WO2011028368A1

WO2011028368A1 - Uplink resource assignment in a wireless communication network

Info

Publication number: WO2011028368A1
Application number: PCT/US2010/044939
Authority: WO
Inventors: Dan Tayloe; Shubhodeep Adhikari; Shalini Gulati
Original assignee: Motorola Mobility, Inc.
Priority date: 2009-09-02
Filing date: 2010-08-10
Publication date: 2011-03-10

Abstract

An apparatus and method for uplink resource assignment in a wireless communication network includes a first step (400) of queuing requests for uplink resources for a subframe. A next step (402, 404, 406) includes selecting users, based in part on their channel conditions, into a linear superposition of fractional groups having similar channel conditions. A next step (410) includes assigning uplink resources for the subframe after the selected users have been selected for the available resources in the subframe.

Description

UPLINK RESOURCE ASSIGNMENT IN A WIRELESS COMMUNICATION

NETWORK

FIELD OF THE INVENTION

The present invention relates generally to wireless radio communication and, in particular, to uplink resource assignment in a wireless communication network.

BACKGROUND OF THE INVENTION

In 4G communication systems, such as the Long Term Evolution (LTE) and Worldwide Interoperability for Microwave Access (WiMAX) communication systems, system operators attempt to keep subscribers satisfied with the communication throughput that they receive. Subscribers that get the lowest throughput are generally those with the poorest signal conditions, such as edge-of-cell users. Improving the throughput of these users implies providing them with more resources. Since more resources are used for low throughput edge-of-cell users, this implies that the overall sector communication throughput will go down.

In practice, scheduling algorithms attempt to maximize the edge-of-cell user throughput, while simultaneously keeping the fifty-percentile user throughput high, and the sector total throughput as high as possible in order to increase the capacity of the cell. One solution for this has been to use a "proportional fair" (PF) algorithm to schedule users for uplink transmission. However, the PF algorithm often fails to meet the above defined goals.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention is pointed out with particularity in the appended claims. However, other features of the invention will become more apparent and the invention will be best understood by referring to the following detailed description in conjunction with the accompanying drawings in which:

Figure 1 illustrates an example of a communication system in accordance with the present invention; Figure 2 is a graphical representation of a first simulation example in accordance with the present invention;

Figure 3 is a graphical representation of a second simulation example in accordance with the present invention; and

Figure 4 illustrates an example of a method in accordance with the present invention.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

The present invention provides significantly better uplink resource allocation than the proportional fair algorithm. In particular, the present invention uses three different user selection techniques. These three techniques are; Head Of Queue, Windowed Spectral Efficiency, and Windowed Better-Than-Average Radio Frequency (RF) Conditions, which are described below. Each of these three selection techniques is assigned a fraction of the total user uplink resource assignments. Inasmuch as the present invention assigns uplink resources using queues, the present invention is particularly suited for non-delay sensitive traffic.

The following description focuses on embodiments of the invention applicable to 4G communication systems such as LTE and WiMAX. For example, the present invention can be implemented for LTE evolved NodeBs (eNB). The present invention could also be applied to the WiMAX base stations. However, it will be appreciated that the invention is not limited to these applications but may be applied to many other cellular communication systems such as a 3 GPP (Third Generation Partnership Project) E-UTRA (Evolutionary UMTS Terrestrial Radio Access) standard, a 3GPP2 (Third Generation Partnership Project 2) Evolution communication system, a CDMA (Code Division Multiple Access) 2000 1XEV-DV communication system, a Wireless Local Area Network communication system as described by the Institute of Electrical and Electronics Engineers 802.xx standards, for example, the 802.1 la/HiperLAN2, 802.1 lg, 802.16, or 802.21 standards, or any of multiple other proposed ultrawideband communication systems. Therefore, as used herein the term evolved NodeB can also represent a base station, access point, NodeB, or other similar device, and the term user equipment (UE) can also represent a mobile station, subscriber station, access terminal, and the like.

Figure 1 shows a communication network in accordance with the present invention. An eNB 100 is serving one or more UE 102. During a communication time slot (i.e., uplink subframe or transmission time interval (TTI)), one or more UE will request 104 uplink access to the eNB through the transceiver 108 to the processor 110. The processor 110 of the eNB allocates 106 uplink resources (i.e., physical resource blocks or PRBs) to these requesting UEs 102 through the transceiver 108, in accordance with the present invention, which defines how the uplink resources are assigned among three selection techniques: Head Of Queue, Windowed Spectral Efficiency, and Windowed Better-Than- Average Radio Frequency Conditions. This assignment of these three selection techniques is known to the eNB 100 and stored in memory 112. Specifically, the present invention provides a user section technique that allocates user data bearing resources after a subset of schedulable users have been selected, using the three selection techniques as follows.

Head Of Queue: For a given user queue, the first user in the queue is selected for uplink transmission. If at least one complete higher level packet is finished being transmitted by the first user, and the user has more data to transmit, that user is moved to the back of the queue to wait for more resources to continue transmissions, otherwise the user stays at the front of the queue until one complete higher level packet is delivered on the uplink. This is an optional feature. Otherwise, even with a partial packet sent, the user is sent to the back of the queue after scheduling in order to wait his turn again. If the user has completed sending its last packet, it is removed from the queue. Moving a user to the back of the queue happens when not all the data in the queue gets sent. This can occur under several conditions. First, the user has a lot of data and simply was not able to get it all sent. Secondly, there may have been a lot of competition among other users for RF resources so that the number of PRBs the first user received was not large enough to carry the complete message. Finally, the first user may be in poor RF condition such that it can only be granted a few PRBs that are encoded to the degree that each PRB contains relatively small amounts of data. In the uplink, the UE power guides how many uplink resources the UE can support. A UE in excellent RF coverage can support up to as many PRBs as the system has. In such a case, each PRB could be minimally encoded so that each PRB can carry a relatively large amount of data. In poor RF environments, the UE power is stretched to support only one or two PRBs which are highly encoded, supporting relatively few data bits per PRB. The present invention will preferably allocate as many uplink users as possible. This means that fewer PRBs will be allocated per user, which implies that the UE will be able to split its relatively small amount of uplink power (e.g., 200 mW) across fewer PRBs, which in turn means every PRB assigned gets higher than normal throughput. Higher throughput per PRB in turn implies higher total sector throughput. Also, a user in the worst RF environment with the power to support only one (or a few) PRBs that are highly coded will receive a relatively low uplink throughput. They will only actually realize this relatively low throughput if they get their one (or a few) PRBs every TTI (1 millisecond frame). Scheduling as many uplink users as possible each time maximizes the power per PRB and thus throughput per PRB and means that these edge of cell, poor RF users will be scheduled more often and thus get better edge of cell throughput.

Windowed Spectral Efficiency (Best RF users): N users with the very best RF conditions are selected from a set of M users. The selected users are kept in the same relative location in the queue after completing an uplink transmission unless a user is at the front of the queue. Selected users are not moved to the end of the queue unless they are at the front of the queue. The next N users are selected from the next M users. The set of M users wraps around to the front of the queue if and when the set of M users gets to the end of the queue. The preferred embodiment sets M at 15 users and N at 1 to 3 users.

Windowed Better-Than- Average Radio Frequency Conditions: The average RF condition of each user is maintained. In particular, an RF metric (specifically, Max Uplink C/I Capability) is tracked and averaged. This metric is updated periodically (nominally every ten milliseconds). An individual user's instantaneous ten millisecond metric is compared to the averaged metric for that user, and within a window of Y UEs (nominally 15 users), the X users (nominally 1 to 3 users) with the largest instantaneous/average RF difference (best "better-than-average" RF) are chosen. This is considered a fairness metric since it is assumed that all users are fading and that over time, all users will be selected as better-than-average RF as many times as other users, although this is not strictly true. By scheduling a user only when that user is in better-than-average RF conditions, a better-than-average throughput will be obtained from the PRBs assigned to that user. As a next step, the difference between the last known RF condition and the average RF condition is calculated. The X users with the largest delta ("better than average") are selected from a set of Y users. Selected users are not moved to the end of the queue unless they are at the front of the queue. The next X users are then selected from the next set of Y users. The set of Y users wraps around to the front of the queue if and when the set of Y users gets to the end of the queue.

Each of these three selection techniques is assigned a fraction of the total user uplink resource assignments, which can be fixed or changed dynamically. One preferred mix is 30% windowed spectral efficiency user picks, 50% windowed better- than-average RF condition user picks, and 20%> head of queue user picks. For example, if ten users were going to be picked using the above ratios, three would be picked using the spectral efficiency algorithm, five would be picked using the better- than average algorithm, and two would be picked using head of queue algorithm.

The windowed aspect of the above algorithm improves searching speed for picking users by preventing the calculation of these metrics across all users. Rather than searching potentially up to 800 users for the best better-than-average RF user, a pick is made using a "window" of users (e.g., fifteen) at a time. It is much faster to search fifteen users for the best one (or two or three) picks than searching all 800 users. Any users that get skipped can be passed to the next subframe. In addition, the maximum number of users per subframe that can be selected is limited by the Physical Downlink Control Channel (PDCCH) capacity. Maximizing the number of uplink users per subframe forces each user to use fewer resources on average, which increases the amount of power available per resource, which in turn increases the overall sector throughput over selecting fewer users. The combination of the three selection techniques provides an advantage over any single one of the selection techniques. For example, a user experiencing poor RF conditions will tend to be picked more often by the "head-of-queue" process since that user will likely not have good enough RF conditions to send an entire upper layer packet, such as a 1460 byte TCP packet. As a result, this user is kept at the front of the queue until a packet boundary is crossed allowing it to get at least one complete packet out in a timely manner and tends to give a bit of a throughput boost to such poor RF users, e.g., edge-of-cell users. The other two selection techniques, and the PF technique, favor users with average or better RF conditions.

Users in good to very good RF conditions tend to be in line-of-site environments that have little to no fading. The PF algorithm is tuned to look for these "better-than-average" users while users that are not better than average (i.e., poor RF conditions) largely tend to get ignored. In a real world system, these good to very good RF condition users are the very best users. In the present invention, these users will get picked up by both the "spectrally efficient" (Best RF) as well as the "head-of- queue" user selection techniques.

The "windowed" aspect of both the "better-than-average" and "spectrally efficient" algorithms can substantially reduce the processing overhead of computing selection metrics over a possible user population of hundreds of users. "Better-than- average-RF" is normally the best way of selecting users, but the other two selection techniques help the particular cases of very good RF users who tend to have little or no fading, and edge-of cell poor RF condition users. The window size (M or Y) chosen and the number of picked users (N or X) can be fixed or dynamically altered. In addition, M can equal Y, and N can equal X.

Example

Referring to Figure 2, a computer simulation was performed for edge-of cell performance using the present invention as compared to the existing proportional fair algorithm. The model uses Uplink Single Sector 100 User Sector Throughput vs. 5% User Throughput for 3 Kmph & Case 1 Distances, as is known in the art. As can be seen the present invention gives both better edge-of-cell (vertical axis) and total sector throughput (horizontal axis) than the PF algorithm.

In Figure 3, another simulation was performed for 50% user throughput performance using the present invention as compared to the existing PF algorithm. This model uses Uplink Single Sector 100 User Sector Throughput vs. 50% User Throughput for 3 Kmph & Case 1 Distances, as is known in the art. As can be seen the present invention gives both better 50% user throughput (vertical axis) and total sector throughput (horizontal axis) than the PF algorithm.

The PF calculation makes use of a and b coefficients to "tilt" the user selections to either favor poor RF users, favor best RF users, or provide for equal treatment between the two. Gamma is simply the ratio of a/b and is just another way to specify these parameters. A small gamma value tends to emphasize poor RF users, while a large gamma value tends to emphasize good RF users. Notice that a high gamma value tends to give best sector throughput due to the emphasis on good RF user, while the best edge of cell throughput is gained by using small values of gamma as poor RF users are emphasized over best RF users. In such a case the total cell throughput drops as is expected when edge of cell users are preferred over good RF users.

In the present invention, a fraction of the assignments are spectrally efficient, and a fraction are fairness user selections. The fairness fraction is given on the chart. This is one spectrally efficient user selection algorithm (windowed Best RF) and two fairness algorithms (Better-than-average RF and head of queue). Better-than-average- RF is a fairness pick since it is assumed that due to fading, everyone has an equal opportunity at being the best windowed "better-than-average" pick when better-than- average is measured on a user-by-user basis. Equal fading is not the case for the very best RF users though because users in the best RF conditions tend to be close to the site in line-of-sight conditions and as such don't tend to fade. These users do not tend to get picked by "better-than-average RF" but will get picked by the head of queue and best RF user selection algorithms. For this curve, the two fairness user selection algorithms were set to 70% better-than-average-RF and 30% head-of-queue whenever a fairness user selection method was used. On this curve, maximal fairness (Fair 1.0) produced the best edge of cell user throughput (as defined by the 5% user throughput) but sacrificed reduced total sector throughput. However, the Fair = 1.0 (100% fairness picks) point gives better edge of cell throughput than PF at a total sector throughput that is only matched by the best RF user point of the PF algorithm.

Overall, the range of operating points offered by the present invention allows higher throughput trades between edge of cell (5 percentile user throughput) and total sector throughput than does the more traditional proportional fair algorithm. It should be noted the present invention tends to get both better total sector throughput and edge of cell throughput.

Figure 4 illustrates a method for uplink resource assignment in a wireless communication network, particularly for non-delay sensitive traffic, in accordance with the present invention. Preferably, as many users as possible are assigned uplink resources. In particular, for the LTE communication system, there is a control channel used to make user uplink and downlink resource assignments. The channel (i.e., the PDCCH) can carry only a limited number of user assignments. Each user assignment on the PDCCH takes up a variable amount of space since users in very good RF conditions are encode more lightly and require only one control channel element (CCE). A user in a very poor RF environment requires heavier encoding to deliver the resource assignment in the poor RF conditions and may need up to 8 CCEs. Uplink or downlink resource assignments may require 1, 2, 4, or 8 CCEs. In a 10 MHz LTE bandwidth, there are approximately 41 CCEs available in a common pool to be used for uplink and downlink user assignments as well as power control messages, pages, system information messages. The present invention attempts to makes sure the reverse link is packed as full with uplink users as possible, alternating assignment of uplink and downlink PDCCH users.

The method includes a first step 400 of queuing requests for uplink resources for a subframe. In particular, the uplink resources are physical resource blocks. A next step 402, 404, 406 includes selecting users, based in part on their channel conditions, into a linear superposition of fractional groups having similar channel conditions.

A next step 402 includes selecting the first user of the queue for uplink resources, wherein if at least one complete packet is finished being transmitted on the uplink resources the user is moved to the back of the queue. Optionally, the first user stays at the front of the queue until one complete higher level packet is delivered on the uplink resource. At the very minimum, the user gets to send some data, and then is sent to the rear of the queue if the user still has data to send. It should be noted that just because a user gets selected, this does not mean that the user gets used. If it is time for a head-of-queue user to be selected, the user may not actually get used because the PDCCH channel "hashes" user assignments. Due to this hashing function, even if there is space on the PDCCH channel for an additional user, a particular user might not hash to that open spot. Thus the head-of-queue user could be rejected, although that user will keep his head-of-queue position, and the next user in the queue will be tested to see if it fits on the PDCCH. Given the difficulty in finding a last user to fit these remaining PDCCH assignment spots, there is a PDCCH assignment termination condition. There are two approaches considered in the present invention. One approach is to set a time limit on the uplink user search time. The second approach is to go until "Z-in-a-row" user assignments fail and then stop. A combination of these two approaches can be used (Z-in-a-row assignment failures plus a time limit). Other termination conditions are possible. For example, the algorithm must stop if there are no more users to check. If there are only ten users, the algorithm will end up stopping simply because it has checked all ten users for a fit on the PDCCH.

A next step 404 includes selecting at least one user, from among a window of users, experiencing the best channel conditions for uplink resources. The selected at least one user is kept in the same relative location in the queue after completing an uplink transmission unless one selected user is at the front of the queue, whereupon that one selected user is moved to the back of the queue, per step 402. A next step 406 includes selecting at least one user, from among a window of users, experiencing the highest positive difference between its present channel condition and its average channel condition for uplink resources. The selected at least one user is kept in the same relative location in the queue after completing an uplink transmission unless one selected user is at the front of the queue, whereupon that one selected user in moved to the back of the queue, per step 402.

A next step 408 includes repeating the two latter selecting steps 404, 406 for each of subsequent respective windows, wherein the windows of the selecting steps 404, 406 are allowed to wrap around to the front of the queue when the windows get to the end of their respective queues so as to maintain the given window size. Figure 4 is really a loop for steps 402, 404, 406, 408. Conceptually it can be thought of as a random variable which is then used to schedule one user based using one of these three selection techniques where the technique selected is based on the random number and the predefined percentages for the three techniques (such as 20%/30%/50% mentioned above). The process loops until the PDCCH is filled or the termination goal as discussed above is reached or the algorithm runs out of users to check. Note that 408 is part of 404 and 406. When a windowed 404 or 406 user selection is run to get one user, that one user returned might be the second or third result of a windowed selection if the goal was to pick the top two or three best users.

A next step 410 includes assigning uplink resources for the subframe after the selected users have been selected for the available resources in the subframe. Preferably, a fixed fraction of users are selected for each of the combined selecting steps 402, 404, 406.

Claims

A method for uplink resource assignment in a wireless communication network, the method comprising the steps of:

queuing (400) requests for uplink resources for a subframe;

selecting (402, 404, 406) users, based in part on their channel conditions, into a linear superposition of fractional groups having similar channel conditions; and

assigning (410) uplink resources for the subframe after the selected users have been selected for the available resources in the subframe.

The method of claim 1 wherein the uplink resources are physical resource blocks.

The method of claim 1 wherein a fixed fraction of users are selected for each of the fractional groups.

The method of claim 1 wherein the selecting step includes the steps of:

selecting the first user of the queue for uplink resources, wherein if at least one complete packet is finished being transmitted on the uplink resources the user is moved to the back of the queue;

selecting at least one user experiencing the best channel conditions for uplink resources; and

selecting at least one user experiencing the highest positive difference between its present channel condition and its average channel condition for uplink resources.

The method of claim 4 wherein the latter two selecting steps select their respective users from among a respective window of users.

The method of claim 5 further comprising the step of repeating the two latter selecting steps for each of subsequent respective windows.

7. The method of claim 4 wherein for the latter two selecting steps the selected at least one user is kept in the same relative location in the queue after completing an uplink transmission unless one selected user is at the front of the queue, whereupon that one selected user in moved to the back of the queue.

8. The method of claim 4 wherein for the first selecting step the user stays at the front of the queue until one complete higher level packet is delivered on the uplink resource.

9. A method for uplink resource assignment in a wireless communication network, the method comprising the steps of:

queuing (400) requests for uplink resources for a subframe;

selecting (402) the first user of the queue for uplink resources, wherein if at least one complete packet is finished being transmitted on the uplink resources the user is moved to the back of the queue;

selecting (404) at least one user experiencing the best channel conditions for uplink resources;

selecting (406) at least one user experiencing the highest positive difference between its present channel condition and its average channel condition for uplink resources; and

An evolved NodeB (100) for assigning uplink resources in a wireless communication network, the evolved NodeB (100) comprising:

a transceiver (108), the transceiver (108) operable to receive (104) uplink requests from user equipment (102) and to provide (106) uplink assignments to the user equipment (102) such that the user equipment (102) can use the assigned uplink resources for uplink communications with the evolved NodeB (100); and

a processor (110) coupled to the transceiver (108), the processor (110) operable to receive subframe uplink requests of the user equipment (102) from the transceiver (108), queue (400) the requests for uplink resources, select (402, 404, 406) users based in part on their channel conditions into a linear superposition of fractional groups having similar channel conditions, and assign (410) uplink resources for the subframe after the users have been selected for the available resources in the subframe.