of each sensor node importance.For sensor node n, define standardized the number of transmissions and be

x _n＝T _n/b _nO

Suppose to have N user's competition network resource, the resource that nth user obtains is x _n, the fair coefficient of define grid is so

$f=\frac{{\left(\mathrm{\&Sigma;}{x}_n\right)}^2}{\mathrm{N\&Sigma;}{x_n}^2}$

F is value between 0 and 1.If the fair coefficient f=1 of network, represents that network is completely fair.The fair coefficient of network is larger, and fair degree is higher, and vice versa.

Below in conjunction with design parameter and accompanying drawing, the present invention will be further described:

Parameter declaration: WBAN artificial network model as shown in Figure 1, has three sensor nodes (node 1,2 and 3) and an AP node (node 4) in assumed wireless body area network.Except AP, other nodes all gather human body physiological parameter regularly, and the data of collection are sent to AP by common channel.Transmitting energy grade w={1,2,3}, to all sensor node n, channel distribution network lifetime represents with the expectation of the number of transmissions.

In Fig. 2 emulation the network lifetime that obtains of optimal policy, and compare with other several host-host protocols.Here, ignore the expense of obtaining global channel information, only paid close attention to the difference of optimum limiting performance He other sub-optimal performance.Optimal policy and other four kinds of host-host protocols are made comparisons: 1) choose at random sensor node transmission; 2) choose the chance scheduling protocol of the best sensor node transmission of channel conditions; 3) choose the conservative scheduling protocol of the sensor node transmission that dump energy is maximum; 4) DPLM agreement, considers channel condition information and dump energy information and chooses sensor node transmission data.Maximization network need to have been weighed two conflicting targets life cycle: the dump energy when minimizing the energy consumption of each time slot and minimizing network death.The sensor node that it is good that the former tends to selective channel situation carrys out transmission information and the latter tends to select the sensor node that dump energy is many to send.Above-mentioned chance scheduling protocol and conservative scheduling protocol have only been paid close attention to one of them of two conflicting targets, so the scheduling protocol which neither be optimum.DPLM protocol synthesis has been considered channel condition information and dump energy information, but its application is local channel condition information, and each sensor node is only known channel condition information and the dump energy information of oneself, and does not know the information of other sensor nodes.Although cannot use Markovian decision process modeling, DPLM has still obtained the performance of near-optimization.The agreement of application optimal policy has certainly obtained optimum performance, and becomes the benchmark of other agreement contrasts.

Fig. 2 is expressed as network lifetime the function of primary power, and along with the increase of sensor node primary power, also increase accordingly the life cycle that variety of protocol is corresponding.The agreement of application optimal policy has well shown performance life cycle that it is optimum, and DPLM agreement has also obtained the performance of near-optimization.The network lifetime that conservative scheduling protocol obtains is slightly better than chance scheduling protocol, and it is the poorest to choose at random performance life cycle of agreement of sensor node.

Fig. 2 is in the emulation not having under fairness constraints, below with fairness being required to the emulation respectively of different situations with constrained optimal policy at network, as shown in Figure 3.

Simulation parameter is as follows: supposing has three sensor nodes in body area network, transmitting energy grade w={1, and 2,3}, to all sensor node n, channel distribution given fair coefficient threshold value is respectively f _n=0.3,0.5,0.7,0.9, network lifetime represents with expected transmission times.If the weight of each sensor node is identical, be b _n=[13,13,13].As can be seen from Figure 3, the constrained optimum of network reduces along with the increase that network fairness is required life cycle.Higher at the fairness performance requirement to network, as f _nthe network lifetime of=0.9 o'clock drops to very low.And not high to the fairness performance requirement of network, as f _nperformance life cycle of=0.3 o'clock is still fine.This explanation is in certain scope, and be a pair of conflicting performance index the life cycle of fairness and network.The method can be issued to according to application scenarios the life cycle of constrained optimum in the requirement that meets different fairness in practice.

The weight of establishing each sensor node in Fig. 4 is respectively b _n=[0.7,0.2,0.1], this represents that the importance of first node is far longer than other two sensor nodes.Comparison diagram 3 and Fig. 4 can find out, when transducer weighted value is different, along with the variation that network fairness requires, the variation of life cycle is larger, and weighted value identical time the variation of life cycle less.The scheme of this explanation optimal policy is larger to the different situation effect of transducer weighted value, is more applicable for the situation to a multinomial physiological parameter of patient monitoring.

Claims (2)

1. maximization network transmission dispatching method life cycle that meets fairness condition, is characterized in that, comprising:

(1) wireless body area network, this wireless body area network is one and take the network that human body is monitoring target, the different sensor node of several functions is placed in to the relevant position in body surface or body, monitor periodically and record various physiologic informations, by the data of collection, the human body communication technology by bluetooth, Zigbee, ultra broadband or other modes passes to aggregation node AP (Access Point) to each sensor node, and AP crosses external network by these information exchanges again and is sent to remote control center;

In the mac-layer protocol of IEEE802.15.6 communication standard, time has been divided into isometric superframe structure, superframe structure comprises following four parts: the control stage, contention access stage CAP, competition idle phase CFP, non-activation phase, the CFP stage continues to be again divided into some time slots, in wireless body area network, time is the resource to each sensor node to be allocated, in wireless body area network, all the sensors node shares a channel, at each time slot, can only there is a sensor node that the data of collection are sent to AP node, the selected required energy of sensor node consumption transmission data,

(2) problem of sensor node being selected is modeled as constraint Markovian decision process, and a tactful u in a lucky corresponding CMDP problem of transmitting and scheduling agreement, is L by start maximum network lifetime from state i ^*(i), be expressed as follows:

If the remuneration maximum (being that network lifetime is the longest) that network is used certain tactful u to obtain before reaching halted state, we claim that this strategy is optimal policy so, use u ^*represent,

Definition U _a={ u ∈ U:f>=f _nit is possible strategy collection.If tactful u ^*∈ U _al (u satisfies condition ^*)>=L (u), (u ∈ U _a), u ^*be called as constraint optimal policies;

The target of CMDP is to find an optimum tactful u ^*∈ U _acarry out maximization network L life cycle, optimum transmitting and scheduling agreement is by optimal policy u ^*provide clear and definite optimal policy u ^*just can know at which sensor node of each Slot selection and send data;

Parameters in CMDP model is described below:

(a) state space

At each time slot, the state i of network is by dump energy e, transmits needed energy w and fair coefficient f forms jointly.We define grid state space S is as follows:

S＝{i＝(e,w,f)}

When network lifetime exhausts, network reaches a special state of termination S _tas follows:

E _nfor the dump energy of sensor node n, ε ₁be the energy that a time slot sensor node adopts minimum emissive power transmission data to need, e _n: e _n< ε ₁the dump energy that represents sensor node n all can not complete once transmission under any channel conditions, and e<w means a bust this;

(b) actionable space

With A, represent the set of action, when network is during in state i=(e, w, f) ∈ S, actionable space can be expressed as:

A(i)＝A[(e,w,f)]＝{n:e _n≥w _n}

According to the definition of state of termination, can draw, the actionable space of any nonfinal state is non-NULL;

(c) transition probability

When the state of network is i, after the effect of action a, the probability that next state is j is

Wherein p (w')=Pr{W=w'} is the probability density function of W, by channel fading, is determined;

(d) transmission remuneration

After every data transfer, network all can obtain the return of a unit, until network enters halted state, that is to say, if the state of network is the return immediately of this time slot is

(e) constraints

Suppose that network is in state i, if the fair coefficient f of network is more than or equal to given threshold value f after the effect of action a _n, we a that claims to take action is feasible action, formula can be expressed as follows:

f(i,a)≥f _n

F (i, a) represents the fair coefficient of network when state i arrives next state after the effect of a, network at the feasible action collection of state i by A _a(i) represent;

(f) strategy

Tactful u in set of strategies U is a sequence u={u ₀, u ₁... }, u wherein _n: S → and 1 ..., N} is illustrated in the sensor node of n Slot selection, u _n(i ₁, a ₁, i ₂, a ₂..., i _n-1, a _n-1, i _n) be the conditional probability measure of A, the moment exhausting with first node energy is measured life cycle, so L life cycle of whole network can arrive halted state S by network _tall return sums are described before, definition L _u(i) for network is from state i, the life cycle of usage policy u, i.e. all return sums;

(3) solve optimal policy

From state i, meet the maximum network lifetime L of constraints ^*(i) be the unique solution of the optimum equation of following Bellman,

s.t.f _j≥f _n

F wherein _jthe fair coefficient of network under state j, f _nit is the fair threshold value providing according to different application scenarioss;

In fact, formula above can also be write

The optimal policy u of transmitting and scheduling scheme can be drawn by following formula

(4) reduce computation complexity

An optimum equation of equivalent Bellman is write as following form:

So constraint optimal policies also can be obtained by following formula

In implementation procedure, each time slot at first, beacon signal of AP node broadcasts is waken all the sensors node up, in order to make AP node obtain global channel information, all sensor nodes all need to send the beacon signal that a pilot signal is replied AP, the signal that the utilization of AP node receives is estimated the channel conditions of all nodes and is obtained the energy w that transmission needs, then according to state (e, w, f) calculate optimal policy u, clear and definite optimal policy, known the sensor node that should select at each time slot, finally, No. ID of the selected sensor node of AP node broadcasts, this sensor node starts to transmit the data that it collects, because AP node is known the channel conditions of all the sensors node, so can continue to upgrade network state and be the arrival of next time slot prepares.

2. a kind of maximization network transmission dispatching method life cycle that meets fairness condition according to claim 1, is characterized in that, described fair coefficient f is defined as follows: definition (T ₁, T ₂..., T _n) representing the number of transmissions of each sensor node reality, O represents the total degree having transmitted, (b ₁, b ₂..., b _n) mean the weights of each sensor node importance.For sensor node n, define standardized the number of transmissions and be

x _n＝T _n/b _nO

Suppose to have N user's competition network resource, the resource that nth user obtains is x _n, the fair coefficient of define grid is so

F is value between 0 and 1.If the fair coefficient f=1 of network, represents that network is completely fair.The fair coefficient of network is larger, and fair degree is higher, and vice versa.