CN101478517A - OFDM system dynamic bit loading method based on ordering - Google Patents

Abstract

The invention is an improvement to a conventional bit loading method based on a Greedy algorithm for dynamic allocation of a loading bit of a sub-carrier and suitable for multiuser and single user systems employing an Orthogonal Frequency Division Multiplexing (OFDM) technology, wherein the biggest improvement to the Greedy algorithm is to sort and group the sub-carrier by using real feedback channel state information, realize the bit loading of multi-sub-carriers at a time, so as to accelerate the bit loading process. Comparing with the conventional Greedy algorithm for bit loading of single sub-carrier at a time, the method of the invention greatly reduces the complexity of the algorithm at the same time of insuring performance.

Description

Ofdm system dynamic bit loading method based on ordering

Technical field

The present invention relates to wireless communication field, a kind of dynamic bit loading method that is used for orthogonal frequency (OFDM) system.

Background technology

OFDM (OFDM) is the multi-carrier transmission technology under a kind of mobile communication environment, its maximum characteristics are divided into available band the experimental process carrier wave exactly, thereby frequency-selective channel is changed into the flat fading channel of a series of quadratures, in conjunction with Cyclic Prefix, OFDM can effectively resist multipath effect, has greatly improved message transmission rate; Because subcarrier overlapping on frequency spectrum, OFDM has greatly improved the availability of frequency spectrum; Because IFFT/FFT can be used for realizing the modulating/demodulating of OFDM, has greatly simplified the design of multicarrier system transmitter and receiver.OFDM has obtained to use widely in digital video/audio broadcasting (DVB-T/DAB), WLAN (wireless local area network) (IEEE802.11Serial, Hiper-LAN/2), Digital Subscriber Line fields such as (xDSL).

OFDM is easy to combine with the Dynamic Resource Allocation for Multimedia technology, can significantly improve power system capacity, thereby is subjected to extensive concern.In theory, water flood can obtain optimum bit loading performance.But the bit loading result that adopts water flood to obtain is continuous any real number, and with the BPSK in the practical communication system, QPSK, the discrete integer bit that the MQAM modulation requires loads and contradicts.At present, the algorithm that discrete integer bit loads can be divided into two classes: a class be on the basis of water-filling algorithm, round off approximate, as Chow algorithm, Fischer algorithm etc.; Another kind of is directly to solve discrete integer resource allocation problem, as Greedy (greediness) algorithm, Campello algorithm.Can prove that greedy algorithm is the optimal algorithm that discrete integer bit loads, when comparing ofdm system resource allocation algorithm performance, often be that the people adopts.But more for a long time, complexity is very high in user resources for greedy algorithm, can not satisfy the requirement of real time communication.In view of this, the bit loading algorithm that complexity is low to performance is excellent is studied and is had a great deal of practical meanings.

Summary of the invention

The distinguishing feature of greedy algorithm is only to be that a subcarrier loads several bits at every turn, and each bit loading all needs to carry out necessary calculating to determine target sub-carriers.Its complexity mainly is present in the selection of each target sub-carriers and calculates.When speed more when the subcarrier that distributes for the user or that require was higher, the real-time of algorithm was difficult to guarantee.The low bit loading algorithm of existing complexity is with respect to the greedy algorithm existence loss of energy often.In order to take into account two aspects of performance and complexity, the present invention at first determines to have the border that identical greedy algorithm bit loads the subcarrier of general character, subcarrier grouping after will sorting subsequently is a plurality of subcarrier bit loading at every turn simultaneously, has greatly reduced the complexity that target sub-carriers is selected.

The invention provides in the ofdm system based on the improving one's methods of Greedy algorithm, its principle is: at first the subcarrier that distributes for the user is carried out descending according to the size of its channel gain square, according to the boundaries of packets that calculates subcarrier is divided into groups then.With the purpose of subcarrier grouping is when carrying out the bit loading, and all subcarriers in the selected group are bit loading simultaneously, has greatly simplified the bit loading procedure.The loading sequence of each group is during by the group bit loading: the group that loads is for the first time: A ₁The group that loads is for the second time: A ₁, A ₂The group of Jia Zaiing is for the third time: A ₁, A ₂, A ₃The group that loads for the 4th time is: A ₁, A ₂, A ₃, A ₄... consider that the randomness of sub-carrier channels gain reaches the uncertainty of subcarrier number in every group that causes thus, for the speed that meets the demands, utilization of the present invention is drawn the method or the Greedy algorithm of group of molecules and is done further refinement bit loading, progressively approaches the speed of requirement.

Bit loading method provided by the invention comprises the steps:

Consider an ofdm system, for certain user, for the channel gain of the subcarrier of its distribution and respective sub is all known, total bit number that the user need load is designated as R.

At first the base station is carried out descending to all subcarriers of distributing for the user according to the size of channel gain square, and the maximum of this user's sub-carrier channels gain square is designated as | α _Max| ²The method of calculating class boundary is as follows: establishing speed-power function is that (f (c) is illustrated in channel gain and equals at 1 o'clock f (c), on some subcarriers, realize c required transmitting power of information bit of reliable reception), when loading Δ B bit for the first time, first group of all subcarrier all loads Δ B bit, its border is decided by whether subcarrier has subcarrier will load second Δ B bit in this group, so the channel gain of first group of interior norator carrier wave j square | and α _j| ²Should satisfy:

$\frac{f\left(2\mathrm{\&Delta;B}\right)-f\left(\mathrm{\&Delta;B}\right)}{{\left|{\mathrm{\&alpha;}}_{\max }\right|}^2}>\frac{f\left(\mathrm{\&Delta;B}\right)-f\left(0\right)}{{\left|{\mathrm{\&alpha;}}_j\right|}^2}.$ Consider that the modulation system of radio communication mostly is MQAM (M=4,16,64) in the actual conditions, c is an even number, and the formula that embodies of speed-power function can adopt: $f\left(c\right)=\frac{N_0}{3}{\left[Q^{-1}\left(\frac{P_e}{4}\right)\right]}^2(2^c-1),$ N ₀Be noise variance, P _eBe bit error rate, $Q\left(x\right)=\frac{1}{\sqrt{2\mathrm{\&pi;}}}\underset{x}{\overset{\&infin;}{\&Integral;}}{e}^{-{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A200810222668E00162.png,A200810222668E00171.png"\; state="\{\{state\}\}">t</figure-callout>}}^2/2}\mathrm{dt};$ The arrangement following formula can get, ${\left|{\mathrm{\&alpha;}}_j\right|}^2>\frac{{\left|{\mathrm{\&alpha;}}_{\max }\right|}^2}{2^{\mathrm{\&Delta;B}}},$ If get Δ B=2, then have ${\left|{\mathrm{\&alpha;}}_j\right|}^2>\frac{{\left|{\mathrm{\&alpha;}}_{\max }\right|}^2}{4},$ Promptly ${\mathrm{<figure-callout\; id="1"\; label="b"\; filenames="A200810222668E00162.png"\; state="\{\{state\}\}">b</figure-callout>}}_1=\frac{1}{4};$ For the border of n group, similar above-mentioned derivation can get: $b_n=\frac{1}{4^n}.$ According to the border proportionality coefficient of calculating be: $b_n=\frac{1}{{\left(2^{\mathrm{\&Delta;B}}\right)}^n},n=\mathrm{0,1,2},L,$ Determine respectively to organize the boundary of sub-carrier channels gain square.First group of A ₁: (b ₁| α _Max| ², b ₀| α _Max| ²), second group of A ₂: (b ₂| α _Max| ², b ₁| α _Max| ²) ..., because subcarrier by square descending of its channel gain, only needs according to the order of sequence subcarrier sequence number branch to be gone into each group, even | α _n| ²∈ (b ₁| α _Max| ², b ₀| α _Max| ²), α then _n∈ A ₁If | α _n| ²∈ (b ₂| α _Max| ², b ₁| α _Max| ²), α then _n∈ A ₂L L.

Bit at first is that whole group loads when loading, and during whole group bit loading, order of each group is: A for the first time ₁Interior all subcarriers load Δ B bit simultaneously; A for the second time ₁, A ₂Interior all subcarriers load Δ B bit simultaneously; Be A for the third time ₁, A ₂, A ₃And the like, and behind each loaded, with present user bit number that has loaded and the total bit number R that needs to load relatively, three kinds of results can appear in this moment: one, both equate that the bit loading procedure finishes in such cases; Two, the former cancels this bit in such cases and loads greater than the latter, finishes the loading of remaining bits with drawing group of molecules or greedy algorithm; Three, the former enters bit loading procedure next time in such cases less than the latter.

When the whole group of user bit loading〉during R, each the set of dispense result when writing down the last bit loading＜R, what be used for the back draws the group of molecules operation.Group of molecules is drawn in grouping to above-mentioned record: A ₁=A ₁₁+ A ₁₂, the public boundary of two son groups is got A ₁The mid point on border, the left and right sides, wherein A ₁₁Be called A ₁Left side son group, A ₁₂Be called A ₁Right son group.The bit loading principle and the order of each son group are: at first suppose all left side son group (A ₁₁, A ₂₁, A ₃₁, A ₄₁L) all subcarriers in load Δ B bit, with user bit number that has loaded and the total bit number R that needs to load relatively, four kinds of results can occur: one, and both equate that the bit loading procedure finishes in such cases; Two, the former once more divides new child group with all left side son groups as original packet greater than the latter, and the child group that new division obtains is operated; Three, the former less than the latter and greater than the latter's β doubly finishes the loading of remaining bits this moment with greedy algorithm; Four, the former is less than β times of the latter, and divide new child group with all right son groups once more as original packet this moment, and the child group that new division obtains is operated.

Characteristics of the present invention are methods of utilizing subcarrier ordering, grouping, once be a plurality of subcarrier bit loading, the complexity when having reduced the each chooser carrier wave of Greedy algorithm.

Description of drawings

Fig. 1 shows the general flow chart to the subcarrier bit loading.

Fig. 2 shows and adopts the flow chart of drawing group of molecules or Greedy algorithm loading remaining bits.

Fig. 3 respectively organizes an example of loading sequence when having provided whole group of bit loading.

The concrete parameter that table 1 has related to when having provided the emulation of complexity numerical value.

Table 1

Simulation parameter	Note
		User's sub-carrier number (N)	8～512
The bit number R=ratioN that the user is transmitted in each OFDM symbol	ratio：2～4
		The group that participates in Bit Allocation in Discrete is counted g	6 (mean values)
Draw the number of times k of group of molecules	6 (mean values)
		Each bit number Δ B that loads	2

Fig. 4 has provided when the scope of subcarrier number is 8～128, the proportionate relationship of two kinds of algorithm computation complexities (improving the complexity of the complexity of algorithm divided by the Greedy algorithm).

Fig. 5 has provided when the scope of subcarrier number is 128～512, the proportionate relationship of two kinds of algorithm computation complexities (improving the complexity of the complexity of algorithm divided by the Greedy algorithm).

Embodiment

The present invention will be described in detail below by accompanying drawing.

Fig. 1 shows the general flow chart to the subcarrier bit loading.

Suppose the channel gain of base station known users, R represents total bit number that the user need load.

This flow process enters step 102 from step 101, finds out the maximum of user's sub-carrier channels gain square | α _Max| ², and according to | α _Max| ²Determine the border of each subcarrier grouping.

In step 103, subcarrier is pressed the channel gain ordering, and subcarrier is divided into groups according to 102 borders of determining.

In step 104, begin to enter circulation process, set counting variable initial value n=1.

In step 105, finish whole group bit loading process, the order of the n time each group loading is: A ₁, A ₂..., A _n

In step 106, the bit number that has loaded is compared with R, if both equate that then enter step 110, the bit loading procedure finishes; If both are unequal, then change step 107.

In step 107, the bit number that has loaded is compared with R, if the former is less than the latter, then enter step 108; If the former is greater than the latter, then enter step 109.

In step 108, counting variable increases unit of measurement, i.e. a n=n+1.

In step 109, cancel the n time bit and load, adopt stroke group of molecules or Greedy algorithm to finish remaining bits and load.

Fig. 2 shows and adopts the flow chart of drawing group of molecules or Greedy algorithm loading remaining bits.

This flow process enters step 202 from step 201, determines the border of each height group, son group about subcarrier is divided into.

In step 203, the subcarrier that is assumed to be in all left side son groups loads Δ B bit simultaneously.

In step 204, the bit number that has loaded is compared with R, if both equate that then enter step 210, the bit loading procedure finishes; If both are unequal, then change step 205.

In step 205, the bit number that has loaded is compared with R, if the former is less than the latter, then enter step 207; If the former is greater than the latter, then enter step 206.

In step 206, with left side son group as original packet continue to divide new about the son group, and enter step 203.

In step 207, the bit number that has loaded is compared with β R, if the former is less than the latter, then enter step 208; If the former more than or equal to the latter, enters step 209.

In step 208, all left side son groups are all loaded Δ B bit simultaneously, and with right son group as original packet continue to divide new about the son group, enter step 203.

In step 209, finish remaining bit with greedy algorithm and load.

Fig. 3 respectively organizes an example of subcarrier loading sequence when having provided whole group of bit loading.

Its abscissa is the group number of subcarrier, and ordinate is the bit number that every group of subcarrier loads.The sub bit loading sequence of organizing similarly after drawing group of molecules.

Performance evaluation

The concrete parameter that table 1 has related to when having provided the emulation of scheme complexity number value, as the sub-carrier number that distributes for each user, the bit number that each user loads and the proportionate relationship of its rate requirement participate in the group number of bit loading, number of times of stroke group of molecules or the like.

Fig. 4, the abscissa of Fig. 5 is represented the sub-carrier number that the user distributes, ordinate is represented the ratio (improving the complexity of the complexity of algorithm divided by the Greedy algorithm) of two kinds of algorithm computation complexities, wherein the scope of the subcarrier number that the user distributes among Fig. 4 is 8～128, and the scope of the subcarrier number that the user distributes among Fig. 5 is 128～512.As can be seen from the figure, complexity of the present invention when user's rate requirement is higher or the subcarrier of distribution be starkly lower than the Greedy algorithm more for a long time, the scale of resource allocation problem is big more, the relative performance of the algorithm that this patent provides is good more.

Analysis of complexity

In the practical application, the modulation system that radio communication adopts mostly is QPSK, 16QAM, 64QAM, each bit number that loads is 2, because the improvement algorithm that this patent proposes is identical with the Greedy algorithm performance when 2 bits of each loading, the complexity with this patent and Greedy algorithm compares here.

Analysis of complexity is as follows:

Greedy algorithm: load Δ B bit for the first time, calculate

I=1,2, L, N needs altogether to calculate N time, and wherein N is expressed as the subcarrier number that the user distributes, and the complexity that therefrom choosing minimum value needs is: Load for the second time Δ B bit, calculate Wherein ii is the last subcarrier of choosing, and only needs altogether to calculate once, therefrom chooses minimum value and need be N time relatively; Subsequent complexity is analyzed with for the second time similar.After assigning, circulation altogether

The bit number inferior, that R need send for this user.So total complexity is: $\frac{R}{\mathrm{\&Delta;B}}\mathrm{gN}+(N+\frac{R}{\mathrm{\&Delta;B}}-1)g(1+1)=2N+(N+2)g\frac{R}{\mathrm{\&Delta;B}}-2,$ Be approximately

Wherein each real number addition needs 1 floating-point operation, and each real multiplications needs 1 floating-point operation.

Improve algorithm: at first choose the subcarrier with maximum channel gain square, its number of times that need compare is N, and all subcarriers of user are carried out descending according to the size of channel gain square, and its number of comparisons that needs is approximately 1.386Nlog ₂N-2.846N.When beginning to divide into groups, at first determine A ₁Interior subcarrier, its method is: with A ₁Right margin b ₁| α _Max| ²Insert in the middle of the subcarrier after sorting, the number of comparisons under its worst case is

Integer less than the minimum of x is got in expression, inserts other group (son group) outland and analyzes situation similarly, only the scale of problem diminish (littler) during interpolation.The actual range of considering the N value inserts the complexity that causes and can ignore fully tens between the hundreds of.

If in the operation, grouping add up to G, in the actual bit loading procedure, when beginning to draw group of molecules, participating in the subcarrier packet count that bit loads is g, the number of comparisons that needs before the then unallocated son group is: g.Draw in the group of molecules process, each really border of stator pack need be calculated b one time _G2=(b _l+ b _r)/2, its complexity are 2g, also need once the comparison with R simultaneously, so complexity is 2g+1; Supposing needs in the assigning process to draw group of molecules k time, and then complexity is kg (2g+1); Consider practical application and for α _MaxSetting, g, k is generally units, and finishes the complexity that remaining bits distributes with the Greedy algorithm at last and can ignore.So total complexity is: N+1.386N1gN-2.486N+k (2g+1) ≈ 1.386N log ₂N-1.486N+k (2g+1).In the actual allocated process, N generally tens to value between hundreds of.

Claims (6)

1, a kind of dynamic bit loading method that is used for orthogonal frequency (OFDM) multicarrier system the method is characterized in that: the base station is according to user's channel condition information, and subcarrier is carried out descending by the size of channel gain square; According to the characteristics in the Greedy algorithm bit loading procedure, the subcarrier after the ordering by the boundary grouping of determining, is carried out bit with the unit of being grouped into to the whole according to the order of sequence group of each group subcarrier then and loads; According to the performance of bit loading, and the parameter setting of system, adopt stroke group of molecules or greedy algorithm to finish remaining bits and load.

2, according to the channel gain described in the claim 1 according to subcarrier, the subcarrier that will distribute to the user carries out descending by the size of channel gain square, subcarrier after the ordering is divided into groups, it is characterized in that, at first the base station square is carried out descending to all subcarriers of distributing to the user according to channel gain, and select the maximum of all sub-carrier channels of user gain square, and be designated as | α _Max| ², according to the border proportionality coefficient of each group of calculating $b_n=\frac{1}{{\left(2^{\mathrm{\&Delta;B}}\right)}^n},$ N=0,1,2 ..., wherein Δ B is the bit number that at every turn loads for each subcarrier.Determine the boundary of the channel gain square of subcarrier in the grouping of each subcarrier: b ₀| α _Max| ²=| α _Max| ², $b_1{\left|{\mathrm{\&alpha;}}_{\max }\right|}^2={\frac{1}{2^{\mathrm{\&Delta;B}}}\left|{\mathrm{\&alpha;}}_{\max }\right|}^2,$ $b_2{\left|{\mathrm{\&alpha;}}_{\max }\right|}^2=\frac{1}{{\left(2^{\mathrm{\&Delta;B}}\right)}^2}{\left|{\mathrm{\&alpha;}}_{\max }\right|}^2,$ ..., then each subcarrier grouping is respectively by the interval of channel gain square division: first group of A ₁: (b ₁| α _Max| ², b ₀| α _Max| ²), second group of A ₂: (b ₂| α _Max| ², b ₁| α _Max| ²) ..., because subcarrier only needs according to the order of sequence subcarrier sequence number branch to be gone into each group by square descending of its channel gain, even | α _n| ²∈ (b ₁| α _Max| ², b ₀| α _Max| ²), α then _n∈ A ₁If | α _n| ²∈ (b ₂| α _Max| ², b ₁| α _Max| ²), α then _n∈ A ₂

3, organize the bit loading according to each group is put in order described in the claim 1, it is characterized in that: the order of whole group bit loading is, for the first time to A ₁Interior all subcarriers load Δ B bits; For the second time to A ₁, A ₂Interior all subcarriers load Δ B bits; For the third time to A ₁, A ₂, A ₃Interior all subcarriers load Δ B bits; The 4th time to A ₁, A ₂, A ₃, A ₄Interior all subcarriers load Δ B bits; ... by that analogy, and behind each loaded, with present user bit number that has loaded and the total bit number that needs to load relatively, three kinds of results can appear in this moment: one, and both equate that the bit loading procedure finishes in such cases; Two, the former cancels this bit in such cases and loads greater than the latter, finishes the loading of remaining bits with drawing group of molecules or greedy algorithm; Three, the former enters whole next time group bit loading procedure in such cases less than the latter.

4, according to the group of molecules of drawing described in the claim 1, it is characterized in that: the border, the left and right sides of remembering the g grouping is respectively b _l, b _r, the border of two son groups then repartitioning is respectively b _G1=b _l, b _G2=(b _l+ b _r)/2, b _G3=b _r, left side son group A _G1For channel gain square at (b _G1| α _Max| ², b _G2| α _Max| ²) interior subcarrier, the right son group A _G2For channel gain square at (b _G2| α _Max| ², b _G3| α _Max| ²) interior subcarrier; Situation when continuing the division of antithetical phrase group similarly; If do not have subcarrier in the sub-class interval of calculating, then the child group of Hua Fening is an empty set.

5, according to the parameter setting condition of the system described in the claim 1, it is characterized in that, 0＜β＜1, parameter beta is chosen according to the actual conditions that algorithm requires, and the purpose of setting β is: reduce the too small complexity that causes of grouping.

6, draw group of molecules or greedy algorithm bit loading according to the employing described in the claim 1, it is characterized in that, the pass of the child group of new division and the grouping of original subcarrier is A ₁₁+ A ₁₂=A ₁, A ₂₁+ A ₂₂=A ₂, The principle of antithetical phrase group bit loading is: suppose the child group (A to all left sides ₁₁, A ₂₁, A ₃₁, A ₄₁) all subcarriers load Δ B bits, with user bit number that has loaded and the total bit number that needs to load relatively, four kinds of results can appear: one, both equate that the bit loading procedure finishes in such cases; Two, the former is greater than the latter, will draw last time all left side son groups of group of molecules gained as original packet according to claim 4 divide once more new about the son group, the child group that new division obtains is operated; Three, the former less than the latter and greater than the latter's β doubly finishes the loading of remaining bits this moment with greedy algorithm; Four, the former is less than β times of the latter, and once more divide new child group as original packet according to claim 4 with all right son groups this moment, and the child group that new division obtains is operated.

Images