CA2253123A1 - Dynamic resource allocation method and apparatus for broadband services in a wireless communications system - Google Patents

Abstract

A dynamic resource allocation method and apparatus for broadband services in a wireless communications system. The communications system can have a number of cells, each of which has multiple sectors. Each sector can contain a number of communications sites. Information is transmitted in time subframes scheduled to avoid interference between the sectors and cells, and different degrees of concurrent packet transmission can be scheduled for different classes of communications sites. The communications sites can be classified based on reception quality, such as by comparing their measured signal-to-interference ratio (SIR) with a SIR threshold.

Description

CA 022~3123 1998-11-0~

FOR
BROADBAND SERVICES IN A WIRELESS COMMUNICATIONS
SYSTEM

CROSS REFERENCE TO RELATED APPLICATIONS

The subject matter of the present application is related to the subject matter of U.S. patent application Serial No. 08/775,466 entitled "Method and 15 Apparatus for Providing High Speed Services Using a Wireless Communications System" to Thomas K. Fong, Paul Shala Henry, Kin K.
Leung, Xiaoxin Qiu, Nemmara K. Shanka~ alayanan and ac~ignecl to AT&T
Corp., filed on December 30, 1996 and U.S. patent application Serial No.
08/832,546 entitled "Method and Apparatus for Resource ~.signment in a 20 Wireless Communications System" to Xiaoxin Qiu and Kapil Chawla, filed April 3, 1997, the entire disclosures of which are hereby incorporated by reference.

FIELD OF THE INVENTION
The invention relates to wireless communications systems. More particularly, the invention relates to a method and appalalus for dynamic resource allocation for broadband services in a wireless communications system.

CA 022~3123 1998-11-0~

The need for high-speed broadband packet services will grow tremendously as tele-commuting and Internet access become increasingly popular. Customers will expect high quality, reliable access to high-speed 5 communications from homes and small bl-einessçe in order to, for example, access: (a) the World Wide Web for information and entertainment; (b) office equipment and data from home at rates comparable to Local Area Networks (LANs); and (c) multimedia services such as voice, image and video. Although varying with application, effective bro~db~n~l communication requires a 10 bandwidth sufficient to permit a data rate up to the range of several tens of Mega-bits per second (Mbps).
Traditional wireless communications systems have a problem delivering high-speed services because of the amount of bandwidth these services require.
Bandwidth is a key limiting factor in detçrmining the amount of information a 15 system can transmit to a user at any one time. In terms of wireless networks,bandwidth refers to the difference between the two limiting frequencies of a band expressed in Hertz (Hz).
The concept of bandwidth may be better understood using an analogy. If information carried by a network were water, and links between communication 20 sites were pipes, the amount of water (i.e., information) a network could transmit from one site to another site would be limited by the speed of the water and thediameter of the pipes carrying the water. The larger the diameter of the pipe, the more water (i.e., information) can be transmitted from one site to another in a given time interval. Likewise, the more bandwidth a communications system 25 has available to it, the more information it can carry.
Traditional wired communications systems using modems and a physical tr~n~mi~.sion medium, such as twisted pair copper wire, cannot currently achievethe data rates necessary to deliver high-speed service due to bandwidth CA 022~3123 1998-11-0~

limitations (i.e., small pipes). Promising wired-network technologies for broadband access, such as Asymmetrical Digital Subscriber Loop (ADSL) and Hybrid Fiber-Coax (HFC), can be expensive and time consuming to install.
The benefit of wireless systems for delivering high-speed services is that 5 they can be deployed rapidly without in~t~ tion of local wired distribution networks. However, traditional wireless systems such as narrowband cellular and Personal Communications Services (PCS) are bandwidth limited. As an alternative, wireless solutions such as Multichannel Multipoint Distribution Service (MMDS) and Local Multichannel Distribution Service (LMDS) have 10 become attractive, but these solutions presently offer limited uplink channel capacity and may not be capable of supporting a large number of users.
One solution for solving the bandwidth limitation problem for wireless systems is to m~ximi7~ the available bandwidth through frequency reuse.
Frequency reuse refers to reusing a common frequency band in different cells 15 within the system. Refer, for example, to FIG. 1 which shows a typical wireless communication system. A Base Station (BS) 20 communicates with several Terminal Stations (TS) 22. The BS 20 is usually connected to a fixed network 24, such as the Public Switched Telephone Network (PSTN) or the Internet.
The BS 20 could also be connected to other base stations, or a Mobile Telephone 20 Switching Office (MTSO) in the case of a mobile system. Each TS 22 can be either fixed or mobile.
The BS 20 communicates information to each TS 22 using radio signals transmitted over a range of carrier frequencies. Frequencies represent a finite natural resource, and are in extremely high tlem~n~l Moreover, frequencies are 25 heavily regulated by both Federal and State government.~ Consequently, each cellular system has access to a very limited number of frequencies.
Accordingly, wireless systems attempt to reuse frequencies in as many geographic areas as possible.

CA 022~3123 1998-11-0~

To accomplish this, a cellular system uses a frequency reuse pattern. A
major factor in designing a frequency reuse pattern is the attempt to maximize system capacity while m~int~ining an acceptable Signal-to-Interference Ratio (SIR). SIR refers to the ratio of the level of the received desired signal to the level of the received undesired signal. Co-channel interference is interference due to the common use of the same frequency band by two different cells.
To determine frequency reuse, a cellular system takes the total frequency spectrum allotted to the system and divides it into a set of smaller frequency bands. A cellular communications system has a number of communications 10 sites located throughout a geographic coverage area serviced by the system. The geographic area is org~ni7e~1 into cells and/or sectors, with each cell typically cont~ining a plurality of communications, sites such as a base station and termin~l stations. A cell can be any number of shapes, such as a hexagon.
Groups of cells can be formed, with each cell in the group using a different 15 frequency band. The groups are repeated to cover the entire service area. Thus, in e~sçn~e, the frequency reuse pattern represents the geographic distance between cells using the same frequency bands. The goal of a frequency reuse pattern is to keep co-channel interference below a given threshold and ensure successful signal reception.
The most aggressive frequency reuse pattern is where the same frequency band is used in every cell. One example of such a system is Code Division Multiple Access (CDMA) systems, which spread the transmitted signal across a wide frequency band using a code. The same code is used to recover the transmitted signal by the CDMA receiver. Although CDMA systems reuse 25 the same frequencies from cell to cell, they require a large amount of frequency spectrum. In fact, the amount of spectrum required by CDMA systems to offer high-speed bro~(lb~n~l services to a large number of users is commercially unrealistic.

CA 022~3123 1998-11-0~

Another example for aggressive frequency reuse Time Division Multiple Access (TDMA) systems, an example of which is discussed in United States Patent No. 5,355,367, which use the redlln~l~nt tr~n~mi~.eion of information packets to ensure an adequate SIR. The use of redlln~nt packet tr~n.~mi~sions, 5 however, merely trades one inefficiency for another. Although a frequency bandcan be reused from cell to cell, rednnd~nt packet tr~n~mi~.cion means that a smaller portion of that frequency band is now available for use by each cell in the system since multiple packets are required to ensure the successful reception of a single packet.
In addition to the frequency reuse problem, traditional cellular systems are not engineered to allow a communications site to use the entire bandwidth available to the system (or "total system bandwidth"). Rather, traditional cellular systems employ various techniques in both the frequency domain and time domain to m~ximi7P the number of users capable of being serviced by the 15 system. These techniques are predicated on allocating smaller portions of the total system bandwidth to service individual communication sites. These smaller portions are incapable of providing sufficient bandwidth to offer high speed services.
An example of a technique employed in the frequency domain is 20 Frequency Division Multiple Access (FDMA). FDMA splits the available bandwidth into smaller sections of bandwidth under the concept of providing less bandwidth for a greater number of users. Using the water/pipe analogy, a single large pipe is separated into a number of smaller pipes, each of which is ~.cign~cl to a sector or cell. Unfortunately, the smaller frequency bands are too 25 small to support high-speed broadband packet services. Moreover, by definition, a communication site is not capable of using the total system bandwidth, but rather is limited to a discrete portion of the total system bandwidth.

An example of a technique employed in the time domain is TDMA, described above. Using the water/pipe analogy, each cell or sector has access tothe entire pipe for a fixed amount of time. These systems allocate a specific time slot of a fixed duration for a specific communication site. As a result, a 5 communication site cannot transmit more information than can be accommodated by its assigned time slot. Traditional TDMA systems are designed to handle circuit switching and, therefore, are static in nature. Thus,traditional TDMA systems are not designed to take advantage of new switching technology, such as packet switching.
Some systems employ a combination of FDMA and TDMA to improve the call capacity of the system. These FDMA/TDMA systems, however, merely combine the disadvantages of both and do not permit a user access to the total - system bandwidth on a dynamic basis. To solve this problem, some systems employ a concept called "dynamic resource allocation" to share the radio 15 resource among communications sites efficiently. Dynamic resource allocation methods, however, require a central controller or complicated algorithms to dynamically clçtçrmin~ available time slots and coordinate their use by the communication sites.
In order to increase spectrum efficiency, other cellular systems have 20 employed multiple frequency reuse patterns within the same system. For example, United States Patent No. 4,144,41 1 issued to Frenkiel on March 13, 1979, teaches static reuse of frequencies in a system employing a mini~hlre-sized overlay in each cell, with the mini~hlre-sized overlay using thesame type of reuse pattern as the large cell reuse pattern. This is achieved 25 through yet lower transmit powers and m~int~ining the same site spacing to cell radius as the large-cell. This concept is typically referred to as cell splitting.
An enhancement to Frenkiel is discussed in an article authored by Samuel W. Halpern entitled Reuse Partitioning in Cellular Systems, presented at CA 022~3123 1998-11-0~

the 33rd IEEE Vehicular Technology Conference on May 25-27, 1983 in Toronto, Ontario, Canada. The Halpern article sets forth a cellular system having multiple frequency reuse levels (or patterns) within a given geographicalarea. For example, a cluster of cells normally employing a seven-cell reuse S pattern may simultaneously operate on a three-cell reuse pattern and a nine-cell reuse pattern. One set of frequencies is dedicated to the three-cell reuse pattern while another set of frequencies is dedicated to the nine-cell reuse pattern.
Generally, the principle behind the Halpern system is to allow a degradation of Carrier-to-Interference (C/I) performance for those subscriber units that already 10 have more than adequate C/I protection while providing greater C/I protection to those subscribers that require it. Therefore, a subscriber with the best received signal quality will be assigned to the set of channels for the three-cell reuse pattern since they are able to tolerate more co-channel interference than a subscriber whose signal quality is poorer. The subscriber having the poorer 15 received signal quality is therefore assigned to a channel collespolldent to the nine-cell reuse pattern.
The Halpern system, like previous multiple frequency reuse partitioning systems, is llne~ti~f~ctory for a number of reasons. For example, in practice the Halpern system permits only a small fraction of the total traffic to use the closer 20 reuse pattern for the mini~hlre-sized overlay, leading to little or no gain in system capacity. Further, the Halpern system is designed for circuit switched systems, and not for the modern packet switched systems. More specifically, circuit switched systems can tolerate a lot of measurement overhead and delay when connecting to the user. If the same techniques were applied to a packet 25 switched system, however, several measurements would be required before transmitting each packet. The overhead and delay introduced would be excessive, and therefore the method described in the Halpern reference would not be feasible. In fact, the Halpern method is designed for the conventional CA 022~3123 1998-11-0~

telephony system, and not packet switched systems in general.
Moreover, previous systems were designed to do the reuse partitioning in the frequency domain, that is they were focused on dividing the total frequency bandwidth available to the system and allocating one portion of this total 5 frequency bandwidth to one reuse pattern, and another portion to another reusepattern. Dividing the available frequency, however, limits the maximum data rate that can be provided to any single user or application by the system.
Therefore, frequency reuse partitioning schemes are not suitable for supporting high data rate applications such as those envisioned for wireless broadband 1 0 systems.
A specific implementation of frequency reuse partitioning is disclosed in United States Patent No. 5,038,399 (the "Bruckert patent"). The Bruckert system is directed towards a mech~ni~m for measuring various signal strengths from base stations and subscriber stations throughout the system, constructing a15 reuse level gradient, and using this gradient as a basis for switching between multiple frequency reuse patterns.
As with the Halpern system, the Bruckert system is unsatisfactory for a number of reasons. For example, the Bruckert patent is also targeted towards a circuit switched system and is not ~lçci~n~l towards modern packet switched 20 systems. As a result, the bandwidth available to a user is fixed during the call duration, thus becoming inflexible for h~ntllin~ data bursts as anticipated in bro~(lb~n-l services. Furthermore, the Bruckert patent describes a method for acsigning different users to different reuse levels according to the "reuse level gradient," which is another way of stating the assignment is based upon different 25 interference levels. In many instances, however, an integrated system providing different services to the same user may require different reuse levels due to their differing service requirements, even though they experience the same interference. The Bruckert patent fails to disclose how the quality of service CA 022~3123 1998-11-0~

(QoS) is m~int~ined for each application using this method. In addition, the Bruckert patent fails to disclose any techniques for ensuring f~irness among communication sites in terms of each site gaining access to the communication resource in a uniform manner. Finally, the Bruckert patent fails to disclose the5 use of multiple reuse patterns in the time domain, as with the previously discussed systems.
In view of the foregoing, it can be appreciated that a substantial need exists for a dynamic resource allocation method and appaldllls for broadband services in a wireless communications system that efficiently provides high 10 quality bro~clb~ntl packet services to a large number of users, and solving the other problems discussed above.

SUMMARY OF THE INVENTION

The disadvantages of the art are alleviated to a great extent by a dynamic resource allocation method and apparatus for broadband services in a wireless communications system. The communications system can have a number of cells, each of which has multiple sectors. Each sector can contain a number of communications sites. Information is transmitted in time subframes scheduled 20 to avoid interference between the sectors and cells, and different degrees of concurrent packet tr~n~mi~ion can be scheduled for different classes of communications sites. The communications sites can be classified based on reception quality, such as by comparing their measured signal-to-interference ratio (SIR) with a SIR threshold.
With these and other advantages and features of the invention that will become hereinafter ap~ cllt~ the nature of the invention may be more clearly understood by reference to the following detailed description of the invention, the appended claims and to the several drawings attached herein.

CA 022~3123 1998-11-0~

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram of a typical wireless communication system suitable for an embodiment of the present invention.
FIG. 2 shows a cell layout and frame structure according to an embodiment of the present invention.
FIG. 3 shows the order of slot ~.ei~nment.c for the staggered resource allocation method according to an embodiment of the present invention.
FIG. 4 is a block diagram of a terminal classification process according to an embodiment of the present invention.
FIG. 5 shows the use order of subframes and mini-frames in the enhanced staggered resource allocation method according to an embodiment of the present invention.
FIG. 6 is a block diagram of a time slot scheduling process according to an embodiment of the present invention.
FIG. 7 shows a typical ~ntçnn~ pattern, with a front-to-back ratio of 25 dB and a beamwidth of 60 degrees, suitable for use with an embodiment of the present invention.
FIG. 8 shows the impact of the base station ~ntçnn~ beamwidth on throughput and coverage according to an embodiment of the present invention.
FIG. 9 shows the impact of the base station ~nt~nn~ front-to-back ratio on throughput and coverage according to an embodiment of the present invention.
FIG. 10 shows the impact of the terminal ~ntPnn~ front-to-back ratio on throughput and coverage according to an embodiment of the present invention.

DETAILED DESCRIPTION

CA 022~3123 1998-11-0~

The present invention is directed to a dynamic resource allocation method and appaldl~ls for broadband services in a wireless communications system. In particular, a resource allocation algorithm, referred to herein as the 5 Enhanced Staggered Resource Allocation (ESRA) method, is used for bro~clb~n~l wireless networks with directional antennas at base stations and termin~ The ESRA method uses Staggered Resource Allocation (SRA) but also considers reception quality at terminal locations. This is done by categorizing tçrmin~l~ into multiple classes depending on the ability to tolerate 10 concurrent packet tr~n.cmi~sions. The bandwidth is divided into multiple timesubframes, each of which has a number of mini-frames which allow different degrees of concurrent tr~n.cmi~ions. The packets of different classes are sent in the corresponding mini-frames. In accordance with the ESRA method, concurrent tr~n.cmi~.sions are maximized up to an extent tolerable by the 15 receiving terminals for throughput improvement, while avoiding major co-channel interference in the networks where the same frequency band can be re-used in every sector of every cell. In a reasonable radio environment, with practical ~ntçnn~ patterns and choices of system parameters, the ESRA method could provide 98.69% coverage, and yield a maximum throughput of 36.10%
20 per sector, with a packet tr~n.cmi~.~ion success probability of one given a specific SIR threshold. This translates into a very large network capacity and the high quality of service reveals the applicability of the ESRA method to support real-time traffic, such as voice and video, in addition to data.
Although the system described in detail herein is a fixed broadband 25 packet-switched network using TDMA techniques with user data rates of 10 Mb/s, link lengths typically less than 10 kilometers and an operating frequency in the range of 1 to 5 Ghz, the ESRA method can of course be used in other wireless communications systems.

CA 022~3123 1998-11-0~

Referring now in detail to the drawings wherein like parts are designated by like reference numerals throughout, FIG. 2 shows a service area in a wirelessnetwork divided into hexagon shaped cells. Each cell is further divided into multiple sectors numbered 1 to 6, and each sector is covered by a sector antenna5 co-located with a Base Station (BS), not shown in FIG. 2, at the center of thecell. Because of the co-location, sector ~ntçnn~ are also called BS antennas.
Terminals (users) can use directional antennas mounted on roof tops and pointed to respective BS ~ntPnn~c The beamwidth of each BS ~ntenn~ can be just wide enough to serve the whole sector, while a te min~l ~ntenn~ can have a smaller 10 beamwidth to suppress intc~lr~lence. The Front-to-Back (FTB) ratio for BS and .
terminal ~ntenn~c are assumed to be finite. Time is slotted such that a packet can be transmitted in each slot, and the downlink and uplink between tennin~l~
and BS are provided by Time-Division Duplex (TDD) using the same radio spectrum.
In the context of packet-switched networks, time slots naturally become the bandwidth resources. Time slots need to be dynamically allocated to various transmitters to send data packets such that a given SIR can be achieved at the intPn-led receiver for successful reception. This results in the concept of Dynamic Resource Allocation (DRA). The problem of time-slot ~c~ignment to 20 achieve certain optimal performance while meeting a SIR requirement can be m~thPnn~tically considered Non-Polynomial (NP) complete or hard, which implies a very high degree of computational complexity in deriving the optimal ~csignm.?nt~
In the fixed wireless network of the present invention, cell sectorization 25 and directional ~ntPnn~ at fixed terminal locations are used to reduce interference from neighboring sectors and cells through the Staggered Resource Allocation (SRA) method. This results in a distributed DRA algorithm where the same radio spectrum is used (shared) by each sector in every cell on a CA 022~3123 1998-11-0~

dynamic time basis. With the use of directional antennas to suppress interference, the SRA method is particularly effective in avoiding both inter-cell and intra-cell interference.
However, depending on terrain and fading, a certain terminal (e.g., 5 house) may be constantly unable to receive a signal with a satisfactory SIR due to its fixed location. The tr~n~mi.csion for other tçrmin~lc may always be successful. Thus, terminals at "good" and "poor" locations should be served according to different time-slot reuse patterns, which is called Time Slot ReusePartitioning (TSRP), allowing for many BS's to transmit simultaneously if the 10 intçn-led receiving tçrmin~l~ are located at good positions. When the receiving .
locations are poor, few BS's should be scheduled to transmit at the same time sothat a target SIR threshold can be met for successful reception at the receivingends.
TSRP divides the time frame (i.e., bandwidth) into a dedicated portion 15 and a shared portion. At most one packet is transmitted among four neighboring cells during each time slot in the dedicated portion and up to three packets canbe transmitted simultaneously in every cell in the shared portion. The purpose is to allow t~rmin~l~ at "good" and "poor" locations to use time slots in the dedicated and shared portion, respectively. Because of the bandwidth 20 partitioning into the dedicated and shared portion, many tçrmin~l locations with moderate reception quality either may be over-protected when transmitting in the dedicated portion or may not receive the packets succes~fully when sent during the shared portion. This results in a potential waste of bandwidth.
The present invention enhances the SRA method by considering the 25 reception quality of tçrmin~l~. This Enhanced Staggered Resource Allocation (ESRA) method has the capability of avoiding major interference as the SRA
method does, and makes use of the knowledge of the reception quality at terminals to improve throughput and m~int~in the success probability of one for CA 022~3123 1998-11-0~

packet tr~n.cmi~.sion. As discussed in detail below, ESRA also yields close to full coverage for a set of typical radio and system parameters.
Refer again to FIG. 2 with the regular, hexagonal cell layout. Each cell is divided into six sectors, each of which is served by a BS ~ntenn~ with 60 5 degree beamwidth, and terminal ~ntçnn~c can have a beamwidth smaller than 60 degrees. In the SRA method, time slots are grouped into 6 subframes and sectors are labeled by 1 to 6 anti-clock-wise, as shown in FIG. 2. The sector labeling patterns for adjacent cells are rotated 120 degrees, creating a cluster of 3 cells whose p~tt~rn~ can be repeated across the entire system. Note that the time 10 frame shown in FIG. 2 is applicable to both the downlink and uplink, which are .
provided by TDD using the same spectrum.
Each sector assigns time slots for transmitting packets to or from its tçrrnin~l~ according to a special order shown in FIG. 3. It is assumed that the BS is informed when a tennin~l needs to send packets, perhaps via a separate 15 multi-access channel or piggyback requests. For example, Sector 1 first schedules packets for tr~nemi.c~ion in time slots of subframe 1 (denoted by a). If it has more traffic to send, it then uses subframe 4 (b), subframe 5 (c), etc. until subframe 6 (f). The reasoning behind the particular order is as follows. If interference due to concurrent packet tr~n~mi~.cion in the same cell can be 20 tolerated, then after using all slots in the first subframe a, Sector 1 should use the first subframe of the opposite sector (Sector 4) in the same cell to make the best use of the BS directional ~nt.onn~. Following that, time slots in the first subframes for the sectors next to the opposite sector are used. To avoid interference due to overlapping ~ntenn~ patterns of neighboring sectors, their 25 first subframes are used as the last resort. For simplicity (while causing very minor throughput degradation), FIG. 3 does not show the ~csi~nment from the left and right-hand side of the subframes. The acsignment order for the next sector is "staggered" by a right rotation by one subframe based on the order for CA 022~3123 1998-11-0~

the previous sector. The ~.cignment order, regardless of the associated sector, is generally referred to as the "staggered" order.
It is easy to see from FIG. 3 that if all sectors have traffic load of less than one-sixth of total channel capacity, all packets are transmitted in different 5 time subframes (labeled "a" in each sector), thus causing no interference within the same cell. Of course, as the traffic load increases packets will be transmitted simultaneously, and this increases the level of interference. Nevertheless, the staggered order exploits the characteristics of directional Ant.?nn~c to allow multiple concurrent packet tr~n~mi~sions while reducing the intra-cell 1 0 interference.
Besides m~n~eing intra-cell interference, the SRA method helps avoid interference from major sources in the neighboring cells. This is particularly so when the traffic load is low to moderate. Consider the downlink for Sector I in the middle cell of FIG. 2. Sector 2 in the bottom cell and Sector 3 in the upper15 cell are the major sources of interference. By e~minin~ the staggered order for Sector 1, 2 and 3, note that they will not transmit simultaneously, and thus will not interfere with each other, provided each has a traffic load of less than one-third of total channel capacity (i.e., using only subframes a and b for tr~n.cmi~ion). The same comment also applies to the uplink, where Sector 2 and 20 5 of the bottom cell in FIG. 2 now become the major sources of interference.
Due to the symmetry of the staggered order and cell layout, the comment applies to each sector in every cell.
For a given radio environment and ~nt~nn~ characteristics, the SRA
method can be used in conjunction with a control mechanism to improve the SIR
25 at the receiving ends. Specifically, the control limits packet tr~n.cmi.~.sions only in the first few subframes in the staggered order for each sector. For example, if at most three packets can be sent simultaneously by various BS or terminal antçnn~ in the same cell to ensure the required reception quality in the given CA 022~3123 1998-11-0~

environment, only time slots in subframes a, b and c as indicated in FIG. 3 would be used for tr~n~mi.~cion in each sector. The control limits the degree ofconcurrent tr~n~mi.~sions, and thus the amount of interference, to achieve a target SIR for the desirable quality of service.
The ESRA method can include the following components, discussed in detail below: terminal classification; selection of cell and sector; mini-frame structure and schedllling mech~ni~m; and the selection of mini-frame sizes.
Terminal Classif cation The ESRA method uses the same sector labeling as the SRA method shown in FIGS. 2 and 3. The basic idea of termin~l classification in the ESRA
method is to categorize terminals ba~sed on their ability to tolerate various - degrees of concurrent packet tr~ncmi~ions according to the staggered order.
The tolerance depends on the reception quality of the termin~l locations, which 15 in turn depends on the distance between the BS's and t~rmin~l.c, tr~n~mic.sion power, antenna characteristics, terrain and fading. For the layout in FIG. 2 with six sectors per cell, there are six levels of concurrent tr~n~mi~sion.
Correspondingly, terminals are categorized into six classes, indexed by 1 to 6.
As depicted in FIG. 3, each time frame has 6 subframes, indexed by 1 to 20 6. Let J' be the index of the mth subframe for use by sector i in the staggered order. For example, J/ = 3, J2 = 6, J3 = 1, J3 = 5, J3 = 4 and J3 = 2 for sector 3 as it first uses slots in subframe 3, 6, 1, and so on, as shown in FIG. 2.
Further, for c = 1, 2, ..., 6, let ICG) 5 { 1, 2, ..., 6} denote the set of sectors allowed to transmit in subframe j when each sector can use only the first c 25 subframes in the staggered order for tr~n~mi~ion (which results into c concurrent packet tr~n~mi~sions in each cell). For instance, I2(1) = {1,4}, I3(1) =
{1,4,3},I4(3)= {3,6,5,1} andI5(3)= {3,6,5,1,2}.

CA 022~3123 1998-11-0~

Assume that the system can activate one or a set of BS ~ntenn~c to send a special signal, such as via a pilot tone, for measurement purposes. FIG. 4 illustrates a classification procedure for a terminal located in sector i according to an embodiment of the invention. After beginning at step 400, the process setsc = 6 and k = 1 at step 410. The process then setsj = Jk at step 720. The system instructs the BS antenna in the sector i, where the terminal belongs, to transmit a special signal. At step 422, the received signal strength is measured at the t~nnin~l location. Next, BS ~ntçnn~c for all sectors in ICa) are arranged totransmit simultaneously and the received power at the terminal is measured at 10 step 425. The SIR at the tç min~l when all sectors in ICO transmit can be obtained from these two measurements at step 427. At step 430, if the SIR is less than the threshold required for s~ti~f~ctory signal detection, the process continues with step 470. Otherwise, the process continues with step 440.
At steps 440 and 480, the tçrmin~l is categorized as a class-c tçrmin~l 15 and the procedure is completed if k = c. In other words, the system can sustain the interference with c concurrent packet tr~n.cmi.csions according to the staggered order. Otherwise, at step 450 k is incremented by 1 and the process proceeds with step 420 to check the SIR when transmitting in the next subframe.
If at step 470 c > 1, c is decreased by 1 and k is set to 1 at step 460 before 20 step 420 is repeated. Otherwise, at step 485 the terminal cannot be served by the ESRA method because the terminal is unable to meet the SIR threshold - even when one packet is transmitted in each cell at a time. Therefore, the procedure stops at step 490.
For a typical radio environment, less than 1.5% of uniformly located 25 terminals cannot be served by the ESRA method. In those cases, termin~l antennas with improved FTB ratio, or sophisticated digital signal processing techniques to lower the SIR requirement, may be used to ensure s~ti~factory reception.

CA 022~3123 1998-11-0 In practice, the tçrmin~l classification can be done when inct~lling the service at a terminal location. In addition, the classification of each terminalcould be updated periodically by monitoring the reception quality through measurements and statistics collection. Periodic monitoring would be helpful 5 because the radio environment tends to change over time due to, for example, seasonal fluctuations and the addition of man-made objects to the radio path.

.~election of CeU and Sector It is well known that cell selection can improve the quality of signal reception. To take advantage of the macro-diversity in the ESRA method, each terminal selects its cell and sector, which may not nlocess~rily be the closest one in distance, according to the fading and the scheduling algorithm in use.
Specifically, for each terminal, the ESRA method applies the terminal 15 classification procedure presented above to determine a terminal class for several combinations of sectors and cells in the vicinity of the terminal. The t~ rnin~l then chooses a home sector and cell that gives the tçrmin~l class withthe largest index (i.e., that can tolerate the highest degree of concurrent tr~n.~mi~sion). If multiple combinations of cells and sectors yield the same 20 terminal class, the one with the highest SIR can be chosen.
With this terminal classification and selection of cell and sector, packets for a class-c terminAI can be received s~lGcescfully, as far as meeting the required SIR is concerned, if each sector uses the first c subframes in the staggered order (which yields c concurrent packet tr~n~mi~ions in each cell). For this reason, 25 the frame structure in FIG. 2 can be modified so that packets for each terminal class can now be transmitted simultaneously up to the maximum tolerable degree of concurrent tr~n~mi.~ions in order to improve the throughput without degrading the success probability of packet reception.

CA 022~3123 1998-ll-0 Frame Structure and .~hed~lling Mech~7ni~m Each time frame in the ESRA method consists of six subframes, indexed 5 by 1 to 6 in FIG. 5. Each subframe is further divided into six "mini-frames,"
which are also labeled from 1 to 6. Each mini-frame with the same label consists of multiple but fixed number of time slots in each subframe. The sizes of mini-frames are chosen to match the expected traffic dçnn~n-l of the terminalclasses and each sector uses the subframes according to the staggered order, 10 given by "a" to "f" in FIG. 5. It is hlll~o~ lt to note that time slots of only those.
mini-frames marked with a solid line are available to the corresponding sector indicated on the left-most side of the figure. Clearly, varying from subframe tosubframe, each sector is allowed to schedule packet tr~n.cmi~sion in one or moremini-frames in some subframes, but not in others. For instance, Sector 2 can use15 all mini-frames in Subframe 2, but it can schedule tr~n.~mi~eion only in Mini-frame 5 and 6 in Subframe 3. The other mini-frames in Subframe 3 are unavailable to Sector 2.
There are different degrees of concurrent packet tr~n.cmi~sion in various mini-frames. For c = 1, 2, ..., 6, as many as c packets are transmitted 20 simultaneously during mini-frarne c in each subframe. On one extreme, only one packet is transmitted in each cell during Mini-frame 1, while on the other extreme, up to six packets are sent during Mini-frame 6. The various mini-frames allow different degrees of concurrent packet tr~n~mi~.cions. Thus, the mini-frame structure is compatible to the terminal classification because 25 packets for class-c tçrrnin~l~ transmitted in mini-frame c will be successfully received as verified in the classification procedure. In fact, as discussed in detail below, packet tr~n~mi~sion for a class-k terminal in mini-frame c with c < k, referred to as "upgraded sharing," will also be succes~fully received.

CA 022~3123 1998-11-0~

In the ESRA method, the procedure shown in FIG. 6 is invoked for each time frame by each sector in every cell to assign available time slots in the frame to pending packets for tr~n~mi~sion. Once a packet is scheduled for tr~n~mi~sion in a time slot, the slot becomes unavailable to other packets.
After beginning at step 600, the procedure sets c = 1 and i = c at step 610. At step 620, the sector schedules pending packets of terminal class i for tr~n.cmi~sion in the available time slots of mini-frame c, starting from the first subframe in the staggered order (denoted by a to f) and according to the availability of mini-frames in the subframes to the sector as shown in FIG. 5.
10 The sche~uling continues until either (i) all available time slots in mini-frame c have been assigned at step 630; or (ii) all pending packets for the terminal class have been scheduled for tr~n~mi~sion at step 640. If condition (i) occurs, the process proceeds with step 660, otherwise the process proceeds with step 650.
At step 660, if c < 6, then c is increased by 1 and i is set to c at step 670 15 before procee~ling with step 620. Otherwise, the procedure stops at 690 as all available time slots in the time frame have been assigned.
At step 650, if i < 6, then i is increased by 1 and the process proceeds with step 620 to schedule tr~n~mi.c~ion of packets for the next termin~l class in mini-frame c. Otherwise, the procedure stops at step 690 as all pending packets have been scheduled for tr~n.cmi~sion. It is worth noting that as long as time slots are available, packets are tr~n~mittç~ by the upgraded sharing to further enhance the SIR at the receiving ends.

.~election of Mini-Frame Sizes The mini-frame structure can be viewed as a division of bandwidth into multiple "channels" allowing different degrees of concurrent packet tr~n~mi~sions tolerable in terms of SIR by various tç min~l~. To m~ximi7P the CA 022~3123 1998-11-0~

system throughput the sizes of mini-frames should be chosen to match the trafficload from the respective terminal classes. Without loss of generality, consider that terminals of all classes have an identical traffic load. Let aj be the fraction of class-i terminals (relative to the total number of terminals served by the 5 ESRA method) in the whole network for i = 1 to 6. Further, let Nt be the "target" number of time slots in each subframe frame, which is determined by considering packet delay requirements, schedllling overhead and so on. In addition, let mini-frame i in each subframe have nj time slots. As explained in detail with respect to FIG. 5, each sector can use mini-frame i in i different n 10 subframes. Therefore, to handle the uniform traffic load among te.min~
where ~ is a proportionality constant and the rounding of integer is ignored at ~ ni = N, the present moment. Since:

N, ~, aj / j the following relationship holds:
Substituting this into the equation above for ni yields:

CA 022~3123 1998-11-0~

N, ~j/i jPHANTOM ~ / j where [x] denotes the integer closest to x. With these mini-frame sizes, each subframe has:
N= ~n i time slots. The frame size is K~ where K is the number of sectors in each cell, which equals 6 for the setting under consideration.

Performance Analysis of the ESM Method Using the terminal classification method described herein, packet tr~n~mi~sions for each t~.?rmin~l class in its respective mini-frame will be successful. That is, the success probability of packet tr~n~mi~sion is one as far 10 as meeting a specific SIR threshold is concerned.
To analyze the packet throughput for the ESRA method, assume that terminals of all classes have identical traffic load and that there are always packets pending for tr~ncmi~cion. Based on the size of each mini-frame i, the maximum throughput for class-i termin~lc is in~/KNpackets per time slot in each 15 sector. This is because: (1) each sector can transmit during mini-frame i in i different subframes of each frame, and (2) each packet tr~n~mic~ion for class-i termin~l~ in mini-frame i will be suGces~ful by the definition of termin~l classification. Thus, the maximum throughput in each sector for all terminal classes IS:

CA 022~3123 1998-11-0~

rm = ~ ~n; = ~ i N, o~i/i KN j KN ~ ~ZjPHANTOM-h Ignoring the rounding for integer, and applying the facts that ~ cci = 1 and N, K ~, c~j / j j ~N, the maximum throughput per sector is obtained:
Since the success probability of packet tr~n~mi.c~ion for the ESRA
method is one, its throughput is only limited by the availability of pending 5 packets associated with each terminal class for a given mini-frame structure. As a desirable consequence, once the maximum throughput is reached for sufficient traffic, further increase of traffic load will not cause any throughput degradation.
Although it may appear that the bandwidth partition into mini-frames in the ESRA method could lead to loss of "trunking efficiency," note that packets 10 for class-k termin~l.e can be transmitted s~lcces.cfully during any mini-frame c available to the associated sector with c < k. When such a packet is sent in mini-frame c < k, the SIR can actually be improved at the receiving ends. Such sharing of mini-frames is referred to as "upgraded sharing."
To show the SIR improvement, let q> be the SIR threshold for correct 15 signal detection at a receiver. Further, use Pj to denote the received signal or interference strength from the BS ~ntenn~ of sectorj. Without loss of generality, consider a terminal in a particular sector i. As before, use Jm to denote the index of the mth subframe in the staggered order for use by the sector i in the time-slot assignment. According to the termin~l classification, the terminal is CA 022~3123 1998-11-0~

~ Ps S ~ I k (iJ
categorized to be class k if k is the largest integer such that:
for all m = 1, 2, . .., k and j = Jm . If a packet for the class-k terminal in sector i s ~ 1/ ~') .
is transmitted in mini-frame 1 < k, then the SIR at the receiver is given by wherej' = Jm for some m ~ { 1,2,...,1} because the sector can use mini-frame I in 5 any one of the first I subframes in the staggered order. Since I < k, which indicates different degrees of concurrent packet tr~n~mi~sion, 1/0 c Ik(j) for any subframe j. This is certainly true for all j = Jm with m = 1, 2, ..., 1. C;~ombining this and the fact that Ps 2 O~ the denominator in the equation for <~ must be less than or equal to that in the equation associated with (p, thus (p 2 ~. In other 10 words, the upgraded sharing, or transmitting class-k packet in any mini-frame c with c < k, can meet the SIR detection threshold. Since it is possible that Ps = ~, or tr~l.cmi.csion is at a lower power level, when a sector s does not have sufficient traffic to send, the upgraded sharing actually improves the SIR at the receivers.
In contrast, a similar analysis reveals that transmitting class-k packets during available time slots in mini-frame c > k, to be called "downgraded sharing," cannot guarantee the satisfactory SIR. That is, the downgraded sharingdoes not yield succe~sful packet tr~n~mi.c~ion with probability one. For this reason, the scheduling algorithm described with respect to FIG. 6 does not CA 022~3123 1998-11-0~

include such sharing. Nevertheless, the downgraded sharing can still be applied to packets without tight delay requirements. This is particularly so if the BS can schedule class-k packets for tr~n.cmic.cion in time slots of mini-frame c > k when the traffic load is sufficiently low that the degree of concurrent packet 5 tr~ncmicsions in those slots can be kept to be no more than k. It is important to note that packet tr~ncmicsions by the downgraded sharing have no impacts on the original tr~ncmiccions, or those by the upgraded sharing, because the downgraded sharing does not increase the degree of concurrent tr~ncmicsions.
Thus the interference level remains unchanged. In the worst case, if packets are10 not received sllcces.cfully the first time by the downgraded sharing, they can be .
re-transmitted in their corresponding or upgraded mini-frames.

N~ , .cal Performance Results for the ESRA Method Typical radio and ~ntçnn~ parameters and a simulation model created with OPNET, a computer-aided engineering tool for communication networks and analysis developed by MIL 3, Inc. of Washington, DC, were used to obtain the fraction of termin~l.c of different classes. Based on the fractions and the assumption of uniform traffic among tennin~l classes, the results were then 20 applied to compute the maximum packet throughput for the ESRA method.
A two-tier, hexagonal-cell layout with a total of 19 cells was used. That is, an outer tier of 12 cells is added to the configuration shown in FIG. 2. Each cell was divided into 6 sectors, each of which is served by a BS ~nt~nn~
co-located at the center of the cell. Unless specified otherwise, the beamwidth 25 (where the signal strength drops by 3 dB) of each BS and tçrminAI ~ntenn~ is 50 degrees and 30 degrees, respectively, while each terminal ~ntenn~ points directly to its BS antenna. Practical antenna patterns, such as the one shown in FIG. 7, were used. Although the back/side lobe is not shown, signal arriving at the CA 022~3123 1998-11-0~

back/side lobe is ~ttçnl~t~d according to the FTB ratio. Due to the overlapping of ~ntenn~ patterns, it is likely that certain terminals, especially those located at the sector boundary, will receive a significant amount of interference from the neighboring sectors.
Each radio path between a transmitter and a receiver was characterized by a path-loss model with an exponent of 4 and lognormal shadow fading. For the downlink, since there is only one radio path between all BS antennas in the same cell (which are co-located) and any terminal in the cell, the intendçd signal and interference should experience the same lognormal fading and path loss.
10 However, the fading from BS ~ntçnn~e at other cells were assurned to be different and independent. Unless stated otherwise, the typical FTB ratios for BS and termin~ tt?l~ e (denoted by B and T) are 25 and 15 dB, respectively.
The standard deviation for the shadow fading is 8 dB. Furthermore, with standard modulation and equalization schemes, such as Quadrature Phase-Shift 15 Keying (QPSK) and Decision Fee~b~çl~ Equalization (DFE), the SIR threshold for satisfactory detection could be from 10 to 15 dB, so a SIR threshold of 15 dB
was selected. For each packet tr~n~mi eeion, if the SIR at the intended receiver -exceeded the threshold the packet was considered to be ~ucces.efully received.
Only the statistics in the middle cell are collected and the following results were obtained for about 1,000 tçrmin~le uniformly placed throughout sector 1 of the center cell.
Following the classification method of FIG. 4, the following Table l presents the fraction of termin~le in various classes for the SIR threshold of 15 dB.

Table 1. Fraction of Terminals in Various Classes Terminal Class Without BS Selection With BS Selection CA 022~3123 1998-11-0~

Table 1. Fraction of Terminals in Various Classes Terminal Class Without BS Selection With BS Selection 0.1146 0.0813 2 0.5250 0.6375 3 0.0333 0.0479 4 0.0115 0.0208 0.0219 0.0292 6 0.1531 0.1698 Coverage 0.8594 0.9864 Throughput 0.3402 0.3610 Results with or without BS and sector selection, referred to as "BS selection" in short, are included. The sum of the fractions for all classes gives the total 5 fraction of terminals that can be served by the ESRA method, or "coverage."
Since the ESRA method can elimin~te intra-cell i~llelr~rel1ce entirely by allowing only one packet tr~n.cmi~ion in each cell at a time, the coverage is detçrminecl mainly by inter-cell interference and fading. On the other hand, themaximum throughput strongly depends on the fractions of terminals in various 10 classes. That is, a higher degree of tolerable concurrent tr~n~mi~sions results in higher throughput.
Without BS selection, the coverage is 85.94%, the rem~ining 14.06% of terminals cannot be served even when there is only one packet tr~n~micsion in each cell at a time. In contrast, the coverage increases to 98.64% with BS

CA 022~3123 1998-11-0~

selection due to macro-diversity, and such coverage is adequate in practice.
It is also worth noting that Table 1 reveals that a majority of the terminals are in class 2 and a smaller fraction are in classes 3 to 5. This is because the staggered order is particularly good at avoiding both intra- and 5 inter-cell interference when each sector transmits in the first two subframes in the order. For higher degrees of concurrent tr~n~mi~.cions, however, the amount of intra-cell interference increases due to the overlapping ~ntçnn~ patterns between adjacent sectors in the same cell. The 16.98% class-6 terminals are likely located near the BS with favorable fading and are not affected by adjacent 1 0 sectors.
FIGS. 8 to 10 show how the antenna beamwidth and FTB ratios can impact the performance of the ESRA method. FIG. 8 shows the effect BS
~ntenn~ beamwidth can have on the maximum throughput and coverage. The coverage is insensitive to the beamwidth, and the maximum throughput can be 15 improved if the beamwidth is reduced from 60 degrees to a smaller value. Thisis because a narrower beamwidth reduces interference from neighboring cells and sectors. A beamwidth of about 50 degrees for BS ~ntenn~ may be appro~l;ate to serve a 60 degree sector.
The beamwidth of the terminal ~ntenn~ was also varied from 10 to 40 20 degrees, while keeping the BS ~ntçnn~ beamwidth at 50 degrees and other system parameters unchanged. ESRA p~lrollllance is also insensitive to the range of tçnnin~l ~nterln~ beamwidth because as long as the beamwidth is less than 60 degrees each terminal ~ntenn~ faces to the front lobe of the BS ~ntçnn~
of one sector in the first-tier neighboring cells, which contributes most of the25 inter-cell interference for the terminal.
As shown in FIG. 9, which illustrates the performance impacts due to the FTB ratio of BS antenna, the coverage is relatively insensitive to the FTB ratioof BS antenna. On the other hand, as the FTB ratio increases, interference CA 022~3123 1998-11-0~

decreases, allowing a high degree of concurrent tr~n~mi~.cions and improving themaximum packet throughput. However, when the FTB ratio reaches 25 dB, further increases of the ratio yield only marginal throughput improvement because other parameters, such as the terminal antenna FTB ratio and SIR
S threshold, become domin~ting factors in deterrnining the throughput.
In contrast, FIG. 10 shows that both coverage and throughput strongly depend on the FTB ratio for terminal ~nt~nn~e Generally, when the ratio is high, inter-cell interference can be suppressed sufficiently so that almost all terminals meet the SIR threshold and terminals can tolerate a high degree of 10 concurrent tr~n~mi~.cions. As a result, both coverage and throughput improve as.
the tçrmin~l FTB ratio improves.
In summary, the ESRA use of mini-frames can apply various limits to control the degree of concurrent tr~n~mi~sions, depending on the reception quality at the tçrmin~l locations. Using ESRA terminal classification and 15 sche~l-ling, packet tr~n~mi~sions for all tçrmin~l classes can be successfully received given a specific SIR threshold. This is in contrast to the uncertainty of ~ucces.sful packet reception for the SRA method, the TSRP approach and most contention based multi-access protocols. Furthermore, ESRA performance is stable because its throughput and successful packet tr~n~mi.~sion do not 20 deteriorate with an excessive amount of traffic. For these reasons, the ESRA
method could be used even for real-time traffic such as voice and video services.
In addition, the success probability of one for packet tr~n~micsion can help simplify call admission control and traffic management to ensure a desired levelof QoS.
ESRA performance depends on the correct categorization of terminals.
As the quality of a radio path between any pair of BS and terminal can change over time, perhaps due to seasonal fluctuation or man-made objects, the reception quality can be periodically monitored and, when needed, a terrninal CA 022~3123 1998-11-0~

can be re-classified. To handle temporary fluctuation, the ESRA method can use the upgraded sharing approach to re-transmit packets (i.e., to make sure that class-c packets are re-transmitted in mini-frame k with k ' c) that are not received properly the first time.
For a reasonable radio environment, using practical ~ntenn~ patterns and system parameters, the ESRA method provides 98.64% coverage, and yields a maximum throughput of 36.10% per sector with a success probability of one for packet tr~n~mi~ion.
Although various embo(~iment.c are specifically illustrated and described 10 herein, it will be appreciated that modifications and variations of the present invention are covered by the above teachingc and within the purview of the appended claims without departing from the spirit and intenflçd scope of the invention. For example, although a TDMA system was used to illustrate various embodiments of the invention, it can be appreciated that other systems fall 15 within the scope of the invention. Similarly, although various embo~iment~ ofthe invention make reference to fixed tçrmin~l stations, it can be appreciated that mobile terminal stations could fall within the scope of the invention. Another example includes the number of sectors and cells discussed in the various embo-limçnt~ It can be appreciated that different nurnbers of sectors or cells 20 also fall within the scope of the invention.

Claims (38)

What is claimed is:

1. A method for operating a communications system having a plurality of communications sites and a service area divided into a plurality of sectors, the communications system using a plurality of time subframes scheduled to avoid interference between the plurality of sectors, each subframe being further divided into a plurality of mini-frames, comprising the steps of:
scheduling a first degree of concurrent packet transmissions in a first mini-frame for a first class of communications sites located within each sector;scheduling a second degree of concurrent packet transmissions in a second mini-frame for a second class of communications sites located within each sector, the second degree being different from the first degree; and communicating said packets according to said scheduling.

2. The method of claim 1, further comprising the step of:
classifying the plurality of communications sites into a plurality of classes including the first class and the second class.

3. The method of claim 2, wherein said step of classifying classifies a communications site based on the reception quality of the communications site.

4. The method of claim 3, wherein said step of classifying classifies a communications site based on the signal to interference ratio of the communications site and at least one threshold signal to interference ratio.

5. The method of claim 4, wherein said step of classifying assigns a communications site to the classification having the highest degree of concurrent packet transmission possible and yet maintain a threshold success probability for the threshold signal to interference ratio.

6. The method of claim 5, wherein the threshold success probability is one.

7. The method of claim 2, wherein said step of classifying is performed at the beginning of the operation of the communications system.

8. The method of claim 2, wherein said step of classifying is performed periodically during the operation of the communications system.

9. The method of claim 2, wherein mini-frames in a subframe have different sizes.

10. The method of claim 9, wherein mini-frames are assigned a size selected based on the expected traffic from the associated class of communications sites.

11. The method of claim 9, wherein mini-frames are assigned a size selected based on the number of communications sites in the associated class.

12. The method of claim 1, wherein packet transmissions for one class of communications sites can be transmitted in an alternate mini-frame associatedwith another class of communications sites having a lower degree of concurrent transmission.

13. The method of claim 12, wherein packet transmissions for the one class of communications sites are transmitted in the alternate mini-frame whenever the alternate mini-frame is not being used to capacity.

14. The method of claim 2, further comprising the step of:
selecting an appropriate sector for a communications site based on the classification of the communications site in a plurality of different sectors.

15. The method of claim 1, wherein said scheduling steps create an excess information transmission schedule indicating when excess information for an initial subframe is to be transmitted in other subframes and mini-frames.

16. The method described in claim 15, wherein said other subframes are selected according to a special order.

17. The method described in claim 16, wherein said special order is created by ordering all subframes from a minimum level of interference to a maximum level of interference.

18. The method described in claim 17, wherein said special order is created using a staggered resource allocation protocol.

19. The method described in claim 18, wherein the service area has a plurality of cells, and each cell has six sectors.

20. The method described in claim 19, wherein the system has six subframes, with each sector within a cell being assigned a different subframe.

21. The method described in claim 20, wherein the pattern is created by rotating each cell 120 degrees.

22. The method described in claim 21, wherein said initial subframe for sector one is subframe one, and said special order comprises subframes four, five, three, two and six.

23. The method described in claim 22, wherein when said initial subframe is subframes two through six, said staggered resource allocation protocol staggers said special order by one subframe, respectively.

24. The method of claim 23, wherein said communications sites are divided into six classes, each of the six classes corresponding to one of the six subframes.

25. The method of claim 1, wherein the communication sites are fixed.

26. A method for reusing a common frequency in each sector of a communications system having a plurality or communications sites, comprising the steps of:
identifying major sources of communications interference for the communications sites located within each sector;
evaluating the communications quality of each of said communications sites located within each sector; and scheduling packet transmissions, including concurrent packet transmission, in all sectors to avoid the communications interference and to achieve a threshold communications quality.

27. The method described in claim 26, wherein said scheduling step schedules packet transmissions to maximize concurrent packet transmissions within a cell.

28. A communications system having a plurality of communications sites and a service area divided into a plurality of sectors, the communicationssystem using a plurality of time subframes scheduled to avoid interference between the plurality of sectors, comprising:
a first class of communications units operably associated with the service area for communicating between the communications sites using at least one time subframe;
a second class of communications units operably associated with the service area for communicating between the communications sites using at least.
one time subframe;
a scheduler for scheduling a first degree of concurrent packet communications for the first class of communications sites and scheduling a second degree of concurrent packet communications for the second class of communications sites, the second degree being different from the first degree.

29. The system described in claim 28, wherein said scheduler creates an excess information transmission schedule.

30. The system described in claim 29, wherein said excess information transmission schedule indicates when excess information for an initial subframe is to be transmitted in other subframes.

31. The system described in claim 30, wherein said other subframes are selected according to a special order.

32. The system described in claim 31, wherein said special order is created by ordering all subframes from a minimum level of interference to a maximum level of interference.

33. The system described in claim 32, wherein said special order is created using a staggered resource allocation protocol.

34. The system described in claim 33, wherein the service area has a plurality of cells, and each cell has six sectors.

35. The system described in claim 34, wherein the system has six subframes, with each sector within a cell using a different subframe.

36. The system described in claim 35, wherein the pattern is created by rotating each cell 120 degrees.

37. The system described in claim 36, wherein said initial subframe for sector one is subframe one, and said special order comprises subframes four, five, three, two and six.

38. The system described in claim 37, wherein when said initial subframe is subframes two through six and the staggered resource allocation protocol staggers said special order by one subframe.