Scheduling method, radio system, base station, and scheduling module

Abstract

There is provided a base station comprising: an estimation unit for estimating a second or higher order statistics for each user terminal in a set of user terminals to be served; an estimation unit for estimating a scheduling metric based on the estimated second or higher order statistics; and a scheduling unit for controlling scheduling based on the estimated scheduling metric.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

In the following, the invention will be described in greater detail with reference to the embodiments and the accompanying drawings, in which

FIG. 1 shows an example of a radio system;

FIG. 2 illustrates an example of a base station according to an embodiment of the present invention;

FIG. 3 illustrates uncoded bit error rates for different scheduling schemes;

FIG. 4 illustrates throughput curves for different scheduling schemes; and

FIG. 5 illustrates an example of a scheduling method according to an embodiment of the present invention.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

FIG. 1 illustrates an example of a radio system to which embodiments of the invention can be applied. A radio system in FIG. 1, known at least as UTRAN [UMTS (Universal Mobile Telecommunications System) Terrestrial Radio Access Network] 130, is taken as an example. The UTRAN belongs to the third generation and is implemented with WCDMA (Wideband Code Division Multiple Access) technology. The solution is not limited to a WCDMA radio interface but applications exist which are implemented with cdma2000, MC-CDMA (MultiCarrier Code Division Multiple Access) or OFDMA (Orthogonal Frequency Division Multiple Access) technologies without restricting the invention to the above-mentioned technologies.

FIG. 1 is a simplified block diagram showing the most important parts of a radio system and interfaces between them at a network-element level. The structure and functions of the network elements are not described in detail, because they are generally known.

The main parts of a radio system are a core network (CN) 100, a radio access network 130 and a terminal (UE) 170. The term UTRAN is short for UMTS Terrestrial Radio Access Network, i.e. the radio access network 130 belongs to the third generation and is implemented by wideband code division multiple access (WCDMA) technology. The main elements of the UTRAN are a radio network controller (RNC) 146, 156, Node Bs 142, 144, 152, 154 and user a terminal 170. The UTRAN is attached to the existing GSM core network 100 via an interface called Iu. This interface is supported by the RNC 146, 156, which manages a set of base stations called Node Bs 142, 144, 152, 154 through interfaces called Iub. The UTRAN is largely autonomous from the core network 100 since the RNCs 146, 156 are interconnected by the Iur interface.

On a general level, the radio system can also be defined to comprise a user, such as a subscriber terminal or a mobile phone, and a network part, which comprises the fixed infrastructure of the radio system, i.e. the core network, radio access network and base station system.

From the point of view of Node B 142, 144, 152, 154, i.e. a base station, there is one controlling RNC 146, 156 where its lub interface terminates. The controlling RNC 146, 156 also takes care of admission control for new mobiles or services attempting to use the Node B 142, 144, 152, 154. The controlling RNC 146, 156 and its Node Bs 142, 144, 152, 154 form an RNS (Radio Network Subsystem) 140, 150.

The terminal 170 may comprise mobile equipment (ME) 172 and a UMTS subscriber identity module (USIM) 174. The USIM 174 contains information related to the user and information related to information security in particular, for instance, an encryption algorithm.

In UMTS networks, the terminal 170 can be simultaneously connected to a plurality of Node Bs in the occurrence of soft handover.

From the point of view of the terminal 170, a serving RNC 146, 156 is provided which terminates the mobile link layer communications. From the point of view of the CN 100, the serving RNC 146, 156 terminates the Iu for this terminal 170. The serving RNC 146, 156 also takes care of admission control for new mobiles or services attempting to use the CN 100 over its Iu interface.

In the UMTS, the most important interfaces between network elements are the Iu interface between the CN 100 and the radio access network 130, which is divided into the interface IuCS on the circuit-switched side and the interface IuPS on the packet-switched side, and the Uu interface between the radio access network and the terminal.

FIG. 2 shows a simplified outline of an embodiment of the present invention. An exemplary base station 142 of FIG. 2 comprises multiple antennas 202, 204 for receiving data streams from several user terminals concurrently. The base station 142 may also comprise the following elements: one or more transceiver units 206, a processing unit 208 for controlling the functions of the base station, an estimation unit 210 for estimating second or higher order statistics, an estimation unit 212 for estimating scheduling metrics, and a scheduling unit 214.

In an embodiment, the estimation unit 212 for estimating the scheduling metric and the scheduling unit 214 are part of a scheduling module 200 of the base station 142. However, it is also possible that the scheduling unit 214 and the estimation unit 212 are located in some other part of the base station or the radio system.

In an embodiment, the estimation unit 210 is configured to estimate the second or higher order statistics for each user terminal in a set of user terminals to be served. The estimated second or higher order statistics of each user terminal k may have the form of:

$E\left\{H_kH_k^H\right\}\in C^{M_R\times M_{R_k}},$

where H_kεC^MR×MT is a channel matrix for the kth user terminal, E denotes a mathematical expectation, and M_T and M_R are the numbers of transmit and receive antennas.

In an embodiment, the estimation unit 212 for estimating the scheduling metric is configured to estimate an ergodic sum capacity of multiple access channels based on the second or higher order statistics received from the estimation unit 210. The estimation unit 212 is further configured to estimate the scheduling metric based on the estimated ergodic sum capacity.

In an embodiment, the estimation unit 212 is configured to estimate the ergodic sum capacity of multiple access channels by using a second-order Taylor series of the ergodic sum capacity.

The scheduling unit 214 is configured to control scheduling based on the estimated scheduling metric by using any suitable scheduling algorithm.

Let us next study some theoretical basis of the proposed solution. In the following example, all expressions refer to a frequency-flat case. In the case of a frequency selective channel, the implementation example can be applied in the same way on each resource unit of an Orthogonal Frequency Division Multiplex (OFDM) system by taking into account that the channel statistics in the case of the channel model stated below do not depend on the subcarrier index of an OFDM system.

A frequency-flat MIMO transmission channel can be modeled as y=Hx+n, where x and y are transmit and receive vectors, HεC^MR×MT is the joint channel matrix of all user terminals, and M_T and M_R are the number of transmit and receive antennas, respectively. Further, n is zero-mean, circularly symmetric additive Gaussian noise with per-component variance NO, i.e., E{nn^H}=N_oI_MR. The channel matrix H can be given as: H=[H_i, H₂, . . . , H_K]εC^MR×MT.

For deriving the scheduling metric, let us assume K active user terminals k=1, . . . , K having M_T,k transmit antennas, respectively. The total number of transmit antennas is, thus,

$M_T=\sum _{k=1}^KM_{T,k},$

which is complemented by M_R receive antennas at the base station.

Let the channel matrix H_kεC^MR×MT,k of each user terminal k be modeled as:

H_k=R_k^1/2H_w,kεC^MR×MT,k  (1)

where R_kεC^MR×MR is a receiver side correlation matrix for a kth user, and H_w,k is a matrix containing independent, Rayleigh-fading random channel coefficient elements of unit average power. For each user terminal k, the matrix R_k mainly depends on the geometry of the current propagation situation of that user terminal, as well as on different properties of the base station, such as the type of antenna array used. While H_w,k constitutes the fast-changing components of H_k, the correlation matrix R_k only contains the slowly changing components (“long-term channel knowledge”). In addition, R_k is constant across neighboring frequency bands as long as the total bandwidth is small compared to the carrier frequency.

When a certain set of users S⊂{1, 2, . . . , K} is tested for possible scheduling, a natural measure of performance is information on the theoretic capacity of a multi-user channel related to S. Assuming that only the R_k,k=1, . . . , K are known to the base station, the ergodic sum capacity of multiple access channels corresponding to S is given by:

$\begin{array}{cc}{C}_S\left(\rho \right)=E\left\{\log \det \left(I_{M_R}+\rho \sum _{i\in S}H_iH_i^H\right)\right\}& \left(2\right)\end{array}$

where

$\rho =\frac{E_S}{M_TN_0}$

is a normalized SNR value and the expectation is with respect to the fast fading terms H_w,k. Because the exact C_S(ρ) has to be calculated by Monte Carlo simulations, its direct use in practice is not possible. Therefore, the second-order Taylor series of C_S(ρ) developed at ρ₀=0 can be used. It can be expressed by:

$\begin{array}{cc}{C}_S\left(\rho \right)\approx \rho ·E\left\{\mathrm{trace}\left(\sum _{i\in S}H_iH_i^H\right)\right\}-\frac{{\rho }^2}{2}·E\left\{\mathrm{trace}\left({\left(\sum _{i\in S}H_iH_i^H\right)}^2\right)\right\}& \left(3\right)\end{array}$

where ρ is a tuning constant, E is a mathematical expectation, H_i is a channel for user terminal i, and H_i^H is a Hermitian conjugate of H_i.The linear and quadratic coefficients can be computed from the channel model described above in equation (3). They can be expressed by:

$\begin{array}{cc}E\left\{\mathrm{trace}\left(\sum _{i\in S}H_iH_i^H\right)\right\}=\sum _{k\in S}M_{T,k}\mathrm{trace}\left(R_k\right)& \left(4\right)\end{array}$

$\begin{array}{cc}\left\{\mathrm{trace}\left({\left(\sum _{i\in S}H_iH_i^H\right)}^2\right)\right\}=\sum _{k\in S}\sum _{l\in S}M_{T,k}M_{T,l}\mathrm{trace}\left(R_kR_l\right)+\sum _{k\in S}M_{T,k}\mathrm{trace}{\left(R_k\right)}^2.& \left(5\right)\end{array}$

Finally, using the expected value results from equations (4) and (5) in the capacity approximation (3), a good approximation of the ergodic sum capacity of the user set S can be given by:

$\begin{array}{cc}{C}_S\left(\rho \right)\approx \rho ·\sum _{k=1}^KM_{T,k}\mathrm{trace}\left(R_k\right)-\frac{{\rho }^2}{2}·\left(\sum _{k=1}^K\sum _{l=1}^KM_{T,k}M_{T,l}\mathrm{trace}\left(R_kR_l\right)+\sum _{k=1}^KM_{T,k}\mathrm{trace}{\left(R_k\right)}^2\right).& \left(6\right)\end{array}$

The approximation of equation (6) enables the base station to obtain an estimate of the sum capacity by only using the receiver-side correlation matrices R_k.

In an embodiment, an exemplary implementation can, thus, comprise:

• • 1. The base station determines a set of K_total users that need to be served,
  • 2. The base station obtains estimates of second or higher order statistics in the form of R_k for each user,
  • 3. The base station makes a scheduling decision based on the scheduling metric in equation (6) by using a suitable design parameter ρ=ρ_ref and an appropriate scheduling algorithm,
  • 4. The base station grants transmit permission to the selected users.

In an embodiment, the trace expressions in equation (6) can be precomputed for each user k or user pair (k,l), respectively, thus reducing the effort for testing each candidate S to a relatively low number of additions. Since the metric is generated from a Taylor series developed at ρ₀=0, it is beneficial to use low values ρ_ref for the metric computation even though the true ρ may, in fact, be much higher. The value of ρ_ref in use can be subjected to optimization.

There are different possibilities in choosing scheduling algorithms using a scheduling metric. In an embodiment, an opportunistic scheduling can be used. In opportunistic scheduling, it is assumed that out of K_total users exactly K_group users are selected for service. Of the

$\left(\begin{array}{c}{K}_{\mathrm{total}}\\ K_{\mathrm{group}}\end{array}\right)\hspace{1em}$

possible groups (i.e. possible sets S), a scheduler has to select the one with the maximum metric according to equation (6). This can be carried out by an exhaustive search or, in order to reduce the computation burden, by suboptimal search approaches.

In an embodiment, a round-robin (RR) based fair scheduling can be used. In the opportunistic scheduling scheme, it may occur that the far away or strongly coupled users never become scheduled (starvation). This situation can be avoided in the following way: in each scheduling run, one user is predetermined to be included in a subset to be served. This “free ticket” user is chosen in a round-robin fashion, thus providing fairness. The remaining members of the subset to be served are chosen according to the scheduling metric of equation (6). Since one user is fixed, exactly

$\left(\begin{array}{c}{K}_{\mathrm{total}}^{-1}\\ K_{\mathrm{group}}^{-1}\end{array}\right)\hspace{1em}$

possible subsets remain to be tested. Once again, either exhaustive or suboptimal search approaches can be used.

The proposed scheduling metric can also be combined with any other scheduling algorithms than those described above. For instance, QoS requirements can be included in the system through a scheduling algorithm that weighs the metric according to user priorities. In addition, only long-term CSI is required, which makes the scheduling metric robust and also computationally attractive.

Let us next consider the following situation: a total number of K_total user terminals wishes to transmit data to a single base station which comprises M_R>1 receive antennas. Multiple user terminals can be accommodated concurrently by exploiting the spatial properties of the joint transmission channel. However, the number of simultaneously decodable streams is in practice limited by M_R; therefore it often occurs that not all user terminals can be served within the same resource unit. A subset of user terminals has to be selected, which is a process also known as multiuser scheduling.

A number of scheduling algorithms is known. A common feature of these schemes is that a metric is needed to evaluate a possible set of user terminals for spatial compatibility among its members. In an embodiment, the sum capacity approximation of equation (6) can be used as a scheduling metric. For each potential user group G, a group-wise scheduling metric is defined:

M _group(G)^def=C_erg^G(ρ_ref)  (7)

where C_erg^G is the ergodic capacity estimate given in equation (6) when applied to the user group G. It is evaluated at the normalized SNR value ρ_ref which in practice may be chosen to differ from the true SNR ρ. Once the scheduling metric has been chosen as a function that evaluates potential user groups, different approaches are feasible for selecting the best group.

In an embodiment, exactly K_group out of K_total users are served, while the others are disregarded. The set of active users is determined as:

$=\arg \underset{\underset{\chi =K_{\mathrm{group}}}{\chi ⋐\left\{1,2,\dots ,K_{\mathrm{total}}\right\}}}{\max }M_{\mathrm{group}}\left(\chi \right),$

where M_group(G) is the group-wise metric defined in equation (7). The number of candidate groups the maximization operates upon is

$\left(\begin{array}{c}{K}_{\mathrm{total}}\\ K_{\mathrm{group}}\end{array}\right)\hspace{1em},$

which is the main influence on the complexity when an exhaustive search is applied.

The users that are not within G, i.e., those from {1, 2, . . . , K_total}−G, may be served in a sufficiently spaced frequency band where R_k matrices result in a different set of active users. Alternatively, the users can be required to wait until their channel conditions improve. It is possible that users with a weak channel or strong mutual coupling are never served (starvation); thus, this approach may lack fairness.

In an embodiment, metric-based scheduling is combined with a fairness-based scheduling rule as follows. When K_group out of K_total users are to be selected, the total group G is split into two parts, G=G⁽¹⁾∪G⁽²⁾, with preset cardinalities |G⁽¹⁾=K_group⁽¹⁾| and |G⁽²⁾=K_group⁽²⁾|, where K_group=K_group⁽¹⁾+K_group⁽²⁾.

The first part, G⁽¹⁾ is chosen independently of the channel knowledge. It constitutes the fairness component of the combined scheme. One possibility is to select the K_group⁽¹⁾ users that have not been scheduled for the longest time. Another possibility is to assign G⁽¹⁾ in a round-robin fashion, i.e. users 1, . . . , K_group⁽¹⁾ are selected in the first run, users K_group⁽¹⁾+1, . . . , 2K_group⁽¹⁾ in the second run, and so forth. Thus, G⁽¹⁾ is chosen independently of R_k. The second part of the group, G⁽²⁾, can be assigned based on the metric, using a best-subset approach as before,

${}^2=\arg \underset{\underset{\underset{\chi \bigcap {}^{\left(1\right)}=0}{\chi =K_{\mathrm{group}}^{\left(2\right)}}}{\chi ⋐\left\{1,2,\dots ,K_{\mathrm{total}}\right\}}}{\max }M_{\mathrm{group}}\left(\chi \bigcup {}^{\left(1\right)}\right).$

In this case, the number of candidates for the maximization can be computed as

$\left(\begin{array}{c}{K}_{\mathrm{total}}-K_{\mathrm{group}}^{\left(1\right)}\\ K_{\mathrm{group}}^{\left(2\right)}\end{array}\right).$

By choosing the decomposition of K_group into its summands K_group⁽¹⁾ and K_groups⁽²⁾, the trade-off between capacity maximization and fairness can be adjusted. The extreme case K_group⁽¹⁾=0 is equivalent to the best subset approach described above. On the other hand, choosing K_group⁽²⁾=0 totally disregards channel state information and focuses solely on fairness.

In an embodiment, a more stringent notion of fairness than that described above can be formulated as follows: within a certain predetermined time frame, all users can be scheduled equally often. Let us assume that during this time frame, the channel state information is static, i.e., the R_k do not change. Further, we restrict the consideration to the special case where each user is to be scheduled exactly once during the given time frame.

Assuming that K_group is an integer divider of K_total, this means that the entire user set {1, 2, . . . , K_total} must be partitioned into N=K_total/K_group subsets of size K_group. The users within any subset are then served concurrently, while different subsets are assigned e.g. different timeslots. Let the best partition be denoted by P={G₁G₂, . . . , G_N} where |G_n|=K_group, ∀n. A natural metric to measure the fitness of a partition is the average of the group-wise metrics as defined in equation (7), namely:

$\begin{array}{cc}{M}_{\mathrm{partitian}}\left(\right)\stackrel{\mathrm{def}}{=}\frac{1}{N}\sum _{n=1}^NM_{\mathrm{group}}\left({}_n\right)& \left(8\right)\end{array}$

The best partition must then fulfill: P=arg_{x—a—valid—partition} max M_partition(χ). The number of candidate partitions for the maximization is

$\frac{K_{\mathrm{total}}!}{N!{\left(K_{\mathrm{group}}!\right)}^N}.$

By combining equations (8) and (7), it can be seen that the optimal choice P is independent of ρ_ref. Thus, the scheduling decision made by the best partition approach does not depend on this parameter.

FIGS. 3 and 4 illustrate the performance of a system using the proposed scheduling metric over a system with random scheduling. The curves are for a case with two base station antennas and two active user terminals out of six total user terminals, where each user terminal has a single antenna. The parameters used are M_R=2, M_T,k=1, K_total=6K_group=2, and ρ_ref=−10 dB. A channel model, SCM C2 (Urban macro) with ν_mobile=3 km/h was used for the figures. The applied modulation is QPSK (Quadrature Phase Shift Keying) and the channel code is a standard 3GPP turbo code with rate ½. An exhaustive search was used. FIG. 3 illustrates uncoded bit error rate curves while FIG. 4 presents the overall system throughput. Based on the figures, it can be seen that the performance of the system using the scheduling metric according to an embodiment is superior over a system using random scheduling.

Finally, FIG. 5 illustrates a scheduling method according to an embodiment. The method starts in 500. In 502, a set of user terminals to be served is determined in a base station. In 504, a second or higher order statistics is estimated for each user terminal of the determined set of user terminals. In 506, a scheduling metric is estimated on the basis of the estimated second or higher order statistics. In 508, scheduling is controlled on the basis of the estimated scheduling metric. The method ends in 510.

The embodiments of the invention may be realized in a radio system module comprising a processing unit. The processing unit may be configured to perform at least some of the steps described in connection with the flowchart of FIG. 5. The embodiments may be implemented as a computer program comprising instructions for executing a computer process for controlling scheduling.

The computer program may be stored on a computer program distribution medium readable by a computer or a processor. The computer program medium may be, for example but not limited to, an electric, magnetic, optical, infrared or semiconductor system, device or transmission medium. The computer program medium may include at least one of the following media: a computer readable medium, a program storage medium, a record medium, a computer readable memory, a random access memory, an erasable programmable read-only memory, a computer readable software distribution package, a computer readable signal, a computer readable telecommunications signal, computer readable printed matter, and a computer readable compressed software package.

Even though the invention has been described above with reference to an example according to the accompanying drawings, it is clear that the invention is not restricted thereto but it can be modified in several ways within the scope of the appended claims.

Claims

1. A scheduling method comprising: providing an estimated second or higher order statistics for each user terminal in a set of user terminals to be served;estimating a scheduling metric based on the estimated second or higher order statistics; andcontrolling scheduling based on the estimated scheduling metric.

2. The method of claim 1, wherein the estimated second or higher order statistics for each user terminal k has the form of:

3. The method of claim 1, further comprising: estimating an ergodic sum capacity of multiple access channels based on the second or higher order statistics; andestimating the scheduling metric based on the estimated ergodic sum capacity.

4. The method of claim 3, further comprising estimating the ergodic sum capacity of multiple access channels by using a second-order Taylor series of the ergodic sum capacity.

5. The method of claim 1, wherein the estimated scheduling metric has the form of:

6. The method of claim 5, further comprising: precomputing trace expressions for each user terminal or user terminal pair before estimating the scheduling metric.

7. The method of claim 1, wherein the estimated scheduling metric has the form of:

8. The method of claim 1, further comprising: using a scheduling algorithm for controlling scheduling based on the estimated scheduling metric.

9. The method of claim 8, further comprising: using an opportunistic scheduling algorithm for controlling scheduling.

10. The method of claim 8, further comprising: using a round-robin based fair scheduling for controlling scheduling.

11. A radio system comprising: a base station;one or more user terminals communicating with the base station;an estimation unit for estimating a second or higher order statistics for each user terminal in a set of user terminals to be served;an estimation unit for estimating a scheduling metric based on the estimated second or higher order statistics; anda scheduling unit for controlling scheduling based on the estimated scheduling metric.

12. The radio system of claim 11, further comprising: an estimation unit for estimating an ergodic sum capacity of multiple access channels based on the second or higher order statistics; andan estimation unit for estimating the scheduling metric based on the estimated ergodic sum capacity.

13. The radio system of claim 11, wherein the estimated scheduling metric has the form of:

14. A base station comprising: an estimation unit for estimating a second or higher order statistics for each user terminal in a set of user terminals to be served;an estimation unit for estimating a scheduling metric based on the estimated second or higher order statistics; anda scheduling unit for controlling scheduling based on the estimated scheduling metric.

15. The base station of claim 14, wherein the estimated second or higher order statistics of each user terminal k has the form of:

16. The base station of claim 14, further comprising: an estimation unit for estimating an ergodic sum capacity of multiple access channels based on the second or higher order statistics; andan estimation unit for estimating the scheduling metric based on the estimated ergodic sum capacity.

17. The base station of claim 16, wherein the estimation unit for estimating the ergodic sum capacity of multiple access channels is configured to use a second-order Taylor series of the ergodic sum capacity for estimating the ergodic sum capacity.

18. The base station of claim 14, wherein the estimated scheduling metric has the form of:

19. The base station of claim 18, further comprising an estimation unit for precomputing trace expressions for each user terminal or user terminal pair before estimating the scheduling metric.

20. The base station of claim 14, wherein the estimated scheduling metric has the form of:

21. The base station of claim 14, wherein the scheduling unit is configured to use a scheduling algorithm for controlling scheduling based on the estimated scheduling metric.

22. A scheduling module for a radio system, comprising: an estimation unit for estimating a scheduling metric based on second or higher order statistics of each user terminal in a set of user terminals to be served; anda control unit for providing information for controlling scheduling based on the estimated scheduling metric.

23. The scheduling module of claim 22, wherein the estimated second or higher order statistics of each user terminal k has the form of:

24. The scheduling module of claim 22, further comprising: an estimation unit for estimating an ergodic sum capacity of multiple access channels based on the second or higher order statistics; andan estimation unit for estimating the scheduling metric based on the estimated ergodic sum capacity.

25. The scheduling module of claim 24, wherein the estimation unit for estimating the ergodic sum capacity of multiple access channels is configured to use a second-order Taylor series of the ergodic sum capacity for estimating the ergodic sum capacity.

26. The scheduling module of claim 22, wherein the estimated scheduling metric has the form of:

27. The scheduling module of claim 26, further comprising an estimation unit for precomputing trace expressions for each user terminal or user terminal pair before estimating of the scheduling metric.

28. The scheduling module of claim 22, wherein the estimated scheduling metric has the form of:

29. The scheduling module of claim 22, wherein the scheduling unit is configured to use a scheduling algorithm for controlling scheduling based on the estimated scheduling metric.

30. The scheduling module of claim 29, wherein the scheduling unit is configured to use an opportunistic scheduling algorithm for controlling scheduling.

31. The scheduling module of claim 29, wherein the scheduling unit is configured to use a round-robin based fair scheduling for controlling scheduling.

32. A scheduling module for a radio system, comprising: estimation means for estimating a scheduling metric based on second or higher order statistics of each user terminal in a set of user terminals to be served; andprocessing means for providing information for controlling scheduling based on the estimated scheduling metric.

33. The scheduling module of claim 32, wherein the estimated scheduling metric has the form of:

34. A computer program on a computer-readable medium having computer executable components for executing a computer process for scheduling, the process comprising: providing an estimated second or higher order statistics for each user terminal in a set of user terminals to be served;estimating a scheduling metric based on the estimated second or higher order statistics; andcontrolling scheduling based on the estimated scheduling metric.

35. The computer program of claim 34, wherein computer-readable medium comprises at least one of the following media: a program storage medium, a record medium, a computer readable memory, a computer readable software distribution package, a computer readable signal, a computer readable telecommunications signal, and a computer readable compressed software package.