CN102209369B - Method based on wireless network interface selection to improve a smart phone user experience - Google Patents

Method based on wireless network interface selection to improve a smart phone user experience Download PDF

Info

Publication number
CN102209369B
CN102209369B CN201110131397.4A CN201110131397A CN102209369B CN 102209369 B CN102209369 B CN 102209369B CN 201110131397 A CN201110131397 A CN 201110131397A CN 102209369 B CN102209369 B CN 102209369B
Authority
CN
China
Prior art keywords
node
probability
time
total
user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN201110131397.4A
Other languages
Chinese (zh)
Other versions
CN102209369A (en
Inventor
牛建伟
高宇航
童超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beihang University
Original Assignee
Beihang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beihang University filed Critical Beihang University
Priority to CN201110131397.4A priority Critical patent/CN102209369B/en
Publication of CN102209369A publication Critical patent/CN102209369A/en
Application granted granted Critical
Publication of CN102209369B publication Critical patent/CN102209369B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention brings forward a method based on a wireless network interface selection to improve a smart phone user experience. According to the invention, in view of the vehicle-mounted mobile situation, an optical user experience is obtained by dynamically distributing a wireless network interface to an application node with data independence. Two solutions are brought forward in the invention. One solution is as follows: enumerating all probability execution events of an application node by carrying out a depth-first search on a solution space; carrying out pruning based on constraints; and selecting a wireless network interface for each node to obtain an optical user experience; the other solution is as follows: decomposing a problem into a problem of a smaller subgraph by a dynamic programming; distributing and determining a probability event for each node having a plurality of father nodes to avoid a conflict in selecting a wireless network interface; deleting a suboptimal solution and an infeasible solution by using a redundant node concept to obtain an optical user experience and a selection of a wireless network interface of each node. According to the invention, a good user experience can be obtained; moreover, the user can set an experience parameter by himself/herself, so that more freedom and selection space are provided for the user.

Description

The method that the enhancing smart phone user of selecting based on radio network interface is experienced
Technical field
The invention belongs to the communications field, be specifically related to a kind of method that enhancing smart phone user of selecting based on radio network interface is experienced.
Background technology
In the information age, every smart mobile phone is all a hand-held mobile terminal that constantly receives and send data.Meanwhile, mobile subscriber expects to obtain and take the optimal user that low delay, cheap any wireless network services and low-power consumption be feature and experience.But in actual conditions, the performance of smart mobile phone is subject to the impact of the many factors such as wireless signal decay, power management policy and environmental condition.In order to improve user, experience and requirement of real time, existing research is emphasized to realize from interface Adaptive Technology by wireless network: by detecting current wireless network environment or predicting current network state according to wireless data in the past, be different mobile application Dynamic Selection radio network interfaces.
The mobile device of considering most has all been equipped with a plurality of network interfaces, Wi-Fi for example, Bluetooth, GPRS, 3G etc., and each network interface has different characteristics, for example GPRS is low in energy consumption but transmission rate is low, Wi-Fi power consumption is large but higher transmission rate can be provided, network power consumption accounts for a large portion (>=30%) of electric quantity consumption of mobile phone simultaneously, and researchers propose to fully utilize a plurality of network interfaces and to reduce user, use the electric quantity consumption of network service.Trevor Pering, the people such as Yuvraj Agarwal propose Coolspots, can list of references [T.Pering, Y.Agarwal, R.Gupta, and R.Want, " Coolspots:reducingthe power consumption of wireless mobile devices with multiple radio interfaces; " in Proceedings of the 4th international conference on Mobile systems, applications and services, 2006, pp.220-232].Coolspots can automatically for example switch according to current network conditions and application demand mobile device in a plurality of network interfaces between Wi-Fi and bluetooth, when communication data amount little, when bandwidth requirement is low, start the bluetooth of low-power consumption and close Wi-Fi interface to reduce electric quantity consumption, when communication data amount large, when bandwidth requirement is high, be switched to fast the Wi-Fi of two-forty to meet communication requirement.But Coolspots has special requirement to hardware, be not suitable for promoting.Footprint is the related data of utilizing equipment to obtain from cellular network, as the ID of the cell tower listening to, signal strength signal intensity etc. are come the current position of supposition equipment, only have and when larger change occurs for the position of equipment, just start Wi-Fi and scan to obtain available AP (Access Point around, access points) information, reduces power consumption by reducing unnecessary scanning times.Footprint can list of references [H.Wu, K.Tan, J.Liu, and Y.Zhang, " Footprint:cellular assisted wi-fi ap discovery on mobile phones for energy saving; " in Proceedings of the 4th ACM international workshop on experimental evaluation and characterization (WINTECH ' 09), 2009, pp.67-76].
One or two in this three aspects: of task execution time, rate and power consumption only considered in existing research mostly, for example under time constraint, guarantee that power consumption is minimum, and reality is obtaining when good user experiences and task execution time, power consumption and rate three comprehensively need to considered.Meanwhile, existing research becomes the uncertainty of mobile application time of implementation that wireless network and conditional order cause etc. while having ignored, only considered the optimisation strategy under worst case.In addition, in vehicle-mounted mobile situation, user need to complete the mobile application that a series of data are relevant, uses 3G wireless network that better transmission speed and less packet loss can be provided, but have higher power consumption and rate compared to WiFi in existing research.
Summary of the invention
The present invention is directed in vehicle-mounted mobile situation, how to pass through dynamic assignment wireless network (3G and WiFi) interface, make time the wireless network that becomes in, with less rate and power consumption, response user request fast, thereby be the problem that user provides best user to experience, proposed a kind of method that enhancing smart phone user of selecting based on radio network interface is experienced.
The inventive method is applicable in vehicle-mounted mobile situation, the application scenarios of the intelligent movable mobile phone of one group of data dependence, first by directed acyclic graph, be that described application scenarios is set up user and experienced model: G=<V, E>, wherein, G is the directed acyclic graph for representing that user experiences, V=<v 1..., v i... v n> is the set of node in directed acyclic graph, each node v irepresent a mobile application, N is natural number, and N represents total number of mobile application in described application scenarios, be the set on limit in directed acyclic graph, every limit represents there is connection between two nodes with the directed line segment with arrow, and representing has data dependence relation between two mobile nodes.
The method that the present invention is based on the enhancing smart phone user experience of radio network interface selection specifically comprises step below:
Step 1, directed acyclic graph G is carried out to topological sorting, obtain a current collating sequence.
Step 2, according to the probability of rate
Figure BDA0000062517830000022
the probability of time of implementation
Figure BDA0000062517830000023
probability with power consumption
Figure BDA0000062517830000024
determine each node v iall probability carry out event P i, n.
Step 3, optimal user experience UX is set minfor infinity, total time of implementation T is set t(G) be 0.
Step 4, enter the 1st layer of solution space, initialization current active node A 1middle information is empty.
The level of the solution space that step 5, basis enter, finds the present node v in G k, establish as the l layer that advances into solution space, present node v kfor l node in the current collating sequence of G.
Step 6, enumerate present node v kall probability carry out events.
Step 7, judge whether to select all probability of present node to carry out events, if not, execution step 8; If so, perform step 14.
Step 8, select present node v successively kprobability carry out event.
Step 9, renewal present node v kexecution end time f k, described execution end time f kfor node v kstarting Executing Time and node v kexecution required time and.
Step 10, upgrade total time of implementation T t(G): judgement present node v kexecution end time f kwhether be greater than total time of implementation T t(G), if so, total time of implementation T is set so t(G) be this present node v kexecution end time f k, otherwise, keep total time of implementation T t(G) constant.
Step 11, according to the probability of selecting, carry out event update current active node A linformation, comprising: upgrade root node from current collating sequence to present node v kthe probability of having done is carried out the selection of event; Upgrading the total time of implementation producing is T t(G);
Upgrade the total rate H producing t(G):
Figure BDA0000062517830000031
h ithe rate that represent the probability execution event that i node of current collating sequence selected;
Upgrade the total power consumption C producing t(G):
Figure BDA0000062517830000032
c ithe power consumption that represents the probability execution event that i node of current collating sequence selected;
Upgrade total confidence probability P t(G):
Figure BDA0000062517830000033
p ithe probability that represents the probability execution event that i node of current collating sequence selected.
Step 12, judgement current active node A lwhether the total time of implementation in information, total rate, total power consumption and total confidence probability, meet time, rate, power constraints <T, H, and C> and confidence probability θ, if meet, execution step 13, otherwise go to step 7 execution.
Step 13, renewal optimal user experience UX min: the user who first determines current active node experiences UX t(G):
UX t(G)＝α*H t(G)+β*T t(G)+γ*C t(G)
Wherein, experience parameter alpha, β and γ and meet alpha+beta+γ=1;
Then judge whether to satisfy condition: pux (UX t, P t)≤pux (UX min, P min), if so, upgrade optimal user experience UX minvalue be UX t(G), if not, keep optimal user experience UX minin value constant.Wherein, pux (UX t, P t) represent to experience UX for user t(G), make the user of figure G experience UX (G) and meet P (UX (G) > UX t(G)) < 1-P t, pux (UX min, P min) represent for current optimal user experience UX min, make the user of figure G experience UX (G) and meet P (UX (G) > UX min(G)) < 1-P minprobability.
Step 14, renewal depend on node v kthe Starting Executing Time of each node of data.
Whether lower one deck (l+1) layer of the current level of the solution space that step 15, judgement enter exists, and if not, execution step 16, if so, enters (l+1) layer of solution space, and by active node A lin information be assigned to current active node A l+1, then go to step 5 execution.
Step 16, by active node A nwith optimal user experience UX minoutput, method ends.
The method that the present invention is based on the enhancing smart phone user experience of radio network interface selection, can also realize by Dynamic Programming, specifically comprises the following steps:
Step 1: directed acyclic graph G is carried out to topological sorting, obtain topological sequences.
Step 2: there is the interstitial content of a plurality of father nodes in statistics G, be designated as t mp.
Step 3: there is the interstitial content of a plurality of child nodes in statistics G, be designated as t mc.
Step 4: if t mp<=t mc, upset topological sequences, then carries out next step, otherwise, directly carry out next step.
Step 5: according to the probability of rate
Figure BDA0000062517830000041
the probability of time of implementation
Figure BDA0000062517830000042
probability with power consumption
Figure BDA0000062517830000043
determine each node v iall probability carry out event P i, n.
Step 6: for t mcthe individual node with a plurality of father nodes, enumerates this t mcthe combination of all possible radio network interface of individual node, rejects the combination that occurs radio network interface selection conflict.
Step 7: the every kind of radio network interface successively step 6 being obtained be combined into line operate: successively to each the node v in topological sequences l, 1≤l≤N, determines the optimal solution of this node, l represents node v lthe l position that is arranged in topological sequences:
(1) determine the solution S of the first mobile application node in topological sorting 1, j, k: S 1, j, k=P 1, n;
(2) determine the 2nd node to the N node v in topological sorting msolution S m, j, k, 2≤m≤N:
First, according to
Figure BDA0000062517830000044
obtain with node v mfor the clique of root node is except node v mouter optimal solution S ' m, j, k, wherein,
Figure BDA0000062517830000045
respectively node v msubgraph
Figure BDA0000062517830000046
optimal solution,
Figure BDA0000062517830000047
respectively with node
Figure BDA0000062517830000048
for the clique of root node,
Figure BDA0000062517830000049
node v mchild node; Then travel through node v mall probability carry out event P m, n, by: S m, j, k=S ' m, j, k⊙ P m, n, obtain all with node v moptimal solution S for the clique of root node m, j, k; Reject optimal solution S n, j, kin redundant node.Finally obtain N node v noptimal solution S n, j, k.
Step 8: to each the optimal solution S obtaining n, j, kdetermine that corresponding user experiences, and then finds minimum probability optimal user experience pus (UX t, P t), described pus (UX t, P t) be to experience UX for user t(G), make to meet P (UX (G) > UXtG < 1-Pt, this minimum pux (UX t, P t) user to experience be exactly optimal user experience UX min(G), finally by the probability execution event output that in optimal user experience and the corresponding optimal solution of this optimal user experience, each node is selected, method ends.
Advantage of the present invention and good effect are: the inventive method is first in vehicle-mounted mobile situation, mobile phone application to one group of data dependence, by dynamic assignment wireless network (3G and WiFi) interface, make time the wireless network that becomes in, with less rate and power consumption, response user request, has higher efficiency fast, and provides better user to experience; And user can arrange experience parameter as required voluntarily, give user more free and selection space.
Accompanying drawing explanation
Fig. 1 is that the present invention strengthens the flow chart of steps of the method for smart phone user experience by depth-first search;
Fig. 2 is a DAG exemplary plot;
Fig. 3 is the schematic diagram of node from i step to the transformation of i+1 in dynamic programming method;
Fig. 4 is that the present invention strengthens the flow chart of steps of the method for smart phone user experience by Dynamic Programming;
Fig. 5 is the result comparison diagram that linear programming method and two kinds of methods of use the present invention obtain.
Concrete enforcement
Below in conjunction with drawings and Examples, content of the present invention is described in further detail.
The inventive method is applicable in vehicle-mounted mobile situation, the application scenarios of the intelligent movable mobile phone of one group of data dependence, by dynamic assignment radio network interface make time the wireless network that becomes in, with less rate and power consumption, response user request fast, to providing best user to experience for user.The mobile application of one group of data dependence in application scenarios, available DAG (Directed Acyclic Graph, directed acyclic graph) show, the problem that then practical application scene is strengthened to smart phone user experience is converted to the time of implementation that how to minimize mobile application process in whole DAG.Whether definite according to the time of implementation of each application process in DAG, application process is divided into the application process of deterministic schedule and the application process of uncertainty scheduling.For deterministic schedule problem, the List scheduling of the dispatching method of existing employing based on classical, such as HLFET, ETF and LAST algorithm etc.Mobile application in the application scenarios the present invention is directed to belongs to the application process of uncertainty scheduling, problem for uncertainty scheduling, consider conditional order and time become the uncertain of mobile application time of implementation that wireless network causes, some researchers have proposed schedule by probability mode.The time of implementation that schedule by probability mode moves application by each is considered as stochastic variable.The people such as Ku and De Micheli have proposed without the schedule by probability mode in the situation of delay lower bound, can be with reference to [W.Qadeer, T.Rosing, J.Ankcorn, V.Krishnan, and G.De Micheli, " Heterogeneous wireless network management; " in Power-Aware Computer Systems, vol.3164, pp.137-184.2005].
The present invention sets up user in application scenarios by DAG, and to experience model as follows:
Set up a DAG:G=<V, E>, V=<v 1..., v i... v n> is the set of node in figure, each node v irepresent a mobile application, N is natural number. the set on limit in directed acyclic graph, e i, j← 0 representation node v iand v jbetween there is no data dependence relation, e i, j← 1 representation node v iand v jbetween have data dependence relation.W=<w 1..., w i..., w m> is one group of radio network interface set.
Figure BDA0000062517830000052
with
Figure BDA0000062517830000053
be respectively used to represent node v iat radio network interface, be w jsituation under, the probability density distribution of rate, time of implementation and power consumption.
Due to conditional order and time become the uncertainty of wireless network, the present invention introduces the concept that probability is carried out event: each probability is carried out event P m, na four-tuple <p m, n, c m, n, t m, n, h m, n>, this probability is carried out representations of events node v mto be more than or equal to p m, nprobability, consumed power consumption c m, n, time of implementation t m, nwith rate h m, nresource complete this node v mrepresented mobile application.P m, nrepresentation node v mn probability carry out event, each node has a plurality of probability to carry out events.Described rate refer to the flow charging of selecting certain radio network interface to produce, and described power consumption refers to the electric quantity consumption of selecting certain radio network interface to produce.
For a directed acyclic graph G, in G, the probability of each node is carried out, with a five-tuple <s i, x i, h i, p i, c i> represents, wherein, and s irepresent node v istarting Executing Time, x irepresent node v iexecution required time, h irepresent node v irate, c irepresent XM v ipower consumption, p irepresent node v iconfidence probability.Described confidence probability refers under certain rate, time, power consumption situation, the probability that node can complete.At this node v iprobability selection event determine after, p ibe exactly the p of corresponding probability selection event m, n, x ibe exactly the t of probability selection event m, n, h ibe exactly the h of corresponding probability selection event m, n, c ibe exactly the c of corresponding probability selection event m, n, determine probability selection event, also just determined radio network interface, the Starting Executing Time s of each node ialso determined.So, the total time of implementation T of directed acyclic graph G t(G), total power consumption C t(G), total rate H t(G), total confidence probability P t(G), user experiences UX t(G) by following formula, can obtain:
T t(G)＝max{s i+x i} &ForAll; v i &Element; V - - - ( 1 )
C t ( G ) = &Sigma; i = 1 N c i - - - ( 2 )
H t ( G ) = &Sigma; i = 1 N h i - - - ( 3 )
P t ( G ) = &Pi; i = 1 N p i - - - ( 4 )
UX t(G)＝α*H t(G)+β*T t(G)+γ*C t(G)，α+β+γ＝1 (5)
When formula (5), used α, β, tri-proportionality coefficients of γ, usage ratio coefficient shows mobile subscriber to experience the formation understanding of model on the one hand, the namely selection of network interface directly has influence on mobile application and carries out rate, power consumption and the time of implementation producing, this three aspects: joint effect mobile subscriber's experience; Because everyone is different to the impression of power consumption, rate and response time and requirement, adoption rate coefficient provides more choices to mobile subscriber and is free on the other hand.
Due to the present invention, introduced the concept of schedule by probability mode and calculated user's experience, defined probability user experience at this by a data flow diagram DFG, described DFG is also a directed acyclic graph:
Set up a DFG:F=<V, E>, can obtain node v by data mining iat radio network interface, be w jsituation under the probability of rate, time of implementation and power consumption
Figure BDA0000062517830000065
with
Figure BDA0000062517830000066
a given confidence probability θ, probability user experiences pux (u, θ) for representing, for user, experience u, total confidence probability constraints condition P (UX (G) > u) < 1-θ is set up, the user of UX (G) presentation graphs G experiences.The user of whole figure G experience u and this user experience total confidence probability description that u is corresponding the probability user of figure G experience, two probability users are experienced and compared, be exactly essence: total the confidence probability of two probability user experience all meets under the prerequisite of described total confidence probability constraints condition, the value that two users are experienced compares.
Through definition above, problem to be solved by this invention is converted to NIASP (Network Interface Assignment and Scheduling with Probability) problem: a given radio network interface collection W, a DAG:G=<V, E>, each node v iat w junder the probability of rate, time of implementation and power consumption with
Figure BDA0000062517830000068
time-constrain T, power constraints C, rate constraint H and confidence probability θ, require as each node distributes suitable radio network interface, makes meeting under the condition of each item constraint, is being more than or equal under the confidence probability of θ, and the user that offers the best experiences UX *.Solution space for NIASP problem is a search tree, therefore can use the depth first method search optimal solution based on depth-first or breadth First.
The method that the present invention adopts the enhancing smart phone user of selecting based on radio network interface of depth first method search to experience, experience after model setting up user by directed acyclic graph, specifically the step by as shown in Figure 1, for each node distributes suitable radio network interface, obtains best user and experiences:
Step 1: for the mobile application DAG:G=<V of one group of data dependence, E> carries out topological sorting, obtains a current collating sequence.
Be illustrated in figure 1 one embodiment of the present of invention, V=<v in this DAG 1, v 2, v 3, v 4, v 5, v 6>, represents to have 6 mobile application nodes, if exist data dependence to have the arrow of line to represent between node, does not exist data dependence not have.Carry out after topological sorting, select current collating sequence a: v 1, v 2, v 3, v 4, v 5, v 6, carry out step below.First node in collating sequence is called root node.In the embodiment of the present invention, root node is v 1.
Step 2: for each the node v in set up DAG i,
Figure BDA0000062517830000071
according to the probability of rate, time of implementation and power consumption
Figure BDA0000062517830000072
with
Figure BDA0000062517830000073
determine each node v iall probability carry out event P i, n, P i, nrepresent node v in probability carry out.The probability distribution of rate, time of implementation and power consumption
Figure BDA0000062517830000074
with
Figure BDA0000062517830000075
can determine by the method for data mining.P i, nin power consumption, time of implementation and rate c m, n, t m, n, h m, ncan be respectively by calculating rate, time of implementation and power consumption
Figure BDA0000062517830000076
with
Figure BDA0000062517830000077
joint probability distribution in key point is tried to achieve.
Step 3: initialization optimal user experience UX minfor infinity, total time of implementation T is set t(G) be 0.
Described optimal user experience UX minfor global variable, and compose a higher value, such as being infinitely great, when calculating when better user experiences, it is replaced.
Step 4: enter the 1st layer of solution space, initialization current active node A 1middle information is empty.
Described active node is the node on the search tree of solution space, the total N layer of search tree one of solution space, and N represents the number of node in DAG, the level that is solution space hereinafter referred to as the level of the search tree of solution space.A lbe illustrated in the active node of solution space l layer.Active node A lin information include: the root node from the current collating sequence of G is to present node v kthe probability of having done is carried out EventSelect, and the total rate, power consumption, time of implementation and the probabilistic information that produce.And the rate, power consumption, time of implementation and the probabilistic information that comprise when active node have been violated time-constrain T, power constraints C, rate constraint H and confidence probability P, this active node becomes and dies for the sake of honour a little.Present node v krefer to l node in the current collating sequence of G.
Step 5: according to the level of the solution space entering, find the present node v in G k, establish as the l layer that advances into solution space, present node v kfor l node in the current collating sequence of G.
For example, in Fig. 2, if the level l of the solution space entering is 2 o'clock, namely current active node is A 2,, in the current collating sequence of corresponding embodiment of the present invention G, present node is v 2.When if the solution space entering is the 1st layer of l=1, current active node is A 1, in the current collating sequence of corresponding embodiment of the present invention G, present node is v 1.
Step 6: enumerate present node v kall probability carry out events.For present node v keach probability carry out event by the step of carrying out below.
Step 7: judge whether to select all probability of present node to carry out event, if not, execution step 8; If so, perform step 14.
Step 8: select successively present node v kprobability carry out event.
Step 9: upgrade present node v kexecution end time f k.Described execution end time f kstarting Executing Time s for node kthe execution required time x that adds node k.
Step 10: upgrade total time of implementation T t(G).Judgement present node v kexecution end time f kwhether be greater than total time of implementation T t(G), if so, total time of implementation T is set so t(G) be this present node v kexecution end time f k, otherwise, keep total time of implementation T t(G) constant.
Step 11: carry out event update current active node A according to the probability of selecting linformation.
The root node of renewal from current collating sequence is to present node v kthe probability of having done is carried out the selection of event.
Upgrading the total time of implementation producing is T t(G).
Upgrade the total rate H producing t(G):
Figure BDA0000062517830000081
h ithe rate that represent the probability execution event that i node of current collating sequence selected.
Upgrade the total power consumption C producing t(G):
Figure BDA0000062517830000082
c ithe power consumption that represents the probability execution event that i node of current collating sequence selected.
Upgrade total confidence probability P t(G):
Figure BDA0000062517830000083
p ithe probability that represents the probability execution event that i node of current collating sequence selected.
Step 12: if the total time of implementation obtaining, total rate, total power consumption and total confidence probability, can not meet time, rate, power constraints <T, H, C> and confidence probability θ, go to step 7 execution, otherwise carry out next step.
Step 13: upgrade optimal user experience UX min, then go to step 7 execution.According to the content through type (5) of current active node, obtain user and experience UX t(G): UX t(G)=α * H t(G)+β * T t(G)+γ * C t(G), alpha+beta+γ=1; Further obtain again probability user and experience pux (UX t, P t), judge whether to satisfy condition: pux (UX t, P t)≤pux (UX min, P min), if so, by UX t(G) assignment is to global variable UX min, if not, keep global variable UX minin value constant.Probability user experiences pux (UX t, P t) represent to experience UX for user t(G) P (UX (G) the > UX that makes to satisfy condition t(G)) < 1-P t, pux (UX min, P min) represent current optimal user experience UX minp (UX (G) > UX makes to satisfy condition min(G)) < 1-P min, wherein the user of UX (G) presentation graphs G experiences.
Step 14: upgrade and depend on node v kthe Starting Executing Time of each node of data.The Starting Executing Time of each node is initially set to 0, when the Starting Executing Time of new node more, by the Starting Executing Time comparison of current time and this node, if current time is large, the Starting Executing Time that upgrades this node is current time, otherwise it is constant to preserve the Starting Executing Time of this node.
Step 15: judgement, when whether lower one deck (l+1) layer of the solution space level advancing into exists, if not, performs step 16, if so, enters (l+1) layer of solution space, and by active node A lin information be assigned to current active node A l+1, then go to step 5 execution.
Step 16: by active node A nwith optimal user experience UX minoutput, method ends.Active node A nin the information that comprises have root node from the current collating sequence of G to present node v nthe probability done is carried out EventSelect, carries out event can obtain the selected radio network interface of this node from the probability of each node of selecting.
The present invention also proposes to use dynamic programming method to solve NIASP problem.
Before explanation dynamic programming method, first provide several definition.G ibe defined as with node v iclique for root node.T a(G i), H a(G i) and C a(G i) represent at node v iradio network interface be assigned as situation figure below G of A itotal time of implementation, total rate and total power consumption.All possible T a(G i) time set T of composition i.
Dynamic Programming, by the subproblem of NIASP PROBLEM DECOMPOSITION Cheng Geng little, completes solving of whole problem by bottom-up solution subproblem.NIASP has optimum minor structure and overlapping subproblem.On the one hand, the optimal solution of whole problem
Figure BDA0000062517830000091
depend on the optimal solution of subgraph
Figure BDA0000062517830000092
node v ichild node,
Figure BDA0000062517830000093
respectively with node
Figure BDA0000062517830000094
for the clique of root node, it is respectively subgraph
Figure BDA0000062517830000096
optimal solution.On the other hand, the existence of many father nodes has caused overlapping subproblem, for example, supposes v ithere are a plurality of father nodes, if record
Figure BDA0000062517830000097
the solution of subproblem, can avoid the double counting of subproblem.
According to the character of optimum minor structure and overlapping subproblem, can obtain following relation:
S &OverBar; i * = P i , n , w = 0 S i 1 * &CirclePlus; S i 2 * . . . S iw * , , w &GreaterEqual; 1 - - - ( 6 )
Figure BDA0000062517830000099
be illustrated in node v ifor removing node v in the clique of root node iouter optimal solution, w is node v ichild node number.
Figure BDA00000625178300000910
be illustrated in node v ifor the optimal solution of the clique of root node,
Figure BDA00000625178300000911
the S arranging from low to high according to confidence probability i, j, knode.S i, j, knode definition is as follows: c i, j, kat all C a(G i) middle minimum power consumption, meet T simultaneously a(G i)≤j, H a(G i)≤k and confidence probability are more than or equal to p i, j, k; J represents total time of implementation, and k represents total rate.Symbol
Figure BDA00000625178300000912
for the solution of subproblem being combined to the more massive problem that solves.Suppose node v ithere are two child nodes
Figure BDA00000625178300000913
with
Figure BDA00000625178300000914
for with node
Figure BDA00000625178300000915
clique for root node
Figure BDA00000625178300000916
optimal solution, represent with total time of implementation j 1, total rate k 1, total power consumption complete figure
Figure BDA00000625178300000918
probability be
Figure BDA00000625178300000919
for with node
Figure BDA00000625178300000920
clique for root node
Figure BDA00000625178300000921
optimal solution, represent with total time of implementation j 2, total rate k 2, total power consumption complete figure
Figure BDA00000625178300000923
probability be
Figure BDA00000625178300000924
in process
Figure BDA00000625178300000925
after computing,
Figure BDA00000625178300000926
obtain j=max{j 1, j 2, k=k 1+ k 2,
Figure BDA00000625178300000927
Figure BDA00000625178300000928
But, the overlapping problem that has caused equally wireless network to select conflict of subproblem, for example, certain node v kbelong to subgraph G simultaneously iand G j.Likely solving G iduring subproblem, be v kselect 3G; And solving G jtime, be v kselect WiFi.Therefore, even subproblem G iand G jall found optimal solution, can failure when merging its result.In order to address this problem, by enumerating all probability of many father nodes, carry out event, by setting to avoid this problem for the radio network interface of certain many father node.
This dynamic programming method has N step.Transformation from from i step to i+1 step is the process of a larger subproblem of solution.Fig. 3 has described the feasible solution S in i step i, j, kwith at i+1, walk possible succession, as shown in Figure 3, through the 1st step, to i-1 step, (stage 1 th... (i-1) th) obtaining some optimal solution of subproblem, in figure, circle represents node, in i step, has 4 nodes, each node to represent a G isolution, the circle of black represents feasible solution, the circle of white represents infeasible solution.In the 1st step, because the first node getting from topological sorting does not have child node, G 1feasible solution be equal to and may carry out chained list P 1, n.In order to obtain S i+1, j, k, need to be first by merging subgraph optimal solution obtain S ' i, j, k.The new S ' obtaining i, j, kmay not be feasible solution, according to each item constraint, judge and see whether it is feasible solution.It is optimum that each node in i ' step (stage i ') travels through to determine which probability is carried out event successively.In order to reach this object, probability, power consumption, rate and the time of implementation of each probability of the node in i+1 step being carried out to event join in the node of i ' step, select to meet confidence probability and provide the node that optimal user is experienced, and enter next step.The process of this selection is completed by operation ⊙, and in other words, ⊙ is the transformation from i ' step to i+1 for the treatment of node.
The present invention is based on dynamic programming method, for each node distributes suitable radio network interface to strengthen smart phone user, experience as shown in Figure 5, comprise the following steps.
Step 1: directed acyclic graph G is carried out to topological sorting, obtain topological sequences Seq:v 1..., v n, N is the number of node in directed acyclic graph G.In the embodiment of Fig. 1, a Seq who obtains is: v 1, v 2, v 3, v 4, v 5, v 6.
Step 2: statistics has the interstitial content of a plurality of father nodes, is designated as t mp.Input data source is designated as t in the mobile application numbers of a plurality of mobile application in the present invention mp, in embodiment as shown in Figure 2, t mpbe 2.
Step 3: statistics has the interstitial content of a plurality of child nodes, is designated as t mc.The data of output are that the mobile application numbers that a plurality of mobile application are used is designated as t in the present invention mc, in embodiment as shown in Figure 2, t mpbe 2.
Step 4: if t mp<=t mc, upset topological sequences, then carries out next step, otherwise, directly carry out next step.As shown in Figure 2, t mpequal t mp, the topological sequences that therefore overturns, the Seq obtaining is: v 6, v 5, v 4, v 3, v 2, v 1.
Step 5: move application node v for each i,
Figure BDA0000062517830000102
according to the probability of rate, time of implementation and power consumption
Figure BDA0000062517830000103
with
Figure BDA0000062517830000104
determine each node v iall probability carry out event P i, n, P i, nrepresent node v in probability carry out.
Step 6: for t mcthe individual node with a plurality of father nodes, enumerates this t mcthe combination of all possible radio network interface of individual node, selects a radio network interface to avoid occurring the situation of wireless network selection conflict by have the node of a plurality of father nodes for certain.
As Fig. 2, node v 5be node v simultaneously 2with node v 3subgraph in node.Likely solving node v 2subgraph time, be node v 5choice for use WiFi, and solve node v 3subgraph problem time, be node v 5select 3G.Like this at merge node v 2with node v 3optimal solution time, can produce infinite network conflict.Wireless network is selected to conflict situations, and can pass through is many father nodes v in advance 5select a radio network interface, avoid.
Step 7: the every kind of radio network interface successively step 6 being obtained be combined into line operate: successively to each the node v in topological sequences l, 1≤l≤N, determines the optimal solution of this node, l represents node v lit is the l position that is arranged in topological sequences.
(1), for the first mobile application node in topological sorting, determine the solution S of this node 1, j, k: S 1, j, k=P 1, n.
(2) for the 2nd node to the N node v in topological sorting l, 2≤l≤N:
The first step, basis
Figure BDA0000062517830000105
obtain with node v lfor removing node v in the clique of root node louter optimal solution S ' l, j, k, wherein, respectively node v lsubgraph
Figure BDA0000062517830000112
optimal solution,
Figure BDA0000062517830000113
respectively with node for the clique of root node,
Figure BDA0000062517830000115
node v lchild node.
By
Figure BDA0000062517830000116
computing obtains with node v lfor removing node v in the clique of root node louter total time of implementation j, total rate k, total power consumption c l, j, kwith total confidence Probability p l, j, k:
j＝max{j 1，...，j w}， k = &Sigma; i = 1 w k i , c l , j , k = &Sigma; m = 1 w c l m , j m , k m , p l , j , k = &Pi; m = 1 w p l m , j m , k m ;
Wherein, w represents node v lchild node number, j wrepresented node v lchild node
Figure BDA00000625178300001110
total time of implementation, k mrepresented node
Figure BDA00000625178300001111
total rate, represented node minimum total power consumption,
Figure BDA00000625178300001114
represented node
Figure BDA00000625178300001115
total confidence probability.
Second step, traversal node v lall probability carry out event P l, n, by: S l, j, k=S ' l, j, k⊙ P l, n, obtain all with node v loptimal solution S for the clique of root node l, j, k.For node v lprobability carry out event P l, n, node v lto carry out required time x l, rate h l, power consumption c lthe probability completing is p l, by ⊙ computing, obtain with node v lfor total time of implementation j ' of the clique of root node, total rate k ', total power consumption c ' l, j, kwith total confidence Probability p ' l, j, k:
j′＝j+x l，k′＝k+h l，c′ l，j，k＝c l，j，k+c l，p′ l，j，k＝p l，j，k·p l；
Wherein, j, k, c l, j, kand p l, j, krepresent respectively with node v lfor removing node v in the clique of root node louter optimal solution S ' l, j, ktotal time of implementation, total rate, total power consumption and total confidence probability.
The 3rd step, rejecting optimal solution S l, j, kin redundant node.
Remove optimal solution S n, j, kin redundant node.Described redundant node refers to: for node v mtwo kinds of optimal solution
Figure BDA00000625178300001116
with represent node v mwith power consumption
Figure BDA00000625178300001118
rate k 1the probability completing is
Figure BDA00000625178300001119
represent node v mwith power consumption
Figure BDA00000625178300001120
rate k 2the probability completing is
Figure BDA00000625178300001121
if
Figure BDA00000625178300001123
and k 1≤ k 2, so for redundant node.
N the node v that the 4th step, preservation obtain nall optimal solution S n, j, k.
For example, for Fig. 2, node v 6for the first mobile application that topological sorting obtains, node v 6there is no child node, its solution is node v 6probability carry out.And when carrying out to the 5th step, obtained with node v 2, v 3and v 4solution for the clique of root node, now will solve whole figure, first need by
Figure BDA00000625178300001125
will be respectively with node v 2, v 3and v 4for the solution of the clique of root node is combined, this just completes node from the 5th step to the 5th ' transformation.Then enumerate node v 1all probability carry out events, by operation, ⊙ completes from the 5th ' execution of step to the 6 steps.According to the concept of the redundant node of definition, delete redundant node, select keeping optimization simultaneously.
The available following coded representation of operation of carrying out in step 7:
Figure BDA0000062517830000121
In above-mentioned code, K represents the combination of radio network interface,
Figure BDA0000062517830000122
represent to distribute in the situation of certain radio network interface with node
Figure BDA0000062517830000123
clique for root node
Figure BDA0000062517830000124
the time set of total time of implementation,
Figure BDA0000062517830000125
represent to distribute in the situation of certain radio network interface with node
Figure BDA0000062517830000126
clique for root node
Figure BDA0000062517830000127
the time set of total time of implementation, k 1..., k wrepresent respectively clique
Figure BDA0000062517830000128
total rate.
Step 8: can access some optimal solution of directed acyclic graph G by step 7, then to each the optimal solution S obtaining n, j, k, according to formula (5), determine that corresponding user experiences UX t(G), then further determine minimum probability optimal user experience pux (UX t, P t), this minimum pux (UX t, P t) in user to experience be exactly optimal user experience UX min(G).Finally by optimal user experience UX minwith the probability execution event output that each node of optimal solution corresponding to this optimal user experience is selected, method ends.Each moves the selected radio network interface of application node and can from the probability execution event of corresponding selection, obtain.
Pux (UX t, P t) represent that user experiences UX t(G), make to meet P (UX (G) > UX t(G)) < 1-P tcondition, the user of UX (G) presentation graphs G experiences.
The present invention above-mentioned by depth-first search with by two kinds of Dynamic Programmings, for the result of the method that strengthens smart phone user and experience, be consistent.The comparison diagram of the user experience value obtaining for application the present invention two kinds of methods and application linear programming method as shown in Figure 5.Linear programming (Integer Linear Programming, brief note is for ILP) problem researched and solved of method is that requirement variable is while rounding numerical value, in the next linear function optimal problem of one group of Linear Constraints, to apply an important branch of operational research very widely, when using this linear programming method to solve NIASP problem of the present invention, the suboptimal solution often obtaining.As can be seen from Figure 5,, in the situation that confidence probability θ is 0.7,0.8,0.9 and 1.0, the user that linear programming method obtains than the inventive method experiences and differs from 30.3%, 25.9%, 19.8% and 10.8%.

Claims (2)

1. the method that the enhancing smart phone user of selecting based on wave point is experienced, it is characterized in that, described method is applicable in vehicle-mounted mobile situation, the application scenarios of the intelligent movable mobile phone of one group of data dependence, first by directed acyclic graph, be that described application scenarios is set up user and experienced model: G=< V, E >, wherein, G is the directed acyclic graph for representing that user experiences, V=< v 1..., v i... v n> is the set of node in directed acyclic graph, each node v irepresent a mobile application, N is natural number, and N represents total number of mobile application in described application scenarios,
Figure FDA0000397818010000011
it is the set on limit in directed acyclic graph; In G, the probability of each node is carried out, with a five-tuple < s i, x i, h i, p i, c i> represents, wherein, and s irepresent node v istarting Executing Time, x irepresent node v iexecution required time, h irepresent node v irate, c irepresent XM v ipower consumption, p irepresent node v iconfidence probability; Described confidence probability refers under certain rate, time, power consumption situation, the probability that node can complete; Determined probability selection event, also just determined radio network interface, the Starting Executing Time s of each node ialso determined;
One group of radio network interface set expression is W=< w 1..., w j..., w m>, with
Figure FDA0000397818010000013
be respectively used to represent node v iat radio network interface, be w jsituation under, the probability density distribution of rate, time of implementation and power consumption;
Specifically passing through step is below that each node is selected radio network interface, obtains best user and experiences:
Step 1, directed acyclic graph G is carried out to topological sorting, obtain a current collating sequence;
Step 2, according to the probability of rate the probability of time of implementation
Figure FDA0000397818010000015
probability with power consumption
Figure FDA0000397818010000016
determine each node v iall probability carry out event P i,n;
Described probability is carried out event P i,na four-tuple < p i,n, c i,n, t i,n, h i,n>, for representing node v ito be more than or equal to p i,nprobability, consumed power consumption c i,n, time of implementation t i,nwith rate h i,nresource complete this node v irepresented mobile application; P i,nrepresentation node v in probability carry out event, each node has a plurality of probability to carry out events;
Step 3, optimal user experience UX is set minfor infinity, total time of implementation T is set t(G) be 0;
Step 4, enter the 1st layer of solution space, initialization current active node A 1middle information is empty;
The level of the solution space that step 5, basis enter, finds the present node v in G k, establish as the l layer that advances into solution space, present node v kfor l node in the current collating sequence of G;
Step 6, enumerate present node v kall probability carry out events;
Step 7, judge whether to select all probability of present node to carry out events, if not, execution step 8; If so, perform step 14;
Step 8, select present node v successively kprobability carry out event;
Step 9, renewal present node v kexecution end time f k, described execution end time f kfor node v kstarting Executing Time and node v kexecution required time and;
Step 10, upgrade total time of implementation T t(G): judgement present node v kexecution end time f kwhether be greater than total time of implementation T t(G), if so, total time of implementation T is set so t(G) be this present node v kexecution end time f k, otherwise, keep total time of implementation T t(G) constant;
Step 11, according to the probability of selecting, carry out event update current active node A linformation, comprising: upgrade root node from current collating sequence to present node v kthe probability of having done is carried out the selection of event; Upgrading the total time of implementation producing is T t(G);
Upgrade the total rate H producing t(G):
Figure FDA0000397818010000021
h ithe rate that represent the probability execution event that i node of current collating sequence selected;
Upgrade the total power consumption C producing t(G):
Figure FDA0000397818010000022
c ithe power consumption that represents the probability execution event that i node of current collating sequence selected;
Upgrade total confidence probability P t(G):
Figure FDA0000397818010000023
p ithe probability that represents the probability execution event that i node of current collating sequence selected;
Step 12, judgement current active node A ltotal time of implementation in information, total rate, total power consumption and total confidence probability, meet time, rate, power constraints < T, H, and C > and confidence probability θ, if meet, execution step 13, otherwise go to step 7 execution;
Step 13, renewal optimal user experience UX min: the user who first determines current active node experiences UX t(G):
UX t(G)＝α*H t(G)+β*T t(G)+γ*C t(G)
Wherein, experience parameter alpha, β and γ and meet alpha+beta+γ=1;
Then judge whether to satisfy condition: pux (UX t, P t)≤pux (UX min, P min), if so, upgrade optimal user experience UX minvalue be UX t(G), if not, keep optimal user experience UX minin value constant; Wherein, pux (UX t, P t) represent to experience UX for user t(G), make the user of figure G experience UX (G) P (UX (G) the > UX that satisfies condition t(G)) < 1-P t, pux (UX min, P min) represent for current optimal user experience UX min, make the user of figure G experience UX (G) P (UX (G) the > UX that satisfies condition min(G)) < 1-P min;
Step 14, renewal depend on node v kthe Starting Executing Time of each node of data;
Whether lower one deck (l+1) layer of the current level of the solution space that step 15, judgement enter exists, and if not, execution step 16, if so, enters (l+1) layer of solution space, and by active node A lin information be assigned to current active node A l+1, then go to step 5 execution;
Step 16, by active node A nwith optimal user experience UX minoutput, method ends; Active node A nin the information that comprises have root node from the current collating sequence of G to present node v nthe probability done is carried out EventSelect, from the probability execution event of each node of selecting, obtains the selected radio network interface of this node.
2. the method that a kind of enhancing smart phone user of selecting based on wave point according to claim 1 is experienced, it is characterized in that, renewal Starting Executing Time described in step 14, specifically: by the Starting Executing Time comparison of current time and this node, if current time is large, the Starting Executing Time that upgrades this node is current time, otherwise it is constant to preserve the Starting Executing Time of this node, and the Starting Executing Time of each node is initially set to 0.
CN201110131397.4A 2011-05-20 2011-05-20 Method based on wireless network interface selection to improve a smart phone user experience Expired - Fee Related CN102209369B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201110131397.4A CN102209369B (en) 2011-05-20 2011-05-20 Method based on wireless network interface selection to improve a smart phone user experience

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201110131397.4A CN102209369B (en) 2011-05-20 2011-05-20 Method based on wireless network interface selection to improve a smart phone user experience

Publications (2)

Publication Number Publication Date
CN102209369A CN102209369A (en) 2011-10-05
CN102209369B true CN102209369B (en) 2014-03-19

Family

ID=44697985

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201110131397.4A Expired - Fee Related CN102209369B (en) 2011-05-20 2011-05-20 Method based on wireless network interface selection to improve a smart phone user experience

Country Status (1)

Country Link
CN (1) CN102209369B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102811475A (en) * 2012-08-28 2012-12-05 哈尔滨工业大学 Method for selecting optimal network to access based on android intelligent terminal
CN103781187B (en) * 2012-10-19 2018-05-11 华为终端（东莞）有限公司 A kind of method and terminal for controlling file transmission
CN103220351A (en) * 2013-04-15 2013-07-24 广东欧珀移动通信有限公司 Mobile terminal, file transmission method and system thereof and network system
CN109474521A (en) * 2018-12-22 2019-03-15 Oppo广东移动通信有限公司 Link aggregation implementation method and Related product

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101179516A (en) * 2006-11-10 2008-05-14 北京航空航天大学 Digraph based data distributing method
CN101605332A (en) * 2009-06-26 2009-12-16 北京鼎盛光华科技有限责任公司 Vehicle multi-mode wireless network communication system

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7389279B2 (en) * 2005-06-24 2008-06-17 Palo Alto Research Center Incorporated System and method for time-aware path finding

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101179516A (en) * 2006-11-10 2008-05-14 北京航空航天大学 Digraph based data distributing method
CN101605332A (en) * 2009-06-26 2009-12-16 北京鼎盛光华科技有限责任公司 Vehicle multi-mode wireless network communication system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
CooISpots: reducing the power consumption of wireless mobile devices with multiple radio interfaces;Trevor Pering等;《MobiSys2006》;20060622;全文 *
Trevor Pering等.CooISpots: reducing the power consumption of wireless mobile devices with multiple radio interfaces.《MobiSys2006》.2006,

Also Published As

Publication number Publication date
CN102209369A (en) 2011-10-05

Similar Documents

Publication Publication Date Title
CN109753751B (en) MEC random task migration method based on machine learning
CN109286664A (en) A kind of computation migration terminal energy consumption optimization method based on Lagrange
CN113435472A (en) Vehicle-mounted computing power network user demand prediction method, system, device and medium
CN102209369B (en) Method based on wireless network interface selection to improve a smart phone user experience
CN105472631A (en) Service data quantity and/or resource data quantity prediction method and prediction system
CN103702401A (en) User behavior analysis-based energy-saving method for mobile network
CN114650228B (en) Federal learning scheduling method based on calculation unloading in heterogeneous network
CN105792218A (en) Optimization method of cognitive radio network with radio frequency energy harvesting capability
CN102411594B (en) Method and device for obtaining information
Jiang et al. Q-learning based task offloading and resource allocation scheme for internet of vehicles
CN114126021B (en) Power distribution method of green cognitive radio based on deep reinforcement learning
Bai et al. Joint optimization of computation offloading, data compression, energy harvesting, and application scenarios in fog computing
Leithon et al. Task scheduling strategies for utility maximization in a renewable-powered IoT node
Kashyap et al. DECENT: Deep learning enabled green computation for edge centric 6G networks
CN110061882A (en) The method of License dynamic adjustment
Ji et al. Multi-Agent Reinforcement Learning Resources Allocation Method Using Dueling Double Deep Q-Network in Vehicular Networks
CN103249050B (en) Based on the multiple dimensioned frequency spectrum access method of business demand
CN102711259B (en) Wireless heterogeneous network throughput optimization method based on Markov process
CN116939866A (en) Wireless federal learning efficiency improving method based on collaborative computing and resource allocation joint optimization
CN109561129B (en) Cooperative computing unloading method based on optical fiber-wireless network
Wang et al. Joint job offloading and resource allocation for distributed deep learning in edge computing
CN105426978A (en) Service concurrency prediction method and service concurrency prediction system
CN113840306B (en) Distributed wireless network access decision method based on network local information interaction
CN113411765B (en) Mobile intelligent terminal energy consumption optimization method based on multi-sensor cooperative sensing
CN104540140B (en) A kind of dynamic frequency allocation method of wireless communication networks

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20140319

Termination date: 20150520

EXPY Termination of patent right or utility model