DE102004042757B4 - Model based time slot allocation method for throughput optimization in cellular networks with different asymmetric multimedia traffic - Google Patents

Abstract

A method of allocating time slots to uplink and downlink, and assigning users to time slots for throughput optimization in a cellular network consisting of a plurality of cells, each comprising a base station and mobile stations communicating therewith, with different asymmetric multimedia traffic and time multiplexing , characterized in that the assignments are determined by a model-based control comprising the following steps:
• Calculating a risk indicator ranging between 0 and 1 for each user as a measure of the risk that a user will have a carrier-to-interference power ratio below selectable first thresholds or that a user will be a carrier-to-interference Power ratio below selectable first thresholds in at least one other user causes;
Division of users into critical users and non-critical users, where a critical user has a risk indicator above a selectable second threshold, while an uncritical user has a risk indicator below the selectable second threshold;
• Parameterization of protective tapes, where ...

Description

The The present invention relates generally to the field of mobile communications using time division multiplexing. It concerns in particular a model-based timeslot allocation method for throughput optimization in cellular networks with different asymmetric multimedia traffic.

considered becomes a cellular network consisting of a plurality of cells, each of which is a base station and mobile stations communicating therewith includes.

aim A network operator is to maximize throughput, with at the same time the service needs the participant should be optimally satisfied. These differ Among other things, in terms of the demanded data rates and "uplink / downlink" ratios. In this case, the uplink (UL) is understood to mean the communication of one Mobile station towards base station, and downlink (DL) the Communication from a base station to a mobile station. In networks, which support only voice services, the uplink / downlink ratio is 1. However, the use of data services involves asymmetries the uplink and downlink data rates. For example, one requires Data download almost only the downlink channel. Typical asymmetry values for uplink: downlink are between 1: 4 for low multimedia traffic, over 1:10 for medium multimedia traffic, to 1: 200 for high multimedia traffic. A system that can adapt its resources to this situation has the potential to have a higher spectral Achieve efficiency.

So z. As a UTRA (Universal Terrestrial Radio Access) -TDD (Time Division Duplex) -System the radio room in two dimensions: spreading codes and time slots. Unlike radio systems, which only spread codes Use this has the advantage that timeslots the uplink or Downlink traffic depending on demand can be assigned.

1 illustrates the adaptation of a TDD system to a polling traffic asymmetry. As 1 can be seen leads an asymmetric user request with a user bandwidth requirement 11 ie a user request that is asymmetric with respect to uplink (UL) and downlink (DL), to the transition from a symmetric system function 12 to an asymmetric system function 13 , As 1 It can also be seen that, in the case of an asymmetrical user request, a symmetrically operating system is not utilized optimally, since the share 14 can not be used.

Currently a UTRA-TDD variant is developed, which corresponds to the radio frames in 15 Timeslots divided, as well as another variant, which the Radio frames are divided into 7 time slots (TD-SCDMA). Hereinafter All concrete calculations have been made on the basis of the first variant. However, there is no fundamental change before, if the results on the 7-time slot variant or any transmit different number of timeslots become. The only difference is that the first variant has a finer granularity and therefore more accurately adapted to traffic asymmetries.

traffic asymmetries appear in the network as user requests. It's up to the system to decide to what extent and with what response times it meets these requests can. in principle can be the allocation of time slots to the uplink and downlink direction arbitrarily within a time frame, (which is 15 slots numbered by 0 to 14, contains) be. The end of a time slot, after which the system of one Uplink to a downlink changes, or vice versa, is used as a switching point designated. If the slots are not grouped, there is more as a switching point. Otherwise one has a single switchpoint scenario. For a Single switchpoint scenario is intended to be the uplink phase without restriction of generality precede the downlink phase in each frame, and thus each has Cell a switching point, which by the ordered pair (number of Uplink slots, number of downlink slots), the second number is redundant, but for the sake of clarity:

The allocation of time slots to transmission directions (occupancy of a communication frame with uplink and downlink direction) can be controlled in three ways: The first possibility is to exercise no control at all and each time frame in a symmetric SP _i = (7,8) division for the whole cell system. Or an advantage is drawn from the "adaptive switching point" of UTRA-TDD, which has two variants: Either the entire system moves the occupation of its frames synchronously, if the total UL to DL demand asymmetry makes such a movement desirable (Synchronous Asymmetric Switching point, or SASP), that is, SP _i = SP is constant for the entire system, which therefore has a parameter free to be adapted to the traffic asymmetry, namely the number of uplink time slots. Or, thirdly, each cell can autonomously move the uplink (UL) and downlink (DL) assignment (Asynchronous Asymmetric Switching Point or AASP), that is, the assignment may be different for each cell. Thus, "asymmetry" is defined per cell, while "asynchronicity" is defined only for at least two adjacent cells, which show autonomy in determining the UL / DL occupancy. This gives the system flexibility to respond to locally varying, requested UL / DL data relationships. At the same time, the allocation of resources must be such that a desired overall system behavior is maintained. The provision of additional resources (channels) to meet the heavy use demand by means of a switching point shift opens the possibility of achieving higher throughput as long as the asynchronicity does not cause an increase in interference on the channel which is so large that the Profit from ASP is nullified again.

In There is a multi-cell network with adjacent cells with UL / DL synchronicity two types of interference: those at the mobile stations occur and are caused by neighboring base stations, and those that occur at the base stations and by mobile Stations, both in the same cell (intracellular interference) as well as in neighboring cells (intercell interference) become. An asynchronous switch point results in two new types of Interference: cross-mobile interference and cross-base interference. Both are interferences of the "same Entity "and arise due to the fact that if the same timeslot in a cell the Uplink is assigned and the downlink in another, a receiving station the neighboring transferring ones Stations can "hear". This is special problematic if both are in close proximity to the respective cell boundaries are located.

• •
  
  Zahl von Zeitschlitzen im Intervall, wo beide Zellen im Uplink funktionieren,
• •
  
  Zahl von Zeitschlitzen im Intervall, wo Zelle₁ im Uplink funktioniert und Zelle₁ im Downlink funktioniert,
• •
  
  Zahl von Zeitschlitzen im Intervall, wo beide Zellen im Downlink funktionieren.

An evaluation of switching point control strategies now takes place. The TDD mode has synchronized frames. MS ₁ and MS ₂ denote mobile stations in cell ₁ and cell ₂ , respectively, and BS ₁ and BS ₂ denote the associated base stations. A time slot is further characterized by the respective functional mode of the two neighbors as U ₁ U ₂ , D ₁ U ₂ or D ₁ D ₂ , where U stands for uplink and D for downlink. For each pair of neighbors, there is a time frame of three different time intervals, which are indicated by the number of timeslots with stationary transmission direction:

• •
  
  Number of time slots in the interval where both cells work in the uplink,
• •
  
  Number of time slots in the interval where cell _{1 works} in uplink and cell _{1 works} in downlink
• •
  
  Number of time slots in the interval where both cells work in the downlink.

eine genaue Charakterisierung der relativen Schaltpunktsituation bei jeder Zwei-Zell-Interaktion.Thus is

a precise characterization of the relative switching point situation in every two-cell interaction.

• 1. Die Verkehrsanfragen sind symmetrisch, was typisch für Sprachdienste ist (7:8 für einen Rahmen von 15 Schlitzen). Es gibt keine signifikanten Variationen in der Verkehrsdichte und keine UL/DL-Asymmetrie über den gesamten Netzwerkbereich.
• 2. Die Verkehrsanfragen sind UL/DL-aymmetrisch mit einem Verhältnis von 1:4, was ein typischer Wert für das Surfen im Internet ist. Diese Asymmetrie ist an allen Stellen des gesamten Netzwerkbereichs gleich.
• 3. Die Verkehrsanfragen sind UL/DL-asymmetrisch mit einem Verhältnis von 1:4 in einem Teil der Zellen und symmetrisch mit einem Verhältnis von 7:8 im Rest des Netzwerks (räumlich inhomogener Verkehr).

In the following, the three switching point control strategies for the TDD system set out above are compared. The following traffic situations are evaluated as examples:

• 1. The traffic requests are symmetrical, which is typical for voice services (7: 8 for a frame of 15 slots). There are no significant variations in traffic density and no UL / DL asymmetry over the entire network range.
• 2. The traffic requests are UL / DL-Aymmetric with a ratio of 1: 4, which is a typical value for surfing the internet. This asymmetry is the same throughout the network area.
• 3. The traffic requests are UL / DL asymmetric with a ratio of 1: 4 in one part of the cells and symmetric with a 7: 8 ratio in the rest of the network (spatially inhomogeneous traffic).

• 1. Das Schalten ist auf ein UL/DL-Verhältnis von 7:8 (symmetrisch) festgelegt.
• 2. Die Schaltpunkte werden global auf eine Asymmetrie eingestellt, welche auf die Verkehrsanfragen aller Zellen optimiert ist.
• 3. Die Schaltpunkte werden den Verkehrsanfragen lokal angepasst. Diese Kontrollstrategie impliziert das Auftreten von neuen Interferenztypen (Cross-Mobile, Cross-Base).

The network can now respond to these different traffic situations with different shift point control strategies.

• 1. The switching is set to a UL / DL ratio of 7: 8 (symmetrical).
• 2. The switching points are set globally to an asymmetry that is optimized for the traffic requests of all cells.
• 3. The switching points are adapted to the traffic requests locally. This control strategy implies the appearance of new types of interference (cross-mobile, cross-base).

In Table (I) is the achievable throughput gain and associated computational effort as a function of traffic for evaluated a full load situation. The 100 percent reference value is here the symmetric request / symmetric switching point case.

SYSTEMDURCHSATZ UND KOMPLEXITÄT (KLEIN S, MITTEL M) FÜR VERSCHIEDENE VERKEHRS- UND SCHALTPUNKT-SZENARIOS. TABLE I

SYSTEM THROUGHPUT AND COMPLEXITY (SMALL S, MEDIUM M) FOR VARIOUS TRAFFIC AND SWITCH POINT SCENARIOS.

• 1. Für symmetrische Verkehrsanfragen (7:8) erzielt man eine 100 %-Last des Systems, wenn der Schaltpunkt auch exakt auf dieses Verhältnis eingestellt wird. Dies zu implementieren, bedarf keiner besonderen Anstrengung. Eine globa le Mittelung beträgt in diesem Fall das Gleiche. Per Definition gibt es keine lokalen Unterschiede.
• 2. Wenn der Verkehr überall eine 1:4-Asymmetrie aufweist, gibt es für 8 DL-Slots zwei zugehörige Uplink-Slots. Wird jedoch die 7:8-Aufteilung beibehalten, so bleiben fünf Uplink-Slots frei, was bedeutet, dass das System lediglich zu 2/3 ausgelastet ist, d. h. ein resultierender Durchsatz von 66,6 %. Wenn andererseits die Belegung so geändert wird, dass diese Asymmetrie in der Nutzeranfrage (was drei Uplink- und 12 Downlink-Schlitze bedeutet) erfüllt ist, kann das System wiederum voll ausgelastet werden. Dementsprechend beträgt der Gewinn in diesem Fall +33 % des maximalen Durchsatzes oder 50 % in Bezug auf die Situation, wo eine symmetrische Aufteilung belassen wird. Die Anstrengungen bei der Implementierung bestehen darin, die vom Nutzer angefragte UL/DL-Asymmetrie aus allen Basisstationen, möglicherweise bei einem RNC (Radio Network Controller), zu sammeln und die Belegung diesem Wert anzupassen. In dieser Situation gibt es auch keine lokalen Unterschiede.
• 3. Die stärker herausfordernde Situation ist die, wenn die Asymmetrie an verschiedenen Stellen des Netzwerks verschieden ist. Das erste Szenario ist, dass der Verkehr in 2/3 aller Zellen des Netzwerks symmetrisch ist und in 1/3 der Zellen mit 1:4 asymmetrisch ist. Wenn die Gesamtsystem-UL-/DL-Belegung bei 7:8 belassen wird, sieht man sich dem oben genannten 33 %-igem Leistungsabfall in den Zellen mit asymmetrischem Verkehr gegenüber. Die gewichtete Summe der Durchsatzleistungen über dem gesamten Netzwerk ergibt einen Wert von 88,8 % des möglichen Durchsatzes. Die erste Frage ist, ob die Belegung global auf einen Wert eingestellt werden kann, der die Tatsache berücksichtigt, dass der Verkehr des Netzwerks in Teilen hochasymmetrisch ist. Es stellt sich heraus, dass eine 7:8 Belegung auch in diesem Fall das Optimum ist, d. h. besser als jede andere globale Belegung. Nur eine lokal abhängige Belegung kann hier eine Durchsatzerhöhung versprechen.
• 4. Wenn noch mehr Zellen asymmetrisch sind (1/2 und 2/3 aller Zellen), dann ist der erreichbare Durchsatzgewinn noch höher, wenn eine lokal abhängige Belegung verwendet wird.

This results in the following order:

• 1. For symmetric traffic requests (7: 8) you get a 100% load of the system, if the switching point is set exactly to this ratio. Implementing this requires no special effort. Global averaging is the same in this case. By definition, there are no local differences.
• 2. If the traffic has a 1: 4 asymmetry everywhere, there are two associated uplink slots for 8 DL slots. However, if the 7: 8 split is maintained, then five uplink slots remain free, meaning that the system is only 2/3 full, ie, a resulting throughput of 66.6%. On the other hand, if the occupancy is changed to accommodate this asymmetry in the user request (meaning three uplink and 12 downlink slots), the system may again be fully utilized. Accordingly, the gain in this case is + 33% of the maximum throughput or 50% in relation to the situation where a symmetrical division is left. The implementation effort consists in collecting user-requested UL / DL asymmetry from all base stations, possibly in a Radio Network Controller (RNC), and adjusting the occupancy to that value. In this situation, there are no local differences.
• 3. The more challenging situation is when the asymmetry in different parts of the network is different. The first scenario is that traffic is symmetric in 2/3 of all cells in the network and asymmetric in 1/3 of the cells at 1: 4. Leaving the overall system UL / DL occupancy at 7: 8 compares to the above-mentioned 33% power loss in the asymmetric traffic cells. The weighted sum of throughput across the entire network gives a value of 88.8% of the possible throughput. The first question is whether the occupancy can be globally set to a value that takes into account the fact that the traffic of the network is highly asymmetric in parts. It turns out that a 7: 8 occupancy is the optimum in this case as well, ie better than any other global occupancy. Only a locally dependent occupancy can promise an increase in throughput here.
• 4. If more cells are asymmetric (1/2 and 2/3 of all cells), then the achievable throughput gain is even higher if a locally dependent occupancy is used.

table I, which summarizes the state of the art, reveals that in stronger asymmetric Traffic the network throughput far from the desirable 100% is and thus room for improvement leaves.

Out WO 01/99454 A1 is a method for dynamic resource allocation known in a cellular system. The allocation is here through a compromise between performance and connectivity between determined by the cells.

From Ravii Jain, Li & Fung Chang: "Towards an Asymmetric Air Interface Protocol or Wireless Internet Access", The ^8th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, 1997, "Waves of the Year 2000", PIMRC '97, Publication Date: 1.-4. Sept. 1997, Volume 2, pp 688-692 discloses an asymmetric air interface protocol for wireless Internet access.

It So the question arises whether an increase in throughput through asynchronous asymmetric switching point control is possible, that is, whether it is assuming AASP (Asynchronous Asymmetric Switching Point) function of a cellular system that gives control mechanisms, which hold the resulting interference at a value that is not bigger as in synchronous mode. The height the interference is determined by means of the carrier-to-interference power ratio the quality a transmission, and thus the satisfaction or dissatisfaction of the users. to Increased throughput, it is desirable Allowing cells to switch different switching points in an asynchronous manner in so far as one with a suitable control action the total number of dissatisfied users can stick to a number that is not is larger as in the synchronous case.

It Therefore, the technical problem must be solved, a rule specify the controllable parameters switching point assignment and time slot assignment of the user acts and for a given asymmetric traffic the overall throughput and quality of service of the network optimized.

• • Berechnung eines im Bereich zwischen 0 und 1 liegenden Risikoindikators für jeden Nutzer als ein Maß für das Risiko, dass ein Nutzer ein Carrier-zu-Interferenz-Leistungsverhältnis unterhalb von wählbaren ersten Schwellwerten aufweist, bzw. dass ein Nutzer ein Carrier-zu-Interferenz-Leistungsverhältnis unterhalb von wählbaren ersten Schwellwerten bei wenigstens einem weiteren Nutzer verursacht;
• • Einteilung der Nutzer in kritische Nutzer und unkritische Nutzer, wobei ein kritischer Nutzer einen Risikoindikator oberhalb eines wählbaren zweiten Schwellwerts aufweist, während ein unkritischer Nutzer einen Risikoindikator unterhalb des wählbaren zweiten Schwellwerts aufweist;
• • Parametrisierung von Schutzbändern, wobei ein Schutzband einer Zelle den Zellbereich bezeichnet, für welchen der Risikoindikator oberhalb der wählbaren ersten Schwellwerte liegt.
• • Zusammenfassung der Zeitschlitze jeweils einer Übertragungsrichtung (Uplink oder Downlink) zu einem Block, so dass sich ein einziger Schaltpunkt für eine Änderung der Übertragungsrichtung pro Kommunikationsrahmen ergibt und für jede Übertragungsrichtung "sichere" Zeitschlitze vorhanden sind, nämlich solche, in denen der Teil der Zelle für den eine Kontrollaktion optimiert werden soll, mit den nächstliegenden Nachbarn in synchroner Übertragungsrichtung arbeitet;
• • Zuweisung von kritischen Nutzern (Nutzern innerhalb der Schutzbänder) zu sicheren Zeitschlitzen;
• • Ermittlung von optimalen Arbeitspunkten auf einer System-Oberfläche, aufgespannt insbesondere durch die steuerbaren Variablen Uplink/Downlink (UL/DL)-Schaltpunkte, Vektor der Nutzeranzahl in Uplink und Downlink in jeder Zelle und Schutzband-Parametern, umfassend die folgenden Schritte – Modellierung der Nutzer als kontinuierliche räumliche Dichtefunktion; – für jede Wahl eines UL/DL-Schaltpunkts und einer wählbaren Untermenge von Nutzeranzahl-Vektoren: – Berechnung eines Carrier-zu-Interferenz-Leistungsverhältnisses an jedem Punkt im Raum mittels einer Integration aller Pfadverlust-korrigierten Beiträge aus den Nachbarzellen; – Bestimmung für jede Zelle des Anteils der "unzufriedenen" Nutzer, bei denen das Carrier-zu-Interferenz-Leistungsverhältnis unterhalb der wählbaren ersten Schwellwerte liegt; – Gemeinsame Optimierung der Parameter aller Schutzbänder zur Minimierung des Gesamtanteils der unzufriedenen Nutzer im Netz; – für jede Wahl eines UL/DL-Schaltpunkts: – Ermittlung der maximalen Anzahl von Nutzern, die bei Verwendung der jeweils optimalen Schutzbänder mit einem wählbaren Anteil von unzufriedenen Nutzern bedient werden können; – Ermittlung des mit dieser maximal bedienbaren Nutzerzahl erreichbaren maximalen Netzdurchsatzes für diese Schaltpunktwahl bei einer Datenverkehr-Nachfrage mit vorgegebener Uplink/Downlink-Asymmetrie; – Ermittlung der optimalen Schaltpunkte für jede Zelle und des global maximalen Netzdurchsatzes bei einer Datenverkehr-Nachfrage mit vorgegebener Uplink/Downlink-Asymmetrie.

This object is according to the proposal of the invention by a method for assigning time slots to uplink and downlink, as well as for assigning users to time slots for throughput optimization in a cellular network consisting of a plurality of cells, each of which is a base station and communicating therewith Mobile stations comprises, with different asymmetric multimedia traffic and time multiplexing, which is characterized in that the assignments are determined by a model-based control comprising the following steps:

• • Calculating a risk indicator ranging between 0 and 1 for each user as a measure of the risk that a user will have a carrier-to-interference power ratio below selectable first thresholds or that a user will be a carrier-to-interference Power ratio below selectable first thresholds in at least one other user causes;
• Division of users into critical users and non-critical users, where a critical user has a risk indicator above a selectable second threshold, while an uncritical user has a risk indicator below the selectable second threshold;
• • Parameterization of guard bands, where a guard band of a cell designates the cell range for which the risk indicator lies above the selectable first threshold values.
• • Summary of the time slots in each case one transmission direction (uplink or downlink) to one Block, so that there is a single switching point for a change of the transmission direction per communication frame and for each transmission direction "safe" time slots are present, namely those in which the part of the cell for the one control action to be optimized, with the nearest neighbor in synchronous Transfer direction is working;
• Assignment of critical users (users within the guard bands) to secure time slots;
• • Determination of optimal operating points on a system surface, spanned in particular by the controllable variables uplink / downlink (UL / DL) switching points, vector of user numbers in uplink and downlink in each cell and guard band parameters, comprising the following steps - Modeling the User as a continuous spatial density function; For each selection of a UL / DL switch point and a selectable subset of user number vectors: calculating a carrier-to-interference power ratio at each point in space by integrating all path loss-corrected contributions from the neighboring cells; Determination for each cell of the proportion of "dissatisfied" users where the carrier-to-interference power ratio is below the selectable first thresholds; - Joint optimization of the parameters of all guard bands to minimize the overall content of dissatisfied users in the network; - for each choice of UL / DL switching point: - determination of the maximum number of users that can be served using the optimum guard band with a selectable proportion of dissatisfied users; - Determining the maximum network throughput achievable with this maximum usable number of users for this switching point selection for a traffic demand with a given uplink / downlink asymmetry; Determination of the optimal switching points for each cell and the global maximum network throughput for a traffic demand with a given uplink / downlink asymmetry.

In The method of the present invention is the number of timeslots not to carry it out essential. An essential advantage of the invention is that by the proposed optimal choice of switching points for the cellular Network the highest possible data throughput and thus the maximum profitability of the network is achieved. Here it is additional possible, over the Trade-off of high possible throughput profit in the strongly asymmetrically stressed cells against a small one Throughput loss in the cells with symmetrical traffic to control the guaranteed quality of service.

In an alternative embodiment of the method according to the invention can be the process step of modeling the user as continuous spatial Density function by a step of determining the location a user with a selectable Accuracy to be replaced.

In the method according to the invention can the cellular networks meet the specifications of one or more standards from the group consisting of UMTS-TDD, TD-SCDMA and UTRA-TDD, chosen are.

at an advantageous embodiment of the method according to the invention The risk indicator is calculated by placing the cell boundary for each cell surrounding protective tape introduced within which the risk indicator is 1 while he is outside whose 0 is.

Further can in an advantageous way for using a cell along the cell boundary to different neighbor protective bands a different width can be selected.

at a further advantageous embodiment of the method according to the invention a user within the guard band is rated as critical, while a user outside of the protective tape is rated as uncritical.

In the method according to the invention can the selectable first thresholds, a threshold of carrier-to-interference power (C / I) ratio at a mobile station and a carrier-to-interference power ratio threshold at a base station.

Further Is it possible, that the the system surface spanning controllable variables continue one or more Parameter, selected from the group consisting of a threshold value of the carrier-to-interference power ratio at a mobile station, a threshold of carrier-to-interference power ratio at a base station, a level ratio of a target input power at a base station to a target input power of a mobile station and a radius around a base station within which a power control takes place.

Farther can in the process according to the invention when calculating the path loss-corrected (interference) contributions, as well when calculating carrier performance, a log-normal shadowing be carried out for modeling a small-scale fading consisting of an addition a Gaussian Random variables to the path loss calculation steps, to random power fluctuations on a small scale to take into account.

The inventive method will now be closer explains with reference to the attached Drawings is taken:

1 shows the adaptation of the resource allocation to the two directions of transmission according to the traffic to be transported;

2 shows the three categories of time slots that occur in an asynchronous switchpoint scenario;

3 shows a calculation of the risk indicator p _i in a cellular network;

4 shows an example of the value of the risk indicator p _i as a function of the distance from the base station;

5 shows the value of the risk indicator p _i as a function of the distance from the base station for a single guard point scenario with guard band along the cell boundary;

6 shows an interference situation in U ₁ U _{2 time} slots;

7 shows an interference situation in D ₁ U _{2 time} slots;

8th shows an interference situation in D ₁ D _{2 time} slots;

9 shows a cellular network with a hexagonal cell array;

10 shows a constant user density over cell surface in each timeslot in the absence of guard bands;

11 shows the user densities for cells 1 and 2 with guard bands at asynchronicity of 1 time slot;

12 shows the user densities for the cells 1 and 2 with guard bands at asynchronicity of 3 time slots;

13 shows the user densities for the cells 1 and 2 with guard bands at asynchronicity of 5 time slots;

14 shows carrier and interference powers versus cell radius in U ₁ U _{2 time} slots for cell 1;

15 shows the critical limit in U ₁ U _{2 time} slots for cell 1;

16 shows carrier and interference powers versus cell radius in U ₁ U _{2 time} slots for cell 2;

17 shows the critical limit in U ₁ U ₂ -Zeitschlitzen for cell 2;

18 shows carrier and interference powers versus cell radius in D ₁ U _{2 time} slots for cell 1;

19 shows the critical limit in D ₁ U _{2 time} slots for cell 1;

20 shows carrier and interference powers versus cell radius in D ₁ U _{2 time} slots for cell 2;

21 shows the critical limit in D ₁ U _{2 time} slots for cell 2;

22 shows carrier and interference powers versus cell radius in D ₁ D _{2 time} slots for cell 1;

23 shows the critical limit in D ₁ D _{2 time} slots for cell 1;

24 shows carrier and interference powers versus cell radius in D ₁ D _{2 time} slots for cell 2;

25 shows the critical limit in D ₁ D _{2 time} slots for cell 2;

26 shows unordered time slots, ie uplink and downlink in any change;

27 shows ordered time slots, ie an uplink block followed by a downlink block;

The UL and DL time slots are grouped into blocks in the method according to the invention, so that, as in 2 shown, for each cell results in a single switchpoint scenario.

2 shows the time frames of two adjacent cells. In each two-cell interaction, cell _{1 is} the cell with the former UL / DL switch point, and cell _{2 is} the one with the later. The upper timeline is for cell ₁ , which operates here with four uplink (UL) and 11 downlink (DL) time slots, whereas the bottom timeline is for cell ₂ , ie a base station during the symmetric function. 2 shows the three categories of time slots that occur at an asynchronous switching point. During the D ₁ U ₂ phase, users located in the guard band area near the cell boundary are assigned the same direction timeslot as indicated by the arrows. During each time slot category there is a different interference situation. In the D ₁ U _{2 time} slots, an uplink user in cell _{2 may come} into close proximity to a cell ₁ downlink user at their common cell boundary, the reception of the latter then becoming impossible due to the high level of interference. In order to avoid this situation, a risk indicator p _i ∈ [0,1] is associated with each mobile station i both for transmission and reception, which indicates the risk that the user i experiences or causes an unacceptably high level of interference , In a true cellular system, the risk indicators are renewed (or learned as appropriate) as a function of measurements such as C / I _i , RSCP _i (received signal code power) or OTD _i (observed time difference of arrival), as in FIG 3 is shown. At each point in time, each user is either in transmission or in reception mode and thus identified by exactly one risk indicator.

In order to identify the acceptable risk, a threshold P _T ε [0,1] is maintained. A critical user is a user whose risk indicator is above the threshold P _T , as in 4 is shown.

4 shows an example of the value of a user risk indicator p _i as a function of the distance from the base station BS for a real cell.

A guard band designates the cell area (possibly multiply connected) for which the risk indicator is critical, ie above the threshold value p _T. The essential control action to make the situation created by an asynchronous switching point manageable is to assign users to the time slots of the three categories according to their risk indicator. Time slots in which the relevant neighbors are in transmission / reception synchronicity with the Cell ₁ cell are secure slots for cell _i . D ₁ U ₂ slots are not secure. As a control action, secure time slots are assigned to critical users. In other words, the user density over the guard band area is 0 during D ₁ U ₂ timeslots. Since all users within a guard band are by definition critical, they are assigned D ₁ D ₂ slots in cell ₁ and U ₁ U ₂ slots in cell ₂ , as indicated by the arrows in FIG 2 indicated, resulting in increased user density in the guard band for these slots. To aid visualization, for each timeslot, the possible location of users assigned to this timeslot should be colored and the rest left white. Then, without a protective tape, the entire cell is dyed, while the insertion of a protective tape creates colored spots smaller than a cell area, and white protective tapes. Users are generally allocated time slots via a so-called "channel assignment algorithm".

When the traffic is time invariant with a known spatial distribution, a FCA (Fixed Channel Assignment) algorithm usually works best. B. Wegmann, Performance Evaluation of Channel Assignment Algorithm For UTRA-TDD, Proc. IEEE 3G wireless' 2000, San Francisco, 2000, pp. 340-346, compares channel assignment algorithms for UTRA TDD and finds that DCA schemes are superior for homogeneous Poisson voice traffic. Here one understands under channel or resource one for the communication of the user required medium, which z. B. may be a time slot (TS), but also a spreading code, a transmission frequency or a combination thereof. From: Y. Furuya and Y. Akaiwa, Channel Segregation, A Distributed Adaptive Channel Allocation Scheme for Mobile Communication Systems, IEICE Trans., Vol. E 74, No. 6, June 1991, pp. 1531-1537, it appears that a Channel segregation is a good choice. Inhomogeneous traffic is expected to yield even clearer DCA. For the purpose of an AASP investigation, it is assumed that a well-suited channel assignment algorithm is selected for the expected traffic scenario, and that this should now be improved by the addition of an asynchronous switching point. The switch point asynchronicity feature adds another type of inhomogeneity, which the following control algorithm takes into account as an addition to the channel assignment algorithm.

there This control algorithm assumes that for a given asymmetric traffic demand suitable switching points for throughput maximization already found for given maximum customer dissatisfaction are. How this happens is set out following the control algorithm.

Control algorithm:

Cell boundary guard band: In the following analytical treatment, shadow and multipath effects are ignored. Then, proximity to the cell boundary is a sufficient risk indicator, with the guard band being the annular region of the cell that encloses the cell boundary and p _i = 0 outside the guard band and 1 inside as in 5 is shown.

Furthermore, a guard band is quantified by its width, starting from the cell boundary. g ₁ and g ₂ are intended to denote the protective bands in cell ₁ and cell ₂ , respectively.

wenn durch Einführen von Schutzbändern dieser Typ von Interferenz verringert wird, ein

wenn er erhöht wird, und ein "-" für keinen Effekt.Below, the interference situation and the guard band effect are characterized for each time slot category. The indicator in parentheses is a

if this type of interference is reduced by introducing guard bands

if he is raised, and a "-" for no effect.

1st U ₁ U ₂ phase:

6 shows an interference situation during U ₁ U _{2 time} slots. Cell ₁ has no guard band and uniform user density. In cell _{2 there} is a guard band between the inner hexagon and the hexagonal cell border. At any location within cell ₂ , the density of hatching corresponds to the probability of originating from there, assuming that these timeslots have been assigned.

In Zelle₁ stören die Mobilstationen der Zelle₂ die Basisstation der Zelle₁. Dieser Typ von Interferenz nimmt zu, weil allen Zelle₂-Nutzern in der Umgebung der Zellgrenze diese Zeitschlitze zugewiesen werden. In dem Fall von Leistungskontrolle (Übertragungsleistung wird so eingestellt, dass ein konstanter empfangener Leistungspegel innerhalb der Zelle aufrechterhalten ist) liegt ein zusätzlicher erschwerender Effekt darin, dass diese Nutzer mit einer hohen Leistung übertragen, und zwar wegen ihres mittleren größeren Abstands zu ihrer Basisstation BS.Cell ₁

In cell ₁ , the mobile stations of cell ₂ disturb the base station of cell ₁ . This type of interference increases because all cell ₂ users in the cell boundary environment are assigned these timeslots. In the case of power control (transmission power is adjusted to maintain a constant received power level within the cell), an additional aggravating effect is that these transmit users with high power because of their average greater distance to their base station BS.

Cell ₂ (-): In cell ₂ , the mobile stations of cell ₁ disturb the base station of cell ₂ . The control action does not change the allocation of cell ₁ OVERALL during their uplink phase. Thus, the interference power is unchanged.

2. D ₁ U ₂ phase:

7 shows an interference situation during D ₁ U _{2 time} slots. There is a protective band between the inner hexagon and the hexagonal cell border. No users within the guardband are assigned these timeslots.

In Zelle₁ stören die Mobilstationen der Zelle₂ die Mobilstationen der Zelle₁. Diese so genannte Cross-Mobile-Interferenz wird durch Schutzbänder verringert, weil der Abstand zwischen MS₂ und MS₁ zunimmt. Als ein zusätzlicher günstiger Effekt ist die Übertragungsleistung von MS₂ aufgrund der Leistungskontrolle und der kleineren Abstände von MS₂ zu deren Basisstation geringer.Cell ₁

In cell ₁ , the mobile stations of cell ₂ disturb the mobile stations of cell ₁ . This so-called cross-mobile interference is reduced by guard bands as the distance between MS ₂ and MS ₁ increases. As an additional beneficial effect, the transmission power of MS ₂ is lower due to the power control and the smaller distances of MS ₂ to its base station.

In Zelle₂ stört die Basisstation der Zelle₁ die Basisstation der Zelle₂. Diese so genannte Cross-Base-Interferenz ist geringer, weil BS₁ mit einer niedrigeren Leistung überträgt, da deren angeschlossene Mobilstationen näher sind.Cell ₂

In cell ₂ , the base station of cell ₁ disturbs the base station of cell ₂ . This so-called cross-base interference is lower because BS ₁ transmits at a lower power because its connected mobile stations are closer.

3. D ₁ D ₂ phase:

8th shows an interference situation during D ₁ D _{2 time} slots. In cell _2, there is no guard band and the user density is uniform. In cell _{1, there} is a guard band between the inner hexagon and the hexagonal cell border.

In Zelle₁ stört die Basisstation der Zelle₂ die Mobilstationen der Zelle₁. Dieser Interferenztyp ist erhöht.Cell ₁

In cell ₁ , the base station of cell ₂ disturbs the mobile stations of cell ₁ . This interference type is increased.

Because of the preferred assignment of these timeslots to cell ₁ users who are at the cell boundary, they are more disturbed by the cell ₂ base station than if they were homogeneously distributed throughout the cell. One might assume that a spatially homogeneous cell ₁ traffic distribution suffers the same interference from BS ₂ as an annular distribution. In fact, this is related to the path loss model. For power ^drops according to r ^-χ , χ> 1, it can be shown that the more concentrated the traffic at the boundary is, the more interference an annular distribution suffers.

In Zelle₂ stört die Basisstation der Zelle₁ die Mobilstationen der Zelle₂. Vermehrte Interferenz an den Zelle₂-Mobilstationen in dem Fall einer Leistungskontrolle. Dann hat die Verschiebung von an der Zellgrenze lokalisierten Nutzern zu diesen Zeitschlitzen zur Folge, dass die Basisstation BS₁ mit einer vergleichsweise höheren Leistung überträgt.Cell ₂

In cell ₂ , the base station of cell ₁ disturbs the mobile stations of cell ₂ . Increased interference to the cell ₂ mobile stations in the case of performance control. Then, the shift of users located at the cell boundary to these time slots results in the base station transmitting BS ₁ at a comparatively higher power.

These Qualitative considerations will be quantified below and to analyze how the different interference contributions add up, to evaluate which overall change the interference situation by a change in the user timeslot allocation is effected.

In the further analysis expressions are to be derived which describe a user dissatisfaction as functions of the important system parameters. These functions can then be used to identify the subspace in parameter space where the system can operate at the desired dissatisfaction rate.

It Now a description of the analytical procedure is given. For radio systems the third generation, the majority Support data services, the system performance becomes ordinary through an effective data rate, rather than with 2nd generation systems through speech quality, blocked, dropped and lost calls. The achievable data rate depends strongly dependent on the interference situation, measured by the carrier-to-interference power, the C / I ratio. A low C / I ratio leads to an increase in bit error rate (BER), resulting in erroneous decisions can only be improved by error correction algorithms.

If P ^C and P ^I denote the carrier and interference power, followed by BER = BER (P C / P I ) (3) which in turn determines the effective data rate or throughput j _D for each link j link D (BER (C / P I )) = j D (P C / P I ) (4)

angenähert.This is generally a continuous function of p ^C / p ^I. In the following calculations, this function is performed by

approximated.

worin r → die Probenposition an einem willkürlichen Punkt in dem zellularen System bezeichnet. Wird die Aufmerksamkeit auf eine willkürliche Zelle fokussiert, bildet der Bereich von Punkten r →, die Gleichheit in der obigen Gleichung erzielen, eine allgemein geschlossene Kurve um die Basisstation, welche die Lösung ist vonThe fundamental equation that is assumed is the constraint on carrier-to-interference ratio to ensure adequate communication.

where r → denotes the sample position at an arbitrary point in the cellular system. Focusing attention on an arbitrary cell, the range of points r → that achieve equality in the above equation forms a generally closed curve around the base station, which is the solution of

These Curve is called the critical boundary. Their special meaning lies in the fact that they are the space around the base station in two regions divided: those within, where adequate communication is possible, and those outside, where she is not. In other words, by limiting the User satisfaction versus dissatisfaction with the quality of the communication link the following applies: users within the critical limit are with the quality satisfied with their connection while Users outside it is not.

der Probe zu der i'ten Basisstation ab, sondern auch von der Position der empfangenden Stationen r →_i in der Zelle_i. Der Grund hierfür ist die Leistungskontrolle, aufgrund welcher die Basisstation jeden Empfängerort innerhalb der Zelle mit der gleichen Empfangsleistung bedient.The interference power that is caused by all the base stations B _{i of} the network at a sample point r → within the cell of interest depends not only on the distance

the sample to the i'th base station, but also from the position of the receiving stations r → _i in the cell _i . The reason for this is the power control, by which the base station serves each receiver location within the cell with the same receive power.

The interference caused by the transmitting mobile stations of cell _i at a sample point r → depends on r → -r → _Bi -r → _i , the distance of the sample to the location of the transmitted mobile station in cell _i . This distance is at least as large as the sum of the guard bands of the two cells, g ₁ + g _i .

For two adjacent Cells in transfer asynchronicity can be the Interference by enlarging a the protective straps are limited. However, this guard band increase is only up to a certain point possible, beyond which it is not possible is safe slots for all critical users according to the proposed control algorithm to find.

The maximum guard band g _max for a given cell is easily obtained as a function of the particular switching point. For clarity, consider a cell ₁ where a downlink protective tape ( 2 ) is present. All mobile stations within the guard band must be assigned D ₁ D ₂ timeslots. For a sufficiently small guard band, resources in these timeslots also remain for users in the inner cell area. With an increasing guard band, more users must be allocated a constant number of such time slots until they are exactly filled by the users in the guard band, and the users in the inner circle are assigned exclusive D ₁ U ₂ time slots. Assuming a uniform user distribution within the cell, this case exists when the fraction of cell area which is a guard band, denoted by A _g , corresponds to the proportion of D ₁ D ₂ time slots to the total number of DL time slots in cell ₁ . Consequently:

Or with normalized cell radius:

With the same argument, the maximum guard band for cell _{2 is} :

Critical Limits: When the carrier power and interference power at a sample point r → is inserted into the equation (7), one obtains an integral equation which must be solved for r → in order to obtain the critical limit: r Cherith i, j = r Cherith i, j (θ, g 1 , g 2 ), (12) where i denotes the cell (cell ₁ or cell ₂ ), j denotes the time slot, and θ is the polar angle.

Unsatisfied Users: The number of unsatisfied users in cell _i and TS _j is calculated by integrating the corresponding user density function ρ _ij (r →) over the region outside the critical limit. It should be noted that both the density function and the critical limit depend on the selected cell ₁ and cell ₂ guard bands and the time slot category. The total number of unsatisfied users is then:

The first sum runs over all relevant cells. The second sum runs over all timeslots, which belong to the three categories U ₁ U ₂ , D ₁ U ₂ or D ₁ D ₂ . The number of users in cell ₁ and time slot j is denoted by n _ij . The two sums all comprise users who are not satisfied due to a strong uplink and downlink interference in all time slots. Because of the assumption of an effective handover algorithm, the integrand has no fraction outside the traffic density drop length.

calculation the optimal guard band option: The optimal guard bands yield by dissolving

The substantial shares to nr_dissat_users are further in detail discussed below.

For computer modeling In the following, a special case is selected, this being the method according to the invention restrict in any way should.

Model assumptions: Consider a hexagonal cell grid, as in 9 is shown. In 9 For example, all cells that are identical to cell ₁ are striped while all other cells are identical to cell ₂ . Cell ₁ , with an asynchronous switching point, is surrounded by a ring of six identical cell ₂ cells operating in a symmetric mode, again surrounded by a second ring of 12 cells of alternating cell ₁ and cell ₂ types.

however can the principles presented here on every other cell arrangement be applied, which has appropriate computational modifications brings.

All lengths are normalized to the cell radius. Therefore, R = 1 is the cell boundary for one Polar angle 0, which is the radius of the inscribed circle of the hexagons is equal to.

User densities: Traffic extends into the neighboring cell, the so-called "cell overlap area". The fact that handover algorithms only lead to coarsely defined cell boundaries is taken into account by a traffic _{decay length} a _decay . In order for the total traffic borne by the neighboring cells to be constant even in the cell overlap areas, it is a necessary and sufficient condition for the traffic drop to be symmetrical with respect to the cell boundary. As a plausible instantiation, a linear decrease in traffic around the cell boundary is assumed. In the following calculations, a waste length of 8% of the cell radius is assumed to be a realistic value.

Any kind of spatial traffic distribution could be modeled. In the following, a uniform traffic is assumed within the cells, with a linear decrease in the handover areas. Thus, the iso-lines in the handover area are concentric hexagons. In this way, the cell is equivalent to one of the pairs (R, a _decay) or (R _1, R ₂₎ in the latter correspond to a polar angle of 0 to the beginning and the end of the linear range of waste: R 1 = R - a decay R 2 = R + a decay , (15)

und für andere Werte von θ, indem θ durch θ mod(π/6) ersetzt wird, d. h. durch Replizieren der Verteilung in dem am weitesten rechts befindlichen hexagonalen Sektor durch identische Kopien in die anderen fünf Sektoren. Die Funktion ρ(a →, R₁, R₂, r →BSi ) der räumlichen Dichte des Nutzers an dem Punkt a → in der hexagonalen Zelle mit einer Basisstation bei r →BSi ist definiert über ρ ~ (θ, r, R₁, R₂) durch Bewegen des Ursprungs zu dem Punkt r →BSi , d. h. ρ(a →, R1, R2, r →BSi ) = ρ →(arg(a → – r →BSi ), ||a → – r →BSi ||, R1, R2). (17) At the point with the polar coordinates (θ, r) within the sector θ ε [-π / 6, π / 6], the user density for a user is therefore given by:

and for other values of θ, by replacing θ with θ mod (π / 6), ie, by replicating the distribution in the rightmost hexagonal sector by identical copies to the other five sectors. The function ρ (a →, R ₁ , R ₂ , r → BS i ) of the spatial density of the user at the point a → in the hexagonal cell with a base station r → BS i is defined by ρ ~ (θ, r, R ₁ , R ₂ ) by moving the origin to the point r → BS i ie ρ (a →, R 1 , R 2 , r → BS i ) = ρ → (arg (a → - r → BS i ), || a → - r → BS i ||, R 1 , R 2 ). (17)

The user densities in the absence of guard bands are in 10 are shown and are the same for all timeslots and switching point constellations.

wobei die letzte Gleichung gewonnen wird durch Subtrahieren der einheitlichen Dichte über der reduzierten Zellfläche in den D₁U₂-Zeitschlitzen von der ursprünglichen einheitlichen Dichte über die gesamte Zelle, mit einem geeigneten Skalieren durch die entsprechende Anzahl von Zeitschlitzen. Jede Zelle des Typs Zelle₂, mit BS befindlich bei r →BSi , weist ein Schutzband g₂ in den Uplink-Zeitschlitzen auf, was in analoger Weise zu den folgenden Zelle₂-Nutzerdichten für einen Nutzer bei Punkt a → in der hexagonalen Zelle mit der Basisstation bei r →BSi führt:The effects of introducing a guard band will now be examined. Let cell _{i be} at BS r → BS i considered. And let this cell be of the cell ₁ type, which implies that it has a guard band g ₁ in the downlink timeslots (see 2 ) having. Users with downlink traffic are only assigned D ₁ U ₂ timeslots if they are closer than the cell radius minus the guard band at the base station. Thus, the cell _{1 is a} spatial density function (indicated by the upper index in the densities below) for a user at the point a → in the hexagonal cell having a base station r → BS i in each of the three timeslot categories:

the last equation being obtained by subtracting the uniform density over the reduced cell area in the D ₁ U ₂ time slots from the original uniform density over the entire cell, with appropriate scaling by the appropriate number of time slots. Each cell of type Cell ₂ , with BS present at r → BS i , has a guard band g ₂ in the uplink timeslots, analogous to the following cell ₂ user densities for a user at point a → in the hexagonal cell with the base station r → BS i leads:

The densities for all timeslots are in 11 for the least asymmetry, SP = (6, 1, 8), in 12 for medium asymmetry, SP = (4, 3, 8) and in 13 for the maximum allowed asymmetry SP = (2, 5, 8), each time for a guard band value of g1 = g2 = 0.15 in all cells.

It be on the unchanged User densities in those time slots that are not affected by the guard bands are pointed out, as well as on the asymmetry and guard band dependency the user densities in the other time slots.

Services: In the method according to the invention Any form of transmit power control can be modeled, including the Absence of such. It becomes a realistic example of application ideal receive level-based power control used at a constant reception level of P target / MS → BS at a base station and P target / BS → MS guaranteed at a mobile station is.

worin R_PC der Radius um den Transmitter ist, innerhalb dessen eine Leistungskontrolle effektiv ist, und worin χ der Pfadverlustkoeffizient ist.The following power comes in the case of receive power control at the BS from a connected mobile station at a distance r:

where _{RPC is} the radius around the transmitter within which power control is effective, and where χ is the path loss coefficient.

The following power comes in the case of receive power control at point r → from a mobile station located at point a → and with BS _i located at r → BS i , connected is:

The following Power comes in the case of receive power control at distance r from the BS:

The following performance comes in the case of receive power control at the point r → of one r → BS i BS, which transmits to a mobile station located at a →:

Because of appropriate precautionary measures in the usual standards becomes one Intra-cell interference, d. H. Interference at the own cell not considered.

It Now a time slot to time slot analysis is performed.

It a calculation of the interference situation for all time slots takes place and their associated User dissatisfaction rates.

Cell ₁ , time slots U ₁ U ₂ : The carrier power arriving at the base station of cell ₁ from a connected mobile station at r → is from (20):

The interfering power from the mobile stations, which are connected to the identical neighboring cells in the first ring arrives at the base station of cell _1, using (21) and (19):

14 shows cell ₁ carrier and interference powers across the cell radius in U ₁ U _{2 time} slots (both cells in the uplink), shown for different guard band values g ₂ independent of g ₁ .

wird durch die MS₂-Transmitter verursacht und passiert an der Basisstation 1. Deshalb hat sie keine radiale Abhängigkeit. Sie ist von der Wahl des Schutzbands g₁ unabhängig, da in TS U₁U₂ die Nutzerdichte in Zelle₁ durch g₁ nicht beeinflusst ist. Sie wird umso stärker, je breiter das Schutzband g₂ während der D₁U₂-Zeitschlitze gemacht wird, weil es wegen der veränderten Zuweisung von Nutzern zu den U₁U₂-Zeitschlitzen eine über dem Mittel befindliche Zahl von Nutzern gibt, die sich nahe der Zellgrenze befinden. An mancher Stelle ist es möglich, dass die Interferenz die erforderliche Schwelle überschreitet, was für diesen Zeitschlitz zu einer 100 % Nutzerunzufriedenheit führt. Für die gegebene Parametereinstellung erfolgt dies nicht für

≥ 3. Für

= 2, welches die minimal zulässige Zahl von Uplink-Zeitschlitzen ist, wird der Nullwert für die Kostenfunktion in (26) für g₂ = 0,15 erreicht, und größere Schutzbänder bewirken negative Kostenfunktionswerte. Da g_2,max gemäß (11) 0,1548 beträgt, stellt dies lediglich eine geringe Einschränkung des Schutzbandbereichs dar.The interference

is caused by the MS ₂ transmitters and happens at the base station 1. Therefore it has no radial dependence. It is independent of the choice of guard band g ₁ since in TS U ₁ U ₂ the user density in cell ₁ is not affected by g ₁ . It becomes stronger the wider the guard band g _{2 is made} during the D ₁ U ₂ timeslots because there is an above-average number of users due to the changed assignment of users to the U ₁ U _{2 time} slots located near the cell boundary. In some places, it is possible for the interference to exceed the required threshold, resulting in 100% user dissatisfaction for this timeslot. For the given parameter setting, this is not done for

≥ 3. For

= 2, which is the minimum allowed number of uplink time slots, the zero value for the cost function in (26) for g ₂ = 0.15 is achieved, and larger guard bands cause negative cost function values. Since g _{2, max} according to (11) is 0.1548, this is only a small limitation on the guard band area.

ist die Kurve, auf welcher die Kostenfunktion

über r → zu 0 wird. Die Kurve

ist in 15 gezeigt, für verschiedene Schutzbandwerte von g₂ (unabhängig von g₁).The critical limit

is the curve on which the cost function

via r → becomes 0. The curve

is in 15 shown for different guard band values of g ₂ (independent of g ₁ ).

The critical area A ^crit is the area between the critical boundary and the boundary of the area within which the user density for this cell (cell + handover area) is greater than zero. The number of dissatisfied users in cell ₁ is given by

The relative dissatisfaction rate depends only on the critical limit and not explicitly on g ₂ or g ₁ ; therefore it depends only on g ₂ and not on g ₁ .

Cell ₂ , time slots U ₁ U ₂ :

Cell _{j should} belong to the first ring of cell ₁ neighbors. The carrier power arriving at the base station of cell from a connected mobile station of r → is, with (20):

The interfering power at the base station of cell from the mobile stations connected to its first ring of neighbors comes by the assumption of 9 of three cells that are identical to cell ₁ (cells 1, 8 and 10) and three cells that are identical to cell ₂ (cells 3, 7 and 9). Thus, using (21), (18) and (19),

16 shows cell ₂ carrier and interference powers across the cell radius in U ₁ U ₂ timeslot (both cells in the uplink), for different guard band values g ₂ (independent of g ₁ ).

This interference is analogous to interference 1_U ₁ U ₂ . It is caused by both the MS ₁ and MS ₂ transmitters and happens at the base station j. Therefore it has no radial dependence. The larger the guard band g _{2 is} chosen in the aggressor cells, the larger this second term is. There is no dependence on the choice of protection band g ₁ , because in cells of type cell ₁ the traffic changes only under the downlink time slots is assigned. The comparatively smaller interference power compared to 1_U ₁ U ₂ results from the fact that the interfering user traffic is homogeneously distributed in the three cells of the cell ₁ type and does not occur predominantly at the cell boundary.

ist die Kurve, auf welcher die Kostenfunktion

über r → zu 0 wird. Die Kurve

ist in 17 gezeigt, für verschiedene Schutzbandwerte g₂ (unabhängig von g₁).The critical limit

is the curve on which the cost function

via r → becomes 0. The curve

is in 17 shown for different guard band g ₂ (regardless of g ₁ ).

d. h. die Summe über den ersten Ring von Nachbarn j, j = 2, ..., 7, der Zahl von Zelle_j-Nutzern in diesem Zeitschlitz, die sich außerhalb der kritischen Grenze von Zelle_j befinden. Die relative Unzufriedenheitsrate hängt von der kritischen Grenze und über ρ von g₂ ab (jedoch nicht von g₁). Deshalb hängt sie doppelt von g₂ ab (und überhaupt nicht von g₁).The number of dissatisfied users in the first ring is given by

ie, the sum over the first ring of neighbors j, j = 2, ..., 7, the number of cell _j users in that timeslot that are outside the critical boundary of cell _j . The relative dissatisfaction rate depends on the critical limit and on ρ of g ₂ (but not on g ₁ ). Therefore, it depends twice on g ₂ (and not at all on g ₁ ).

Cell ₁ , time slots D ₁ U ₂ (cross-mobile interference):

The carrier power arriving at a mobile station at r → in cell ₁ from the base station is (22):

The interfering power arriving at this mobile station located at r → in cell ₁ from the mobile stations connected to the identical neighbor cells in the first ring is, using (21) and (19):

18 shows cell ₁ carrier and interference powers across the cell radius in D ₁ U _{2 time} slots (those with uplink downlink asynchronicity), for different guard band values g ₂ (dependent on g ₁ only over the distribution of the users affected by interference).

This interference is caused by the MS ₂ transmitters and happens at the MS ₁ mobile stations. It decreases with an increase in either the buffer zone g ₂ or g ₁ . For the maximum possible choice of g ₂ or g ₁ , it is completely suppressed or ineffective. Therefore, the optimal guard band choice can only be made by taking into account all interference contributions together.

ist die Kurve, auf welcher die Kostenfunktion

über r → zu 0 wird. Die Kurve

hängt nur von g₂ ab (und nicht von g₁), und von P target / MS→BS/P target / BS→MS. Sie ist in 19 gezeigt. Die Zahl von unzufriedenen Nutzern in Zelle₁ ist gegeben durchThe critical limit

is the curve on which the cost function

via r → becomes 0. The curve

depends only on g ₂ (and not g ₁ ), and on P target / MS → BS / P target / BS → MS. she is in 19 shown. The number of dissatisfied users in cell ₁ is given by

This number depends on the critical limit and on ρ of g ₁ (but not on g ₂ ), and thus on both g ₁ and g ₂ .

Cell ₂ , time slots D ₁ U ₂ (cross-base interference):

The cell _{j should} belong to the first ring of neighbors of cell ₁ . The carrier power arriving at the base station of cell _j from a connected mobile station at r → is (20):

The interfering power arriving at the base station of cell _j from its first ring of neighbors comes through the assumption of 9 of three cells that are identical to cell ₁ (cells 1, 8, and 10) that are in the downlink, and three cells that are identical to cell ₂ (cells 3, 7, and 9) that are in the uplink are. Thus, using (21), (18), and (19) follows:

20 shows cell ₂ carrier and interference powers across the cell radius in D ₁ U ₂ timeslot (those with uplink downlink asynchronicity), for different guard band values g ₁ and g ₂ .

Both interferences occur at BS _j . The first term is caused by the three BS _i type 1. It decreases with an increase in the buffer zone g ₁ + g ₂ . At the maximum possible value of g ₁ + g ₂ , it is completely suppressed for the parameters used in this scenario. The second term is caused by the MS _i transmitters in the three neighbors of type 2. It decreases with an increase of g ₂ . Two effects play a role here: a) As g ₁ and g _{2 increase} in the power control area, the transmitters transmit less power because their receiving stations are closer. b) For the first term, the distance between the base stations is roughly a factor of 2 larger than for the other interferences discussed here. Depending on the path loss law used here, this leads to a factor of 8 to 11 in the power decrease.

ist die Kurve, auf welcher die Kostenfunktion

über r → zu 0 wird. Die Kurve

j = 2, ..., 7 hängt von sowohl g₁ als auch g₂ ab und von P target / MS→BS/P target / BS→MS, obgleich weniger stark als in Zelle₁, da nur eine Untermenge von Nachbarn in Uplink-Downlink-Asynchronizität mit Zelle_j ist. Sie ist in 21 gezeigt.The critical limit

is the curve on which the cost function

via r → becomes 0. The curve

j = 2, ..., 7 depends on both g ₁ and g ₂ and on P target / MS → BS / P target / BS → MS, although less strongly than in cell ₁ , since only a subset of neighbors in Uplink downlink asynchronicity with cell _j . she is in 21 shown.

The Number of dissatisfied users in the first ring is given by

The relative dissatisfaction rate depends on the critical limit and on ρ of g ₂ (but not on g ₁ ), and thus on g ₁ and twice on g ₂ .

Cell ₁ , time slots D ₁ D ₂ :

The carrier power arriving at a mobile station at r → in cell ₁ from the base station is (22):

Verkehrsdichte in Zelle₂ während der D₁-D₂-Zeitschlitze hängt nicht vom Schutzband g₂ ab.The interfering power arriving at this mobile station located at r → in cell ₁ from the base stations of the identical neighbor cells in the first ring is, using (23) and (19):

Traffic density in cell ₂ during the D ₁ -D ₂ timeslots does not depend on guard band g ₂ .

22 shows the cell ₁ carrier and interference powers across the cell radius in D ₁ D _{2 time} slots (both cells in the downlink). The interference power is independent of both g ₁ and g ₂ .

ist die Kurve, auf welcher die Kostenfunktion

über r → zu 0 wird. Die Kurve

hängt nicht von g₂ oder g₁ ab und ist in 23 gezeigt.The critical limit

is the curve on which the cost function

via r → becomes 0. The curve

does not depend on g ₂ or g ₁ and is in 23 shown.

The number of dissatisfied users in cell ₁ in these timeslots is given by

This number depends on the critical limit and on ρ of g ₁ (but not on g ₂ ). Therefore, it depends on g ₁ and not g ₂ .

Cell ₂ , time slots D ₁ D ₂ :

Assume that cell j belongs to the first ring of neighbors of cell ₁ . The carrier power arriving from the base station of cell _j to a connected mobile station at r → is (22):

zeigt die D₁D₂-Träger- und Interferenzleistungen in Zelle₂ für eine verschiedene Wahl von Schutzbandwerten.The interference power at this mobile station at r → in cell _j from the first ring of After arrives by acceptance of 9 of three cells that are identical to cell ₁ (cells 1, 8 and 10) and three cells that are identical to cell ₂ (cells 3, 7 and 9), all of which are in the downlink. Thus, using (23), (18) and (19) follows:

shows the D ₁ D ₂ carrier and interference powers in cell ₂ for a different choice of guard band values.

The traffic density in cell ₂ during the D ₁ D ₂ timeslots does not depend on the choice of guard band g ₂ . However, the interference depends on the choice of guard band g ₁ in cell ₁ . The critical users near the boundary are reassigned to D ₁ D ₂ slots and cause additional interference.

j = 2, ..., 7 ist die Kurve, auf welcher die Kostenfunktion

über r → zu 0 wird. Die Kurve

hängt nur von g₁ ab (und nicht von g₂), und ist in 25 gezeigt, für verschiedene Schutzbandwerte g₁ (unabhängig von g₂).The critical limit

j = 2, ..., 7 is the curve on which the cost function

via r → becomes 0. The curve

depends only on g ₁ (and not g ₂ ), and is in 25 shown for different guard band values g ₁ (independent of g ₂ ).

The Number of dissatisfied users in cell, j = 2, ..., 7, in these Time slots are given by

This number depends only on the critical limit and not explicitly on g ₁ or g ₂ and thus only on g ₁ and not g ₂ .

Important system parameters:

Calculation parameters:

R:

Radius of the circle, which is inscribed in hexagonal cell

N _R :

Granularity for iteration along radius

N _α :

Granularity for iteration along angle

worin Nⁱ _i die Einheit von Zelle_i mit seiner Nachbarschaft bezeichnet. Die Anzahl von unzufriedenen Nutzern in Zelle_i hängt von der Schaltpunktkonfiguration sowohl durch die Nutzerdichten der Nachbarzellen in den jeweiligen Zeitschlitzen und direkt von der Anzahl der Nutzer in Zelle_i ab.Tables II and III show the parameter dependence of the critical limit and the number of dissatisfied users. N ₁ denotes the neighborhood of cell N _i . The critical limit in cell _i depends on the switching point configuration by the user densities of the neighboring cells, which are changed in certain time slots by the introduction of a guardband. TABLE II

where N ⁱ _i denotes the unit of cell _i with its neighborhood. The number of dissatisfied users in cell _i depends on the switching point configuration both by the user densities of the neighboring cells in the respective time slots and directly by the number of users in cell _i .

Selected parameter values:

The parameter values used for the 14 to 27 are shown in Table IV.

TABLE III

TABLE IV

The choice of C / I _crit values is worth commenting on. We know that when each cell _{1 has} an SP = (7, 0, 8), there is the synchronous, symmetric mode of operation which is known to be ten users in each timeslot at an acceptable dissatisfaction rate, roughly not in the chosen calculation example over 3%, supported. The same is true for any other synchronous function mode, meaning that all cells have the same switching point (n, 15-n). Then there is no asynchronicity, and thus no guard bands, and the dissatisfaction rate is the same as in the symmetric case. This is the special case that the parameter selection must satisfy. Thus, the parameters must be chosen to satisfy these constraints in both the U ₁ U ₂ and D ₁ D ₂ time slots in the absence of guard bands, namely by leaving ten users in each timeslot at a dissatisfaction rate of not more than about Support 3% (a lower value can be selected, which leads to more stringent conditions for the parameters).

As Table III shows, the dissatisfaction rate depends on (C / I) BS / crit in the U ₁ U _{2 time} slots, and on (C / I) MS / crit in the D ₁ D _{2 time} slots. Setting a maximum for the dissatisfaction rate restricts the respective (C / I) _crit values. The exact dependencies are given in Tables V and VI. The tables are identical for cells 1 and 2 because of complete symmetry in the synchronous case with no guard bands. Since it makes no sense to apply harsher conditions to the mobile station than to the base station (usually the BS has more powerful hardware and more complex algorithms and thus can cope with tougher conditions), it can be seen that a maximum dissatisfaction level of 3% (C / I ) _crit values of 0.14 (0.142 in Table VI - 0.145 is already noticeably too high), while a _maximum dissatisfaction level of 2% requires (C / I) _crit values of 0.13.

TABLE V

TABLE VI

Below is a description of the results. It is now necessary to calculate the guard band dependency of the overall relative dissatisfaction rate in the network for all possible switching points from symmetric (7, 0, 8) to maximum asymmetric (2, 5, 8). The underlying formulas are as follows: For a given SP and user load, the total number of users in the cell is _i :

oder gemessen in %:For a given SP, user load and guard band control action, the number of unsatisfied users in cell _{i is} :

or measured in%:

Of the For simplicity, assume that all cells are the same Type and all timeslots in the same category identical Have loads, with the relevant user numbers in a matrix with 6 entries can be captured wherein the first index indicates the cell type and the second indicates the cell type Time-slot type:

oder gemessen in %:Will the cell grid of 9 infinitely extended by repeating the basic cell ₁ + cell ₂ + cell _{3 building} block, this results in a proportion of 1/3 cells of type cell ₁ and 2/3 cells of type cell ₂ , which is the division of row 3 in Table I. The total number of dissatisfied users in the network is then

or measured in%:

For each switching point SP and load n, the number of dissatisfied users in the network (due to insufficient C / I) depends on the choice of the two guard bands g ₁ and g ₂ . These can be chosen to minimize the number of unsatisfied users in the network due to insufficient C / I. The pair of optimal guardband control and dissatisfaction rate for each switching point is:

The Table VII indicates the result of optimization over the guard bands, d. H. the minimum achievable overall dissatisfaction rate in the network dependent on from load and switching point from symmetrical (7, 0, 8) to maximum asymmetrical (2, 5, 8), along with the optimal guard bands. In this table is only the functional area highlighted, which is considered interesting becomes, with a resulting dissatisfaction rate around DissCeiling around. The load n, d. H. the number of users supported in each timeslot should, was chosen so that they are the highest is for which is a guard band control, which is the setting 3% maximum dissatisfaction enough. It can be read from Table VII, and is convenience half summarized in Table VIII.

It Now throughput calculations are done to find an optimal strategy for one to find asynchronous asymmetric needs.

For a given switching point SP, which is exactly indicated by the number A of overlapping time slots (ie SP = (7-A, A, 8)) and a given dissatisfaction maximum value DissCeiling, a possible load n A / DissCeiling is a load of the form of Equation (52) for which there are guard bands g _i which guarantee a maximum overall dissatisfaction rate no greater than DissCeiling.

TABLE VII

TABLE VIII

worin der Systemdurchsatz T^A anschließend definiert wird. Eine Zunahme der Zahl von Nutzern in einem der sechs Eingänge des Vektors der Nutzerlasten würde zu einer höheren Gesamt-Unzufriedenheitsrate führen. In dem Fall von A = 4 enthält Tabelle VII zwei Lastvektoren, welche DissCeiling genügen, nämlich [8, 8; 10, 10; 9, 9] und [7, 7; 10, 10; 10, 10]. Jedoch ist in diesem Fall ersichtlich, dass der zweite Lastvektor zu einem höheren Durchsatz führt, da die letzten beiden Eingänge den 8 D₁D₂-Zeitschlitze entsprechen, und somit ein höheres Gewicht tragen als die ersten beiden Eingänge, welche den drei U₁U₂-Zeitschlitze entsprechen.As maximum load n ^{*, A} , which can be served with A asynchronous time slots and a selectable DissCeiling, the load vector is defined, which maximizes the system throughput:

wherein the system flow T ^{A is} subsequently defined. An increase in the number of users in one of the six inputs of the user loads vector would result in a higher overall dissatisfaction rate. In the case of A = 4, Table VII contains two load vectors which satisfy DissCeiling, namely [8, 8; 10, 10; 9, 9] and [7, 7; 10, 10; 10, 10]. However, in this case, it can be seen that the second load vector results in a higher throughput since the last two inputs correspond to the 8 D ₁ D ₂ slots and thus bear a higher weight than the first two inputs corresponding to the three U ₁ U _2-time slots correspond.

The following procedure is used to determine whether and by how much the switching point in a cell _{1 should be} shifted, assuming a certain asymmetric traffic request from this cell ₁ , which is asynchronous with the surrounding cells. The asymmetric traffic request can be characterized in two ways. Either through a pair of specific requests for UL and DL, or through the relationship between the two, which is the more common case for large amounts of data, and is thus discussed below.

Determine the optimal throughput for a given ratio a requested UL to DL traffic.

The requested traffic in the type 2 cells is symmetrical. For a given period spanning many frames, the UL and DL traffic demanded in cell ₁ r is.

Important here is that for a transmission available Traffic in a given framework the same ratio r must follow. In other words, if intended, one DL traffic of size DLT in next Frame to use, then is the amount of UL traffic available ULT = r · DLT. If conversely, it is intended to have UL UL traffic in the UL next Frame to operate, then the amount of operable DL traffic DLT = 1 / r × ULT. This is because, if you have to send two large amounts of data, which are lined up in two queues, one for UL and one one for DL, and this data is a UL to DL ratio r then you can not get the ones to send in a given frame Data volume independent pick out. If that were done then the shorter line would be preferred what the relationship r would worsen more and more, and also, each switching point would then have a throughput of 100%, as long as both beats are not empty. The only reasonable service criterion to compare the overall performance of different switching point, the data in the same ratio how to use their appearance in the queues, namely r.

Back to the question, which is the best choice of SP for a given traffic ratio r, and which are the throughputs that can be achieved with each choice of SP. It will use the same model as before, as in 9 a quite typical scenario, namely 1/3 cells with an asymmetric traffic and 2/3 cells with a symmetrical as a concrete instantiation of the simulations performed. Since, as a result of the previous calculations, the maximum number of supported users is known for each potential switch point position, the question of optimum SP and throughput can now be accurately answered, both in cell ₁ and cell ₂ , so as not to exceed the required dissatisfaction value. As before, an excess of DL traffic is assumed, d, h. the DL traffic is available virtually without limits. Let the intended number of asynchronous time slots be A, SP as before SP = (7 - A, A, 8) (58)

In the following, the appended letter a denotes the available traffic for the next frame, s the traffic operable in the next frame (limited by the number of available resources, ie timeslots times number of users of the next frame, and c the traffic The served traffic is the minimum of available and serviceable traffic.This expresses the amount of served DL traffic, with SP = (7-A, A, 8) and a given dissatisfaction rate (which the number n of maximum supported users) in cell ₁ :

And in cell ₂ :

Due to the declared connection between the available UL and DL traffic, the amount of available UL traffic in cell _{1 is} : Ulta A 1 = r · DLTc A 1 (61) and in cell _{2 it} is: Ulta A 2 = 7/8 · DLTc A 2 (62)

Verkehr mit dem gegebenen SP und Unzufriedenheitsrate in Zelle₁ bedient werden, und

in Zelle₂. In dieser Analogie entspricht dies dem Fall, dass man verschieden große Kabinen hat, die auf den Berg hinaufgehen, so dass einige Leute für die nächste Kabine zurückgelassen werden. Somit ist die Menge des bedienten UL-Verkehrs in Zelle_i: ULTcAi = min{ULTaAi ,ULTsAi } (65) If you've ever waited in a long queue (for example, to go up a mountain or in an amusement park), this is the crowd from the long queue that will be brought into the waiting room for the next serving. However, there is no guarantee that all such traffic will be served in the next frame. Regardless of how much UL traffic is lined up, there can be no more than a lack of available resources

Traffic with the given SP and dissatisfaction rate can be served in cell ₁ , and

in cell ₂ . In this analogy, this is the case of having different sized cabins going up the hill, leaving some people behind for the next cabin. Thus, the amount of UL traffic served in cell _{i is} : ULTc A i = min {ULTa A i , ULTs A i } (65)

Hereby the throughput for SP = (7-A, A, 8) in cell _i T A i = ULTc A i + DLTc A i (66) and for the entire network of 1/3 cells of type cell _i and 2/3 of type cell ₂ , one obtains the overall average system throughput T A = 1/3 · T A 1 + 2/3 · T A 2 (67)

Then Of course, the optimum switching point is the one through given A, which is the total mean system throughput maximizes:

In Table IX, the optimization derived above for the example treated in Table I, line 3, namely a 1: 4 UL / DL asymmetry (r = 0.25) in cell ₁ and 7: 8 in cell is performed ₂ . The maximum supported user numbers for each TS configuration are from Table VIII.

TABLE IX

It can be seen from the last two lines of Table IX that in this case of moderately asymmetric traffic, a 7.4% throughput gain can be achieved with respect to the symmetric SP by allowing the switch point in cell _{1 to} be asynchronous to that moved to cell ₂ . Three points are to be noted: First, throughput can actually be increased by allowing an asynchronous SP while maintaining the desired level of user dissatisfaction; secondly, letting the overlapping timeslots carry the whole burden of throughput reduction, it is not optimal, since all null protection bands are a special case of the considered control action, and the throughput optimization results in 0 different optimal guard bands; third, the optimal switching point motion in cell ₁ is not SP = (3, 4, 8), as expected from the offered 1: 4 asymmetry, which would be perfectly served by a 3:12 UL / DL time slot allocation. Rather, as Table IX shows, it is SP = (4, 3, 8) which achieves the greatest system throughput due to the loss of throughput in the symmetric neighbors for greater asymmetry.

Since UL / DL asymmetry of 1: 4 is very moderate, now the achievable system throughput gain for a medium multimedia traffic profile with associated asymmetry ratio of 1:10 in Table X, and for high-multimedia traffic profile with associated asymmetry ratio of 1: 200 in Table XI, in Line with reference UMTS / IMT-2000 Spectrum, Report No. 6 from the UMTS Forum, December 1998. In both cases, the optimal switch point position (3, 4, 8), ie four opposite time slots, is the same order of magnitude (6.7 and 5.8%, respectively) with an overall system throughput increase. However, the increase in throughput in cells of type cell _{1 is} significantly more significant, on the order of 50%, compared to a very moderate increase in the rate of 6% in cells of type cell ₂ . Thus, to achieve low overall delays and satisfied users, it seems highly desirable to accept this small decrease in symmetric traffic cells to gain the 50% increase in throughput in the highly asymmetric cells.

TABLE X

TABLE XI

Conclusions: Herewith the answer to the question at the beginning is that one asynchronous switching point possible is, d. H. the proposed control action on guardbands hold the resulting interference at values not greater than those you find in a synchronous system function. Consequently The potential throughput can be determined by the AASP feature is promised by TDD.

The price in terms of a function controller is not negligible. The controller must allocate and reject users at timeslots, depending on their risk of experiencing unacceptable levels of cross-interference. The reassignment of the users generates an additional signal load. However, it is anticipated that such redistributions will be a fairly "normal" process and that this will not severely affect the performance gain. With regard to an economic question, it must be asked whether the demonstrated throughput gain of the order of 50% for the strongly asymmetric traffic cells, compared to a very moderate throughput loss of 6% in the symmetric traffic cells justifies the development costs. An important finding is that the timeslots should be grouped into a compact uplink and a compact downlink phase. In 26 uplink and downlink directions are randomly assigned for each timeslot. If then a user has a high risk indicator It is computationally very expensive to check whether there are "safe" time slots in which all neighboring cells are in the same mode, so that it is guaranteed that one does not encounter this form of high interference. On the other hand, "safe" time slots are virtually built in by the block design when the time slots are ordered, as in 27 is shown. The guard band control action not only assigns secure time slots to users at risk of experiencing unacceptable large interference of the same entity, but equally makes it so with potential causers thereof.

The Calculations supported on a number of idealizations. The traffic distributions in real systems are spatial not homogeneous, as is assumed here, are cellular systems not hexagonal, and there are shadow and reusable effects. however can be the identified Systemei properties, the relative Sizes of Interference types (base-base, Cross-mobile, base-mobile, mobile-base) and the effect of the proposed Transfer control actions to real systems. The core result is that a "system interface" exists and the having shown properties. For each functional point of the cellular system, there are optimal guard bands in terms of interference minimization. The optimization of the functional point of the system only makes on long Time scales sense (in the order of magnitude the duration of connections, d. H. Minutes). On the short time scale dominate random Fluctuations (eg short-term concentrations of users in a specific location in a cell). The system should not be on this Fluctuation short-term scale can be optimized. Instead, should the realistic system interface for every Switching point arrangement determined by an adaptive algorithm become. To train this algorithm will not be the whole Parameter space because otherwise risk dramatic power losses would. From the investigations it follows that the cross-mobile interference with small protective tapes dramatically increases. Reacted towards the larger guard bands the system is very good and you can make a successive vote, for optimal system performance.

• 1. Bei der Berechnung der Pfad-verlust-korrigierten Beiträge zur Interferenzleistung, sowie bei der Berechnung der Carrierleistung, kann ein Log-normal shadowing zur Modellierung eines small-scale fading durchgeführt wird, bestehend aus einer Hinzufügung einer Gauß'schen Zufallsvariablen zu den Berechnungsschritten für den Pfadverlust, um zufällige Leistungsschwankungen im kleinen Maßstab zu berücksichtigen.
• 2. Das verwendete Zellmodell aus 9 in Richtung nichtperiodischer, asymmetrischer Zellmuster zu erweitern. Zelle₁ wäre dann nicht durch 6 symmetrische Zellen umgeben, sondern durch 6 Zellen mit eigenem zufälligen asymmetrischen Bedarf. In diesem Fall müssen entlang der Zellgrenzen zu verschiedenen Nachbarn innerhalb einer Zelle Schutzbänder mit unterschiedlichen Breiten gewählt werden. Dies ändert vor allem die Komplexität der Optimierung wegen der höheren Dimensionalität des Variablenraums, nicht aber die grundlegenden, hier aufgezeigten Vorgehensweisen und Verhältnisse.

The method according to the invention can easily be extended in the direction of greater generality. Two obvious extensions are:

• 1. When calculating the path loss corrected contributions to the interference power, as well as in the calculation of the carrier power, a log-normal shadowing for modeling a small-scale fading is performed, consisting of adding a Gaussian random variable to the calculation steps for path loss to account for random, small-scale power fluctuations.
• 2. The used cell model 9 towards nonperiodic asymmetric cell patterns. Cell ₁ would then not be surrounded by 6 symmetric cells, but by 6 cells with their own random asymmetric demand. In this case, guard bands of different widths must be chosen along the cell boundaries to different neighbors within a cell. Above all, this changes the complexity of the optimization because of the higher dimensionality of the variable space, but not the basic procedures and relationships shown here.

Claims (9)

A method of allocating time slots to uplink and downlink, and assigning users to time slots for throughput optimization in a cellular network consisting of a plurality of cells, each comprising a base station and mobile stations communicating therewith, with different asymmetric multimedia traffic and time multiplexing characterized in that the assignments are determined by a model-based controller comprising the steps of: • calculating a risk indicator ranging between 0 and 1 for each user as a measure of the risk of a user being carrier-to-interference Power ratio below selectable first thresholds, or that a user causes a carrier-to-interference power ratio below selectable first thresholds in at least one other user; Division of users into critical users and non-critical users, where a critical user has a risk indicator above a selectable second threshold, while an uncritical user has a risk indicator below the selectable second threshold; • Parametrization of guard bands, where a guard band of a cell designates the (possibly multiple connected) cell area for which the risk indicator is critical, i. h, above the selectable first thresholds. • Combining the time slots of each one transmission direction (uplink or downlink) into one block, so that there is a single switching point for a change of the transmission direction per communication frame and for each transmission direction "safe" time slots are present, namely those in which the part of the cell for which a control action is to be optimized, working with the nearest neighbor in synchronous transmission direction; Assignment of critical users (users within the guard bands) to secure time slots; • Determination of optimal operating points on a system surface, spanned in particular by the controllable variables uplink / downlink (UL / DL) switching points, vector of user numbers in uplink and downlink in each cell and guard band parameters, comprising the following steps - Modeling the User as a continuous spatial density function; For each selection of a UL / DL switch point and a selectable subset of user number vectors: calculating a carrier-to-interference power ratio at each point in space by integrating all path loss-corrected contributions from the neighboring cells; Determination for each cell of the proportion of "dissatisfied" users where the carrier-to-interference power ratio is below the selectable first thresholds; - Joint optimization of the parameters of all guard bands to minimize the overall content of dissatisfied users in the network; - for each choice of UL / DL switching point: - determination of the maximum number of users that can be served using the optimum guard band with a selectable proportion of dissatisfied users; - Determining the maximum network throughput achievable with this maximum usable number of users for this switching point selection for a traffic demand with a given uplink / downlink asymmetry; Determination of the optimal switching points for each cell and the global maximum network throughput for a traffic demand with a given uplink / downlink asymmetry. Method according to claim 1, characterized in that that the process step of modeling the users as continuous spatial Density function by a step of determining the location a user with a selectable Accuracy is replaced. Method according to claim 1 or 2, characterized that the cellular networks meet the specifications of one or more Standards, chosen from the group consisting of UMTS-TDD, TD-SCDMA and UTRA-TDD. Method according to one of the preceding claims, characterized characterized in that the risk indicator is calculated by adding for each cell a protective band surrounding the cell boundary is introduced, within which the risk indicator is 1 while he outside whose 0 is. Method according to claim 4, characterized in that that for using a cell along the cell boundary to different neighbor protective bands a different width can be selected. Method according to claim 4 or 5, characterized that a user within the protective tape rated as critical will, while a user outside of the protective tape is rated as uncritical. Method according to one of the preceding claims, characterized characterized in that the selectable first Thresholds, a threshold of carrier-to-interference power ratio at a mobile station and a carrier-to-interference power ratio threshold at a base station. Method according to one of the preceding claims, characterized characterized in that the system surface spanning controllable Variables continue to have one or more parameters selected from the Group consisting of a threshold value of the carrier-to-interference power ratio at a mobile station, a threshold of carrier-to-interference power ratio at a base station, a level ratio of a target input power at a base station to a target input power of a mobile station and a radius around a base station within which a power control takes place. Method according to one of the preceding claims, characterized characterized in that when calculating the path loss-corrected contributions as well as in the calculation of the carrier's performance, a log-normal shadowing is performed to model a small-scale fading, consisting of an addition a Gaussian random variable to the calculation steps for the path loss to random power fluctuations on a small scale to take into account.