METHOD AND ARRANGEMENT FOR IMPROVED CHANNEL-DEPENDENT TIME-AND- FREQUENCY-DOMAIN SCHEDULING

Abstract

In a method of improved channel-dependent time- and frequency-domain scheduling in an OFDM based telecommunication system with multiple user terminals, determining SO a parameter value representative of the system load; pre-selecting SI a subset of user terminals if the determined parameter value is larger than or equal to a predetermined threshold; and performing S2 frequency-domain scheduling of the pre-selected subset, to reduce the downlink signaling overhead and enabling improved efficiency of the channel-dependent time- and frequency-domain scheduling.

Description

TECHNICAL FIELD

The present invention relates to telecommunication systems in general, and specifically to methods and arrangements for improved channel-dependent time-and-frequency-domain scheduling in an Orthogonal Frequency Division Multiplex (OFDM) based telecommunication system.

BACKGROUND

In the area of user resource allocation for wireless fading channels, great effort is presently focused on scheduling [1-2]. Scheduling algorithms can typically be classified into channel-dependent or channel-independent scheduling according to the dependence on the channel. An example of such an algorithm is Round Robin (RR), a typical channel-independent scheduling, which benefits from simplicity at the price of poor performance.

The class of channel-dependent scheduling algorithms utilize the so-called channel state information (CSI) or channel quality indicator (CQI) in order to improve the system performance.

For OFDM systems, the above mentioned channel-dependent scheduling can be further classified into so called time-domain scheduling, where a single user or user terminal per frame is scheduled in a given time scale, and so called time-and-frequency-domain scheduling, where multiple users per frame are scheduled exclusively in a given time scale. The time-and-frequency-domain scheduling, hereinafter referred to as frequency-domain scheduling, has previously been shown to provide better performance than the time-domain scheduling due to the multi-user diversity in the frequency domain, especially for wideband transmissions [1]. However, the frequency-domain scheduling requires CSI or CQI feedback once per frequency-domain resource unit, which requires extensive overhead signaling that is much higher than that for time-domain scheduling, i.e. one feedback for the whole band at a time. In addition, there are many different detailed criteria for the frequency-domain scheduling, such as Max-CIR, Proportional-Fair (PF), weighted-queue-PF etc [3], for both frequency-domain and time-domain scheduling.

For the class of channel-dependent scheduling algorithms the time-domain scheduling has the advantages of low computational complexity and low signaling overhead (for it self and power allocation, link adaptation afterwards as well). However, due to the frequency-selectivity along the wideband, the time-domain scheduling cannot guarantee that the scheduled user performs well on the whole band, therefore, can hardly achieve good performances in capacity and coverage.

Frequency-domain scheduling schemes perform the criteria in the more refined sub-group (e.g. chunk) of the whole band, and utilize the multi-user diversity as well, so that the performances in capacity and coverage are greatly improved as compared to the time-domain scheduling schemes.

However, the disadvantages of the otherwise advantageous frequency-domain scheduling increase as the performance improves. Specifically, the computational complexity increases greatly with the number of chunks and the system load. In addition, since the scheduled user terminals may be different from one chunk to another, a large quantity of DL signaling is required for the frequency-domain adaptation (FDA), including the chunk allocation and the subsequent power allocation and link adaptation per user. The signaling overhead thus increases linearly with the increasing bandwidth, i.e. with the number of chunks, and with the system load, i.e. the number of users.

These disadvantages have prevented, up until now, the further exploitation of channel-dependent time-and-frequency domain scheduling.

Consequently, there is a need for methods and arrangements enabling exploiting the advantages of channel-dependent time- and frequency domain scheduling whilst at the same time reducing the known disadvantages.

SUMMARY

A general problem with known channel-dependent scheduling algorithms is how to utilize the advantageous performance of frequency domain scheduling but without the above described disadvantages.

A general object of the present invention is to provide methods and arrangements for improved channel-dependent frequency-domain scheduling.

These and other objects are achieved according to the attached set of claims.

According to a basic aspect, the present invention comprises determining the load of the system. If the load equals or surpasses a predetermined threshold value, a subset of all the user terminals are pre-selected for scheduling and subsequently scheduled.

Advantages of the present invention comprise:

• • Reduced overhead downlink signaling for channel-dependent frequency-domain scheduling.
  • Reduced overhead signaling for resource allocation.
  • Reduced overhead signaling for link adaptation
  • The complexity of the corresponding frequency-domain adaptation can be limited within a pre-defined scale.
  • The scales of signaling overhead and computational complexity can be pre-defined as fixed, regardless of varying system load, for the given bandwidth, which is favoured by the further signaling design.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention, together with further objects and advantages thereof, may best be understood by referring to the following description taken together with the accompanying drawings, in which:

FIG. 1 is a schematic flow diagram of an embodiment of a method according to the invention;

FIG. 2 is a schematic illustration of an embodiment of an arrangement according to the invention;

FIG. 3 shows the impact for a first cell load of the implementation of an embodiment of the method and arrangement according to the invention;

FIG. 4 shows the impact for a second cell load of the implementation of an embodiment of the method and arrangement according to the invention.

ABBREVIATIONS

• BPSK Binary Phase Shift Keying
• CIR Carrier to Interference Ratio
• CQI Channel Quality Indicator
• CSI Channel State Information
• DL Down Link
• FDA Frequency Domain Adaptation
• HSDPA High-Speed Downlink Packet Access
• LA Link Adaptation
• MCS Modulation and Coding Scheme
• OFDM Orthogonal Frequency Division Multiplex
• PA Power Allocation
• PF Proportional Fair
• PFT Proportional Fair in the Time domain
• PFTF Proportional Fair in the Time and Frequency domain
• QAM Quadrature Amplitude Modulation
• QPSK Quadrature Phase Shift Keying
• RR Round Robin
• SINR Signal to Interference Noise Ratio

DETAILED DESCRIPTION

To provide a further insight into the disadvantages of channel-dependent frequency-domain scheduling as compared to channel-dependent time-domain scheduling a further detailed discussion and analysis is provided below.

The frequency-domain adaptation (FDA) related signaling overhead in the downlink (DL) consists of the signaling used to inform FDA decisions for each user or user terminal, where the pilot signals for signal to interference noise ratio (SINR) estimation are assumed to be the same for all the schemes, therefore not taken into account in the following.

As an illustrative example, consider a system with N_ch chunks in the DL to serve N users per cell, where the number N_bitsDCCH of signaling bits per frame is calculated according to the following:

Time-Domain Scheduling: (Round Robin (RR) or Proportional-Fair (PF) in Time Domain)

• • For user index with uniform power allocation (PA): 1×┌ log₂ N┐bits/frame, where ┌x┐ is the minimum integer that is equal to or larger than x.
  • For user index with On/Off PA: 1×┌ log₂ N┐+1×N_ch bits/frame
  • For link adaptation (LA): k_mod+k_cr bits/frame, e.g. for 20 MHz bandwidth
    • Modulation mode index among {BPSK, QPSK, 16QAM, 64QAM}: k_mod=2 bits/frame
    • Index for coding rate, corresponding to information block size: k_cr=f(N_symsperframe, ModOrder) (e.g. at most 19 bits/frame for 20 MHzm while for HSDPA it is about 6 bits/frame)

For the example for RR time-domain scheduling, uniform power allocation and single modulation and code scheme (MCS) with continuous coding rate, the considered number of DL signaling bits N_bitsDCCH becomes:

N _bitsDCCH=┌ log₂(N)┐+(k_mod+k_cr) [bits/frame],  (1)

Frequency-Domain Scheduling

• • For user index per chunk with uniform PA: N_ch×┌ log₂ N┐ bits/frame;
  • For user index per chunk with On/Off PA: N_ch×┌ log₂ N┐+N_ch×1 [bits/ frame]
  • For LA: f(N)=k_mod·N+k_cr·N bits/frame, depending on how many users are to be scheduled
  • For the example of Proportional Fair in Time and Frequency domain (PFTF), uniform PA and single MCS with continuous coding rate, the total DL FDA signaling N_bitsDCCH is determined according to:

N _bitsDCCH =N _ch┌ log₂(N)┐+N×(k_mod+k_cr) [bits/frame]  (2)

• •  which is substantially larger than that of time-domain scheduling. As can be seen in Equation (2) the scheduling algorithm is the largest contributor to the DL overhead signaling.

Accordingly, as recognized by the inventors, if the DL signaling of FDA can be reduced, the efficiency and usefulness of channel-dependent frequency-domain scheduling can be further improved.

A basic embodiment according to the invention thus provides a method of pre-selecting a subset of the active user terminals in a system and performing frequency-domain scheduling, link adaptation and resource allocation for that subset. The remaining set of user terminals are processed subsequently, thereby significantly reducing the DL overhead signaling in each time instance.

A specific embodiment of a method according to the invention will be described below with reference to the schematic flow diagram of FIG. 1. Initially, a present load of the system is determined S0 and compared S1 to a preset threshold. The load can be determined by measuring some measure or parameter value that is dependent on the total load, i.e. relative signaling overhead for all active users. Consequently, according to a specific embodiment, the above mentioned preset threshold is a specific value of the signaling overhead i.e. 10%.

If the determined load of the system is larger than or equal to the predetermined threshold, pre-selection is deemed necessary, and a subset of user terminals are pre-selected S2 for scheduling. The subset of user terminals comprises at least two user terminals and less than all terminals.

Finally, the subset of user terminals is subjected to scheduling S3 and optionally link adaptation and resource allocation according to known measures.

If the threshold is not surpassed, then all user terminals are scheduled in a known manner.

According to a specific embodiment, the pre-selection process can optionally be repeated for a plurality of subsets until the measured load of the system is below the preset threshold or based on some other criteria.

There are several potential criteria for pre-selecting the above mentioned subset of user terminals:

• • Random selection: pre-select the user terminals randomly;
  • Max-CIR selection: according to a Max-CIR criterion

$\begin{array}{cc}J=\arg \underset{1\le n\le N}{\max }{\gamma }^{\left(n\right)},& \left(3\right)\end{array}$

where γ⁽ⁿ⁾ denotes the estimated SINR of user n, the user(s) with the highest SINR(s) according to Equation (3) are selected frame-wise or chunk-wise.

• • PF-based selection: according to the PF criterion

$\begin{array}{cc}J=\arg \underset{1\le n\le N}{\max }\frac{{\mathrm{TP}}_{\mathrm{cs}1}^{\left(n\right)}}{{\mathrm{TP}}_{\mathrm{av}}^{\left(n\right)}}.& \left(4\right)\end{array}$

where TP_est⁽ⁿ⁾ denotes the estimated throughput of user terminal with index n and TP_av⁽ⁿ⁾ stands for the average throughput of user terminal n each frame, the user(s) with the highest ratio(s) of (4) are selected frame-wise.

There may also be other pre-selection criteria based on chunk-wise PF or including other cost functions or quality of service (QoS) for each user terminal.

By utilizing the step of pre-selection, the users are limited to a pre-defined scale in order to reduce the corresponding DL signaling.

Among the pre-selected users, the frequency-domain scheduling, power allocation and link adaptation can be further performed. The non-elected or discarded user terminals can optionally be queued until the next round of signal processing. In this manner the number of bits for the DL FDA signaling of Equation (2) for the simplified PFTF schemes becomes

N _bitsDCCH =N _sel×┌ log₂(N)┐+N_ch×┌ log₂(N_sel)┐+N_sel×(k_mod+k_cr) [bits/frame]  [5]

where the first term of Equation (5) corresponds to the user index of the pre-selected user terminals.

To illustrate the benefits of the method according to the invention further, a few simulation experiments are presented and described below.

The basic simulation parameters are summarized below in Table 1.

TABLE 1
Basic simulation parameters
Cell Plan	Number of sites	7
	Sectors/site	1
	Frequency reuse	1
	Cell radius [meter]	{500, 1000, 2000, 3000}
System	Duplex mode	FDD
Assumption	Channel model	3GPP SCME, Suburban macro
	Bandwidth [MHz]	20
	Available subcarriers/	1280
	tones
	Chunk size	16
	[tones per chunk]
	Traffic model	Full-buffer traffic model
	Transmitter/receiver	SISO, omni
	antennas
	Transmission	80, uniform power allocation
	power [watt.]
	Average offered	8, 30
	calls per cell
Transmission	Modulation and coding	Single MCS per frame
schemes	scheme (MCS)
	Modulation	BPSK, QPSK, 16QAM, 64QAM
	Link adaptation (LA)	BLER_target = 0.1
	Coding rate	Continuous
Remarks: (*) the same MCS for all chunks allocated to one user in an OFDM frame

For the simulations the Priority-Fair (PF) criteria in the time-domain (PFT) scheduling represents the channel-dependent time-domain scheduling. The PF in time and frequency domain (PFTF) represents the channel-dependent frequency-domain scheduling. For pre-selection schemes, random selection and PFT selection are considered. Therefore, in the following description four schemes are compared for illustrations, namely:

• • Pure time domain scheduling (PFT),
  • Frequency-domain scheduling (PFTF) with Random pre-selection (Rand+PFTF),
  • Priority-Fair (PFT) pre-selection in time domain+PFTF (PFT+PFTF) and
  • Frequency-domain scheduling (PFTF) without pre-selection

Two cases of different system load e.g. 8 and 30 user terminals are considered. However, the invention is not limited to those load scenarios. Moreover, for the load of 30 user terminals the impact of the pre-selection bound (4 and 8) is also shown. The DL signaling of the above schemes under different cases are listed below in Table 2.

TABLE 2
DL signaling of the four schemes
	User pre-selection + PFTF	PFTF
Cell load	Overhead	Relative	Overhead	Relative
[users/cell]	[bits/frame]	overhead	[bits/frame]	overhead
8	256 (bound 4)	5%	408	8%
30	264 (bound 4)	5.1%	1030	20%
	448 (bound 8)	8.7%

It can be seen that with any type of user pre-selection, the resulting DL signaling is fixed and much less than the pure frequency-domain scheduling PFTF. Especially in the case of high load, e.g., 30, the DL signaling with user pre-selection is even smaller than the pure PFTF scheme. The reduction ratios (to the signaling of the pure PFTF) are 74% (bound of 4) and 57% (bound of 8), respectively, which are very remarkable.

It also implies that the resulting computational complexity is greatly reduced by the application of user pre-selection according to the invention.

On the other hand, the slight performance loss as the price of signaling reduction deserves observation. FIGS. 2 and 3 depict the performances of the aforementioned schemes in the average cell throughput and the 5^th percentile user throughput.

It can be seen that the PFT+PFTF scheme shows a little worse performance than the pure PFTF. Especially in the high load case, the PFT+PFTF scheme with bound of 8 shows very close performance to the pure PFTF.

Of course, there are other options for user pre-selection, which might have even better performance than the ones shown. Therefore, user pre-selection provides the potential to further improve the system performance with limited DL signaling.

A node according to an embodiment of the present invention is configured for enabling the above discussed method according to the invention, and will be described with reference to FIG. 4. The node can, according to a specific embodiment, be represented by but not limited to a Node B in a UMTS telecommunication network.

A basic embodiment of a node 1 according to the invention comprises a load determination unit 10 which enables measuring or at least acquiring a measure of the present load in the system, a comparing unit 11 that compares the load measure to a preset load measure threshold, and a pre-selection unit 12 that pre-selects a subset of user terminals for scheduling if the load measure surpasses the threshold value. Finally, the node 1 comprises a scheduling unit 13 for scheduling the pre-selected user terminals.

If the threshold is not surpassed by the present load, the scheduling unit 13 is adapted for scheduling all user terminals in a known manner.

Advantages of the Invention comprise:

• • The DL signaling overhead, which is related to channel-dependent frequency-domain scheduling is greatly reduced by limiting the number of users for scheduling;
    • Less signaling overhead is required by frequency-domain resource allocation due to the reduced amount of users;
    • Less signaling overhead is required by link adaptation due to the reduced amount of users;
  • The complexity of the corresponding frequency-domain adaptation can be limited within a pre-defined scale;
  • The scales of signaling overhead and computational complexity can be pre-defined as fixed, regardless of varying system load, for the given bandwidth, which is favored by the further signaling design.

It will be understood by those skilled in the art that various modifications and changes, including combinations of various embodiments, may be made to the present invention without departure from the scope thereof, which is defined by the appended claims.

REFERENCES

• [1] R. Knopp and P. Humblet, Information capacity and power control in single-cell multiuser communication, in Proc. of Int. Conf. on Communication, June 1995, vol. 1, pp. 331-335.
• [2] W. Anchun, X. Liang, Z. Shidong, X. Xibin, and Y. Tan, Dynamic resource management in the fourth generation wireless systems, Proceedings IEEE International Conference on Communication Technology, April 2003.
• [3] K. Muhammad, Comparison of Scheduling Algorithms for WCDMA HS-DSCH in different Traffic Scenarios, IEEE PIMRC'2003 Beijing, China 2003.

Claims

1-16. (canceled)

17. A method of improved channel-dependent time- and frequency-domain scheduling in an Orthogonal Frequency-Division Multiplexing (OFDM) based telecommunication system, said system comprising a plurality of user terminals, the method comprising: determining a parameter value representative of a system load;pre-selecting a subset of said plurality of user terminals, the subset being limited to a pre-defined scale, if said determined parameter value is larger than or equal to a predetermined threshold;performing frequency-domain scheduling of said pre-selected subset of user terminals to reduce the downlink signalling overhead and to enable improved efficiency of the channel-dependent time- and frequency-domain scheduling.

18. The method according to claim 17, comprising repeating the determining, pre-selecting and frequency-domain scheduling for the remaining user terminals.

19. The method according to claim 17, where said load parameter comprises a relative signalling overhead for all user terminals of the plurality of user terminals.

20. The method according to claim 19, where said predetermined threshold comprises a relative signalling overhead of 10%.

21. The method according to claim 17, comprising pre-selecting said subset randomly from said plurality of user terminals.

22. The method according to claim 17, comprising pre-selecting said subset based on a Max-CIR criterion.

23. The method according to claim 22, comprising pre-selecting user terminals with the highest signal to interference noise ratio.

24. The method according to claim 23, comprising pre-selecting said subset frame-wise and chunk-wise.

25. The method according to claim 17, comprising pre-selecting said subset based on a ratio between an estimated throughput and an average throughput for each user terminal.

26. The method according to claim 25, comprising pre-selecting user terminals with the highest ratio.

27. The method according to claim 17, comprising pre-selecting said subset based on one or a combination of throughput, quality of service.

28. The method according to claim 17, where said subset of user terminals comprises at least two user terminals and not all user terminals.

29. A node in an Orthogonal Frequency-Division Multiplexing (OFDM) communication system, said system comprising a plurality of user terminals, said node comprising: means for determining a parameter value representative of a system load;means for pre-selecting a subset of the plurality of user terminals, the subset being limited to a pre-defined scale, if said determined parameter value is larger than or equal to a predetermined threshold value;means for performing frequency-domain scheduling of said pre-selected subset of user terminals to reduce the downlink signalling overhead and to enable improved efficiency of channel-dependent frequency domain scheduling.

30. The node according to claim 29, further comprising means for repeating said steps of pre-selection and scheduling for the remaining user terminals.

31. The node according to claim 29, where said pre-selection means are adapted for pre-selecting said subset of user terminals based on one or a combination of signal to interference noise ratio, randomly, max-CIR, quality of service, throughput.

32. The node according to claim 29, where said node is a Node B in a Universal Mobile Telecommunications System (UMTS).

33. The node according to claim 30, where said node is a Node B in a Universal Mobile Telecommunications System (UMTS).

34. The node according to claim 31, where said node is a Node B in a Universal Mobile Telecommunications System (UMTS).