Method and system for UMTS packet transmission scheduling on uplink channels

[0001] This application claims priority of European Application No. 01303126.5 filed on Apr. 2, 2001.

[0002] The invention relates in general to a method for packet transmission scheduling and to a packet transmission scheduling system and specifically to a UMTS packet transmission scheduling method and to a system having UMTS packet transmission scheduling functionality.

[0003] In particular, the invention addresses the problem of efficient scheduling of data flows, especially in Universal Mobile Telecommunication Systems (UMTS) for a transfer of data between at least one mobile station also termed user equipment (UE) and a so called NodeB of the core Network (CN), such as a local base transceiver station (BTS), i.e. for an UMTS Uplink shared Channel (USCH) between the mobile station and the base transceiver station.

[0004] As known, in packet switching networks the task of multiplexing essentially reduces to the task of ordering packets and to then send them serially over a shared link. This process of serialization is referred to as scheduling. The benefit of packet switching is based on the multiplexing gain, where some data flows benefit from unused resources from other temporarily inactive data flows. The disadvantage of this service compared to a circuit switched service is a lack of predictability in such systems. Significantly, predictability of system behavior is one important measure of quality. Some services, e.g. internet communication or facsimile transmission services need stronger Quality of Service or “QoS” guarantees than others as for example pure speech data transmission. Quality of Service (QoS) scheduling tries to balance the amount and timing of the received service for each flow according to the respective data flow requirements.

[0005] In the downlink data transmission, the radio access network has a complete or perfect knowledge of how much, when, and to whom packet transmissions are made or will have to be made within a certain interval of time, and hence a central controlled manner can be used without any explicit signaling transmission. For the uplink data transmission, however, the radio access network does not have such a complete or perfect knowledge. Thus, due to the lack of knowledge of e.g. transmission traffic volume and synchronization between a plurality of user equipment's (UEs) there is the need for some signaling between the UE and the NodeB, such as the UMTS-radio access network (UTRAN) of a Universal Mobile Telecommunication System (UMTS) to enable the radio network controller (RNC) to schedule the uplink traffic. Based thereon, some traditional uplink transmission schemes employ a random access scheme, e.g. according to the 3GPP UMTS standards. However, it would be desirable to use a controlled unit for uplink data transmission similar to a central controlled unit for downlink data transmission.

[0006] It is an object of the invention to provide an improved method for packet transmission scheduling, especially on uplink channels and an improved packet transmission scheduling system, both the improved method and the system especially adapted to be used for UMTS systems.

[0007] The inventive solution is characterized by a method, a system, a base and mobile station and in implementation software as set out in the claims.

[0008] Preferred refinements are the subject matters of the dependent claims.

[0009] Accordingly, the invention proposes and uses a QoS-scheduling for handling multiple data flows in a Code Division Multiple Access system by dynamically scheduling protocol data units in dependence of allocated radio resource constraints, using transport block sets for determined transport blocks to be transmitted by the physical layer by using respective assigned transport formats, resulting in the ensuring of the required data rates due to a rate conserving scheduling since an optimization of the usage of radio resources within at least two dimensions, i.e. with regard to a respective cell and to its adjacent cells is enabled.

[0010] The inventive QoS-scheduling preferably is apt to handle the data flows on uplink channels and can mainly be applied to scheduling of multiple data flows for different users on dedicated channels or on a shared channel but can also be applied to in the downlink direction.

[0011] According to preferred embodiments, the invention relies on two schedulers, which are linked together in a novel manner, whereby the first scheduler provides a certain degree of predictable behavior and the second scheduler provides Medium Access Control (MAC access) and also allows for bandwidth conserving segmentation and allocation strategies. Preferably these two schedulers are located centrally within the core network and each associated user equipment is incorporating a further scheduler executing decisions of the second scheduler.

[0012] The two central schedulers are named PDU scheduler and MAC scheduler. For the basics of this scheduling method reference is made to the co-pending European Patent Application 00 310 344.7, “Method of linking two schedulers of a multiplayer network and a network comprising a transceiver having linking functionality for two schedulers”. A first adaptation, especially for the UMTS downlink is proposed in the co-pending European Patent Application 00 310 343.9, “Method and System for UMTS Packet Transmission Scheduling on Shared Downlink Channels”. Since the present invention proposes a further improved scheduling method, especially adapted to the UMTS mobile communication system uplink direction, it may be seen as a complement to the downlink using an improved adaptation of the basic scheduling method. Consequently, the contents of both co-pending European Patent Application 00 310 344.7 and 00 310 343.9 are entirely incorporated to the disclosure of the present application by reference.

[0022] For a better understanding of the invention and especially to ensure the improved performance of the inventive method and devices, certain requirements should be met and a number of assumptions are made in advance.

Assumptions and Requirements

[0023] A certain amount α_schedule of the overall uplink resource is allocated by the Radio Resource Management unit (RRM) to the scheduler radio bearer. The scheduler is apt to use α_schedule autonomously without invocation of the radio resource management unit (RRM).

[0024] If automatic repeat request (ARQ) is applied, the number of retransmissions is assumed to be significantly smaller than the regular traffic.

[0025] All transmissions having certain assigned quality requirements are embedded into a data flow. Accordingly, a data flow is defined as a sequence of data packets from the same source to the same destination in the network, for which the user has certain Quality of Service (QoS) requirements.

[0026] Each radio bearer is related to a single data flow. Because multiple radio bearers might be established for a single user, multiple data flows could exist simultaneously that are related to a single user, too. In the following description all data flows are handled separately.

[0027] Throughout this description the elements of a data flow are defined as Protocol Data Units (PDU or PDUs).

[0028] These PDUs are typically layer 3 elements from an UMTS point of view, but the invention is not necessarily limited to this. Protocol data units (PDUs) are segmented into transport blocks (TB or TB's), which receive their own headers, as specified in the UMTS 3GPP standards. This operation is associated with layer-2. Typically but no necessarily the transport blocks have a fixed size. An arbitrary number of transport blocks can be put together to one Transport Block Set (TBS). Typically but not necessarily only transport blocks of one protocol data unit (PDU) are put together.

[0029] One TBS is scheduled by the medium access control layer within the scheduling interval (typically 10 ms) to the physical layer (PHY-layer) per flow.

[0030] For the uplink channels controlled by the scheduler there are no soft handover (HO) assumed. Therefore, the scheduler handles the data flows addressed to user equipments (UEs) of the own cell, only.

[0031] Any mobility related procedures, e.g. hard handover are handled by the radio resource management system (RRM) independently.

[0032] The Bit Error Rate (BER) of a data flow is a static QoS requirement of the associated radio bearer. Depending on the delay constraints, there is a trade-off between Forward Error Correction (FEC), i.e. the received signal energy to noise ratio E_b/N₀ vs. Automatic Repeat Request (ARC) methods, i.e. the allowed number of retransmissions.

[0033] It is assumed that the required bit error rate always can be received from the core network or radio access network.

[0034] In order to optimize the bandwidth consumption of a data flow padding is minimized as a trade-off for delay where possible. This is indicated by the flow's QoS constraints and the recent flow status.

[0035] In order to optimize the delay, the whole protocol data unit (PDU) is taken by the PDU scheduler preferably at once.

[0036] The uplink channels preferably are time synchronized, i.e. every data flow starts its transmission in order that they are received at the same point of time. Thus, timing advance can be used which will be signaled to the UEs by higher layer before the first transmission starts. For moving UEs the timing advance has to be updated regularly. Generally, a discontinuous transmission (DTX) is not used on the uplink channels.

[0037] There is no physical multiplexing (or PHY MuX) for different data flows in the scheduled uplink channels. As a consequence thereof, the transport format combination sets (TFCS) on these transport channels consist of a transport format set (TFS) for one data flow, only. The transport format set is associated to the data rates R_B of the respective data flow. The transport format sets are directly related to the spreading factor SF of the Code Division Multiple Access (CDMA) transmission system which is used to support that data rate.

[0038] The transport block size within one scheduling interval remains constant for each protocol data unit (PDU). Consequently, only the number of transport blocks needs to be counted for medium access control (MAC) scheduling.

[0039] Subsequently, the Radio Resource Allocation (RRA) according to the invention is described.

Basics on RRA for Uplink Channels

[0040] Due to its nature and in view of using a Code Division Multiple Access (CDMA) method, the main resource in the UMTS mobile communication system is the transmission power, which has to be spent for a certain user. The transmission power P_tri of data flow #I is expressed as

[0041] $\begin{array}{cc}{P}_{\mathrm{tri}}\approx {\left(\frac{E_b}{N_0}\right)}_i·\frac{R_{\mathrm{Bi}}}{W}·\frac{I_0}{h_i}=K·R_{\mathrm{Bi}}·C_i,& \left(1\right)\end{array}$
where

[0042] (E_B/N₀)_i denotes the to be received signal energy to noise ratio (E_b/N₀) for data flow #i,

[0043] R_Bi denotes the current data rate used by flow #i,

[0044] W is the chip rate, which chip rate at the moment is defined for UMTS as W=3.84 Mchip/s,

[0045] I_o denotes the interference at the NodeB of the cell where the user equipment (UE) is locatd and may include an intercell interference and thermal noise, and

[0046] h_i is the path-loss between the user equipment (UE) and Node B with H_i≦1.

[0047] However, the resources in CDMA systems depend not only on the data rate as handled by conventional scheduling methods or prior art schedulers but also on factors K and C which depend on several other parameters, like path-losses and interference. From equation 1 the factor K is given as

[0048] $\begin{array}{cc}K=\frac{E_b}{N_0}·\frac{I_0}{W}.& \left(2\right)\end{array}$

[0049] The chiprate W usually is always known by the respective network used. The signal to noise ratio E_b/N₀ and the interference I₀, however, is preferably measured periodically in the NodeB and reported to the ratio network control unit (RNC) where the scheduling mechanism is located. According to equation 1, the C_i-value is the inverse of the path-loss h_i, i.e.

[0050] $\begin{array}{cc}{C}_i=\frac{1}{h_i}.& \left(3\right)\end{array}$

[0051] The C_i-value is an essential part of the scheduling algorithm and is depending on the distance between the UE and the NodeB. If the UE is located nearby the NodeB, the C_i-value is nearly one. If the UE, however, is far away from the NodeB, then the C_i increases to values much greater than one.

[0052] Preferably, the calculation of the C_i-value is carried out by use of one of the following three methods.

• • [0053] According to equation 3, the C_i-value can be directly obtained from the uplink path-loss. In this case, however, the UE has to report sometimes the estimated path-loss;
  • [0054] If a downlink channel exists, then the downlink path-loss might be used. By utilizing the associated downlink transmission power, which can be measured at the NodeB, the downlink path-loss may be obtained. However, to attain the C_i-value, an additional formula similar to the equation 1 has to be evaluated;
  • [0055] When location services are supported by the network, the path-loss may also be obtained from the knowledge of the UE position within the cell.

[0056] Moreover, there are two main restrictions for an uplink scheduling. The first one is based on the target to hold the interference of the respective own cell within a certain limit. Using a received signal power P_reci=P_tri h_i and equation 1, the first restriction may be obtained by

[0057] $\begin{array}{cc}\sum _{i\in \mathrm{active}}{P}_{\mathrm{reci}}=K·\sum _{i\in \mathrm{active}}{R}_{\mathrm{Bi}}\le P_{\max }\left(\mathrm{intra}\right).& \left(4\right)\end{array}$

[0058] The second restriction is based on the target to hold the impact of transmissions from the respective own cell towards adjacent cells within a certain limit. Accordingly, this restriction may be obtained by

[0059] $\begin{array}{cc}\sum _{i\in \mathrm{active}}{P}_{\mathrm{tri}}=K·\sum _{i\in \mathrm{active}}{R}_{\mathrm{Bi}}·C_i\le P_{\max }\left(\mathrm{inter}\right).& \left(5\right)\end{array}$

[0060] Consequently, based on these two main constraints, in the uplink there is an optimization problem not only in one dimension as compared to a downlink optimization but within two dimensions.

[0061] Moreover, due to limited capacities of e.g. a single channel transmission unit there is a further restriction of the transmission power of a single UE, too. This third restriction can be given by
P_tri=K·R_Bi·C_i≦P_max(UE), ∀i∈active.  (6)
Tasks and Functions of the Improved RRA

[0062] FIG. 1 shows a message flow between a Core Network CN, a Radio Network Controller RNC and User Equipment UE when a new radio bearer RB is added to the scheduling function.

[0063] Although it is an essential part of the exemplar UMTS mobile communication system, the base transceiver station (BTS) is not separately shown as these stations are well known to persons skilled in the art. According to FIG. 1, the following tasks and functions should be performed by the radio resource allocation units (RRA) before the scheduling function or operation starts:

[0064] 1. RB Establishment Request: During this phase a new radio bearer RB establishment is requested from the Core Network CN. This request must contain or specify the quality of service (QoS) requirements of the associated flow, i.e. the requested bit error rates (BER), data rates to be transmitted and delay requirements. If there is no radio resource connection (RRC) established, a radio resource control connection establishment procedure between the radio network controller RNC and the user equipment UE has to be performed which is shown as step 1a in FIG. 1.

[0065] 5. Start Dynamic Scheduling: After a successful establishment and initialization the new data flow is added to the scheduling function. The scheduling function will be performed now also for this flow. Reference is made to FIG. 1 showing the message flow for adding a radio bearer to the scheduler.

[0066] 3. Dynamic Channel Allocation DCA: The Dynamic Channel Allocation procedure DCA allocates the following transmission parameters to the data flow (non exclusive): transport format set (TFS), radio link control (RLC) Info, new channelization code, initial transmission power etc. For the allocation method for transport format set and channelization code see also section “Allocation of Data Rates” of this description. New limits for the transmission powers P_max (intra) and P_max (inter) for the scheduler can also be allocated by DCA.

[0067] 4. Radio Bearer Setup: This function performs a setup of the Radio Bearer and synchronization between the RNC and UE. Additionally, the base transceiver station will be initialized with the parameters that are allocated by DCA which is not shown in FIG. 1 but known to a person skilled in the art.

[0068] 5. Start Dynamic Scheduling: After a successful establishment and initialization the new data flow is added to the scheduling function. The scheduling function will be performed now also for this flow. Reference is made to FIG. 1 showing the message flow for _adding a radio bearer to the scheduler.

Allocation of Data Rates

[0069] The allocation of the data rates for each data flow has a strong impact on the system efficiency that the scheduler can achieve. The data rates relate to TFS and the channelization codes according to the assumption, that there is no physical multiplexing (or PHY MuX) for different data flows in the scheduled uplink channels. As a consequence thereof, the transport format combination sets (TFCS) on these transport channels consists of a transport format set (TFS) for one data flow, only. The transport format set is associated to the data rates R_B of the respective data flow. The transport format sets are directly related to the spreading factor SF of the Code Division Multiple Access (CDMA) transmission system which is used to support that data rate. Thus, for a rough allocation or estimation of the limit data rates the following rules are applied:

[0070] For the maximum data rate R_Bmax transport format sets (TFS) should be allocated to allow data rates up to two to four times of the maximum data rate, i.e. (2 to 4) R_max. There are two reasons for this requirement. The first is that these maximum transport format sets are required from the MAC-scheduler to serve a flow with a temporarily higher data rate than requested to let a flow profit from other flows being idle. This applies when there is remaining capacity on the air-link and if this flow has already data available to send in its transmission queue of its specific rate.

[0071] The second reason is to allow for a time divisional multiplex style multiplexing on the transport block level. The MAC-scheduler's algorithm can be developed towards bandwidth efficiency. Therefore it also may be desirable to use arbitrary sizes of transport blocks to minimize padding. This means the available transport format should be able to temporarily exceed specified rates.

[0072] Fairness, bandwidth and quality of service (BW-QoS) guarantees among flows are maintained by another scheduler, the PDU sceduler.

[0073] For the minimum data rates R_Bmin transport format sets should be allocated to allow data rates below R_Bmin. The availability of smaller transport format sets allows the MAC-scheduler to minimize padding. As this introduces additional delay and lowers the average transmission rate, it is only applicable to certain QoS flows. The optimization of such allocation and transport format combination (TFC) usage is addressed separately.

[0074] In order to comply with the assumption that the uplink channels preferably are time synchronized, i.e. every data flow starts its transmission in order that they are received at the same point of time and a timing advance can be used which will be signaled to the UEs by higher layer before the first transmission starts, only those transport format sets are allowed which would fill up the whole data frame with data. Because the spreading factors of the channelization codes are in the order of SF=2^k, k=2, 3, . . . , according to the above-mentioned assumption, that there is no physical multiplexing, this leads to data rates of R_B=R′_B·2ⁿ, n=0, 1, . . . , where R′_B represents a reference data rate for a certain given spreading factor and may become R_Bmin.

Allocation of Transport Format Set (TFS)

[0075] The transport format set (TFS) is defined as the set of transport formats (TF) that are associated to one data flow. The semi-static part (coding, transmission interval, rate matching) essentially determines the bit error rate. It is defined by the radio resource management. In the following discussion, the focus is only on the dynamic part which consists of the transport block size and the transport block set size. The dynamic part of transport format sets can be used for optimization of the segmentation in the radio link control (RLC). For the choice of this dynamic part there is a trade-off between the granularity of the data rates and the limited size of transport format sets. On the one hand, each data flow intends to have a high granularity in data rates avoiding extensive padding. This would lead to a large transport format set size. On the other hand, a transport format set is used to enable efficient physical or PHY signaling for changing data rates. Due to limitations of that PHY signaling (e.g. TFCI (Transport Format Combination Indicator) encoding) the maximum transport format set size is quite limited. Therefore the following transport format set allocation rules regarding the characteristic of the data flow, similar to the downlink, are proposed and used according to the invention.

[0076] 1. Real Time (RT) Services: This service type needs immediately serving of the offered data. Hence, a high granularity towards higher data rates is desired. Therefore, for real time services a larger transport format set should be allocated.

[0077] 2. Non Real Time (NRT) Delay Sensitive Services: Here, a limited automatic repeat request (ARQ) can be used for protection of the data flow. The granularity is not as high as for the pure RT service because some data can be queued for a limited time. Therefore a limited transport format set can be allocated for such services. For efficient use of automatic repeat request mechanisms the transport block size should be small.

[0078] 3. NRT Unconstrained Delay Services: This type of service is the best candidate for bandwidth optimization. In principle unlimited queuing is possible. Hence, no much granularity is necessary. Therefore a quite limited transport format set can be allocated for this service type. Granularity is used only to avoid padding, only.

[0079] Beyond the delay constraint it is sensible to take further QoS requirements and flow specifications into account. The system may potentially adapt to certain preferred PDU sizes, like that of a transport control protocol acknowledgement (TCP-ACK). Bulk data transfer could guard the selection towards max. PDU size, etc.

UL Channelization Codes

[0080] In the Uplink (UL), each UE preferably gets a unique scrambling code. Since the whole code tree can be used by one UE, a special management of the code tree is not necessary as compared to the Downlink. However, since some codes should be reserved e.g. for UL physical control channel, two alternatives are proposed for allocating the UL channelization code:

[0081] If pre-defined channelization codes are used, e.g. by following a fixed rule, then an allocation and signaling of the UL channelization codes are not necessary. The actual code sequence depends on the currently used data rate. However, the allocation has to be known by both entities, such as the UE and the UTRAN (UMTS Radio Access Network).

[0082] If the channelization codes are determined by the UTRAN, that is, the UTRAN allocates a certain code branch based on different data rates for the specific UE, then specific capabilities of some UEs might be considered. If each UE has its own code tree, the allocation rule is easier in comparison to the Downlink. However, the code branch has to be signaled to the UE.

Control of the Uplink Data Transmission

[0083] Due to the lack of knowledge of e.g. transmission traffic volume and synchronization between the UEs, some signaling must be applied between the UE and a UTRAN for enabling the RNC to schedule the uplink traffic.

[0084] Reference is made to FIG. 2 showing the signaling flow between the UTRAN and a UE during the operation of uplink dynamic scheduling function. This function is similar to a distributed queuing request update multiple access (DQRUMA) as described for example by M. J. Karol et al., “Distributed-queuing request update multiple access (DQRUMA) for wireless packet (ATM) networks” in Proc. ICC'95, pp. 1224–1231, Seattle, Wash., the contents of which is incorporated to the present disclosure by reference. Mainly, the details depend on which transport channels will be used for this signaling. According to FIG. 2 the following signals and messages should preferably be identified.

[0085] 1. PDU request: If the UE has a PDU to send, it will be stored in its own queue. The UE requests for the new PDU which is sent e.g. over a random access channel (RACH), when no other transport channel is active or over another currently active transport channel used by the UE. This request may contain the size of the PDU. Based on the request, the PDU scheduler determines the order of the PDU list, as described below under the section “PDU-scheduler”.

[0086] 2. TF assignment: The MAC-scheduler decides when and how many transport blocks (TB's) could be sent during a scheduling interval. The MAC-scheduler then informs the UE thereabouts by use of a transport format (TF) assignment message. This message is sent over some downlink channel, which can be an associated dedicated chennel (DCH) or a downlink shared channel (DSCH), too. The message may be piggybacked over user data.

[0087] 3. Transmission of first TB's appropriate to TF: The UE sends the TB's to the UTRAN using the TF that was assigned. Therefore the PDU data in the UE, e.g. a mobile station MS is segmented to the TB's, ARQ is performed where applicable and the TBS is created for the given TF. A MAC-scheduler of the UE (UE_MAC) delivers the TB's to the PHY-Layer of the UE (UE_PHY) for sending and a pointer tot he next part of the PDU within the queue is updated within the UE.

[0088] 4. TF assignment: After the scheduling interval the MAC-scheduler again decides on the transmission of the next TB's. If ARQ was applied, the new TF assignment message contains an acknowledgement for the previous sent TB's or in case of error the sequence number of erroneous TB's. When the scheduler decides that no TB's from the specific UE should be sent within the next scheduling interval, a TF assignment is not sent.

[0089] 5. Transmission of next TB's appropriate to TF: The UE sends the TB's to the UTRAN as described in step 3. Erroneous acknowledged TB's will be inserted. If the network assigned no TFS, the UE does not send any data.

[0090] 6. PDU request: There might be a new PDU request within the current MAC-scheduling flow. This should be handled independently on the ongoing transmission as described in step 3. Alternatively, the new PDU request can also be piggybacked to UL traffic.

[0091] 7. Transmission of last TB's appropriate to TF: The signaling sequence of assignment and transmission will be continued until the last TB's will be transported. The UE indicates the end of the PDU, e.g. by appending a tag at the end of the data. After the UTRAN has sent the last acknowledgement, which is not explicitly shown in FIG. 2, the PDU can be deleted from the UE queue and the PDU list.

[0092] Subsequently the Scheduling Method according to the invention is described.

[0093] The invention contemplates the use of two schedulers, which are linked together to achieve a certain degree of predictable behavior, while also allowing for bandwidth conserving segmentation and scheduling, see also the referenced European Patent Application 00 310 344.7. These two schedulers are named PDU scheduler and MAC-scheduler, of which both are preferably located centrally in the RNC. Then, a third scheduler, i.e. a UE_MAC is incorporated in each user equipment UE for executing the decisions of the central MAC-scheduler.

[0094] The PDU-scheduler operates on the input data from Layer 3, the Protocol Data Units (PDU). It receives the QoS requirements of each flows and determines the order in which PDUs should receive service based on the availability of schedulable PDUs in a respective UE notified to the PDU-scheduler.

[0095] The MAC-scheduler serves the PDUs from this list and tries to reflect the order in the list, while also taking timing and power constraints into account.

[0096] In FIG. 3 the principle architecture of the two serial schedulers is shown. The MAC-scheduler is active at every frame, e.g. on a 10 ms base. The PDU scheduler is operated on all active flows, i.e. with a non-empty PDU flow-queue. To avoid undesirable behavior of serially uncoupled schedulers, both schedulers are linked together by means that the MAC-scheduler is driven by the state of the PDU-scheduler.

[0097] According to FIG. 3 depicting the logical view of a RNC scheduling system in the uplink, there are virtual PDU flows. When a UE, such as a mobile station MS, signals a new PDU to be transmitted the PDU-scheduler is served by the MAC-scheduler, i.e. the MAC-scheduler bases the scheduling decisions on the PDU-scheduler's outcome. The MAC-decisions are then signaled to the lIE to drive the UE_MAC-schedulers operation. This operation involves the actual processing of user data, where the PDU in the UE is segmented, ARQ is performed, a TF is filled with TB and ultimately transmitted by the PHY-layer, as depicted in FIG. 4.

Principles of the PDU Scheduler

[0098] For the downlink, the principle funciton of the PDU-scheduler can be directly implemented as described by the referenced European Patent Application 00 310 343.9. However, in the Uplink the acutal data, i.e. the PDU is not physically present at the PDU-scheduler's place, but only the fact, that a PDU of a given size has to be scheduled is known to the PDU-scheduler. Thus, instead of real PDU flows, virtual PDU flows are processed in the PDU-scheduler and these virtual PDU flows are scheduled regarding their QoS requirements into one common list, denoted as the PDU list, reflecting a desired order of transmission as seen from the PDU-scheduler's side and providing the basis for MAC-scheduling decisions. This list is not termed a queue, because due to MAC-constraints it cannot be assured to serve this queue in a first input first output (FIFO) fashion. However, the PDU scheduler tries to serve the data with respect tot he required QoS parameters for each flows, e.g. data rates.

[0099] For this purpose any rate conserving scheduling policy can be applied, see for example Hui Zhang, “Service Disciplines for Guaranteed Performance Service in Packet-Switching Networks”, Proceedings of the IEEE, Vol. 83, No. 10, October 1995, e.g. Weighted Fair Queuing (WF²Q) or Virtual Clock Queuing (VCQ).

[0100] The scheduling elements for this scheduler are taken depending on the following rule: Typically the PDUs are large enough to be scheduled as one unit. In this case one scheduling element equals one PDU.

[0101] If it is possible to serve several PDUs simultaneously on theMAC-layer it may become desirable to have several PDUs available for MAC-scheduling from one flow. This would mainly be the case where PDUs become too small to be served with the required minimum data rate T_schedule (typically T_schedule=10 ms ), i.e. when
PDU_len/R_Bmin≦T_schedule.  (7)

[0102] The problem can be overcome by having several PDUs from the flow grouped together into one container, which then becomes the scheduling element.

[0103] Hence, the scheduling element can be defined as one container that may consist of one (normally) PDU or several PDUs. Throughout this document one scheduling element is defined as PDU and the term PDU scheduler is used for the sake of simplicity.

[0104] As shown by J. Cobb et al. in “Flow timestamps” of the Annual Joint Conference of Information Sciences, 1995, it is equivalent for the considered scheduling method to work with flow time-stamps instead of PDU time-stamps. When doing so, the inventive PDU scheduler becomes active when a Pdu from a flow is fully served and therefore removed from the PDU list or when a formerly inactive flow gets reactivated by a PDU arrival into its empty PDU flow-queue. This is beneficial as it limits the number of elements in the PDU list tot he number of active flows.

[0105] Moreover, it also justifies that a UE only notifies the first PDU of a sequence of packets via a PDU request and notifies the subsequent PDU and their size piggybacked to the last MAC-transmission of each PDU. With only one PDU request a burst of PDU frames may be transmitted. Accordingly, there is no difference for the PDU-scheduler if all or only the next/current PDU is in the virtual PDU-flow.

Creating the Uplink PDU Scheduler

[0106] As mentioned above, in the Uplink there is the problem that the FIFO queues for the incoming data flows are located in each UE separately whereas the scheduling function is located preferably in the RNC within the network. Thus, the usage of a virtual PDU list is proposed, which preferably is created with reference to FIG. 4 as follows:

[0107] 1. If a PDU is generated, the UE sends a “PDU request message” to the RNC (comparing also FIG. 2).

[0108] 2. The RNC then (re)determines the order of all PDUs within the virtual list. This list may then be used by the MAC-scheduler as described subsequently in the sections relating to the MAC-Scheduler.

[0109] The place of each reported PDU is determined according to the data rate that was requested for the associated data flow and the PDU length. According to a very preferred refinement to easily get the length, the PDU length is explicit signaled e.g. via piggybacking. As an alternative, the PDU length may be estimated in the RNC without the knowledge of the exact length, preferably according to the following way.

[0110] The length on the first PDU is taken from a predefined estimate. For example, the length of a usual IP-packet for the requested service could be taken;

[0111] The length of the following PDUs is estimated from the length of the previous PDUs which are already known in the RNC when receiving an end of PDU tag from the UE. According to a preferred method the length of the directly preceding PDU may be taken. Of course, even a kind of average of more PDUs may be used for determining the current PDU length estimate;

[0112] As described in the section “Control of the Uplink Data Transmission” the PDU is not deleted from the list until the acknowledgement for the last TB of that PDU was sent.

Main Function of the MAC Scheduler

[0113] The MAC-scheduler serves the PDUs from the PDU scheduler. The order in the PDU scheduler's list signals the priority in which the PDU scheduler wants the PDU's to be served. Similar to the MAC-scheduler for the Downlink the MAC scheduler for the Uplink tries to achieve this while obeying at least four restricitons:

• • [0114] Bandwidth constraint due to TFC allocation of the flow and availability of a node in the code tree (refers to CBA);
  • [0115] Delay constraints drive the decision how many subsequent TBS transmissions spread over several timing intervals are tolerable to obey timing requirements of the served PDU;
  • [0116] ARQ-constraints, in that a transmission of TB's that receive ARQ service is only possible until the ARQ window size is reached and further transmissions are possible only after the ARQ stage receives the acknowledgement from the receiver;
  • [0117] Power constraints limit both the power for the transmission to one individual mobile and the overall power in the cell. To avoid RRM regulation on these issues, the scheduler should itself take this into account.

[0118] This present proposal mainly consists of a framework that allows for several MAC-scheduling algorithms to obey these constraints without having to explicitly worry for the flow's QoS requirements anymore, as this has been appied by the PDU scheduler already. Hence, in the following an algorithm is used complying with these constraints in a straightforward manner. Later some improvements are shown.

[0119] According to FIG. 5, the basic mechanism for MAC scheduling essentially is the following:

[0120] 1.) Set the queuing pointer in front of the PDU list, i.e. set pointer=0. Reset the current powers P_Current(intra)=0 and P_current(inter)=0.

[0121] 2.) Take the next PDU from the PDU list and consider as much of it in means of TB for scheduling as it is constrained by:

• • [0122] PDU_size/segment size→result max #1 of TB's,
  • [0123] ARQ-constraint→result max #2 of TB's,
  • [0124] TFC-constraint→result max #3 of TB's,
  • [0125] Maximum UE transmission power P_max(UE): R_bimax=P_max(UE)/(K·C_i) (cf. Equation 6), where K and C_i are currently given by equation 2 or equation 3 →result max #4 of TB's,
  • [0126] Intracell interference limit P_limit(inter): Calculate the virtual available data rate R_bi1 (available)=(P_limit(intra)−P_current(intra))/K (cf. equation 4)→result max #5 of TB's,
  • [0127] Intercell interference limit P_limit(inter): Calculate the virtual available data rate R_bi2(available)=(P_limit(inter)−P_current(inter))/(K C_i) (cf equation 5)→result max #6 of TB's.

[0128] 3.) Formal: Result #TB_max=min (max #1 of TB's . . . max #6 of TB's). Perform capacity optimizing decisionson TBS creation. It can be advisable to schedule less TB's, with a smaller TBS than possible from the constraints in step 2. If no optimization is desired, chose max # of TB's from step 2.→new variable #Tb_schedule.

[0129] 4.) Create TF which is associated to #TB_schedule. Set R_Bi used regarding the created #TB_schedule.

[0130] 5.) Store the associated TF for signaling to the UEs in step 8.

[0131] 6.) Compute the total intra cell power by P_new(intra)=P_current(intra)+K·R_Bi (used). Compute the total inter cell power by P_new(inter)=P_current(inter)+K·C_i·R_Bi (used). Compare these values against the power limits P_limit(intra) and P_limit(inter).

[0132] 7.) If total Power check is ok, i.e. if

• • [0133] P_limit(intra)−P_new(intra)≧P_min(intra) and
  • [0134] P_limit(inter)−P_new(inter)≧P_min(inter) (P_min(intra,inter): minimum power for a certain # of TB's regarding intracell and intercell) and there are more PDUs in the PDU list, increase P_cell by one to the next PDU in the PDU list, set P_current(intra)=P_new(intra) and P_current(inter)=P_new(inter) and go to step 2.

[0135] 8.) Signal the stored TF to the UEs by using “TFS assignment” message as described in the section “Control of the uplink data transmission”. This message is sent to all UEs having stored TF simultaneously.

Handling of the power limits P_limit (intra,inter)

[0136] This section describes how the interference limits P_limit (intra,inter) for the cell are allocated for MAC-scheduling. The limits P_limit (intra,inter) for the scheduler should preferably be chosen according to the following rule:
P_limit(intra,inter)=min{P_max(intra,inter), P_current(intra,inter)+ΔP_inc(intra,inter)},  (8)
where ΔP_inc(intra) and ΔP_inc(inter) are certain increases of intra and inter Power, respectively.

[0137] The first term in equation 8 prevents the scheduler to use resources larger than the by RRM assigned P_max(intra) and P_max(inter). The second term guarantees that the increase in the current power P_current(intra) and P_current(inter) is below a given limit ΔP_inc(intra) and ΔP_inc(inter). The intracell limitation is useful in order that the UL power control for all users on other TrCH which are not handled by the scheduler (e.g. users on DCH) can follow the increase of transmission power. The intercell limitation protects the users from adjacent cells. For obtaining the current powers P_current(intra) and P_current(inter) the two following possibilities are preferred.

• • [0138] The values of P_current(intra) and P_current(inter) are directly taken as a result of the MAC-scheduler procedure, or
  • [0139] from time to time the UEs may send a measurement report of its currently transmission power to the UTRAN. From these values P_current(intra) and P_current(inter) can also be estimated.

[0140] The basic limit definition according to equation 8 can be enhanced as following. For the purpose of efficient of the allocated radio recourses the MAC-Scheduler should monitor the goodput, i.e. the throughput R_actual of the scheduler without retransmissions, which simply can be defined by

[0141] $\begin{array}{cc}{R}_{\mathrm{actual}}=\sum _{i\in \mathrm{actual}}{R}_{\mathrm{Bi}}.& \left(9\right)\end{array}$

[0142] The virtual bandwidth is defined by the overall available data rate R_overall, which can be allocated by the MAC-Scheduler. This virtual bandwidth depends on the allocated powers P_max(intra) and P_max(inter) for the scheduler:
P_overall=function(P_max(intra,inter))≈min{P_max(intra), P_max(inter)/C′}/K.  (10)

[0143] The value of C′ represents a kind of estimate from the constants C_i from all data flows. The goodput R_actual is now compared with the virtual bandwidth R_overall. Depending on the comparison result the following actions can be taken:

• • [0144] If R_actual<R_overall, then there is a scheduling problem. The scheduler can process less data than required. In this case RRM has to be informed to take actions. This could involve to allocate larger resource P_max(intra) and P_max(inter) to the scheduler if available. If not, a dynamical resource reallocation to flows has to be performed. This could for example mean to drop or stop certain flows, that formerly were served with QoS BW guarantee. Finally this feedback can be used to alter capacity estimates for future Admission Control decisions.
  • [0145] If R_actual≈R_overall, then the scheduler works efficient and within the limits. In this case equation 8 will be used as scheduling policy.
  • [0146] If R_actual>>R_overall, then the scheduler works in a relaxed manner. That means it is able to schedule much more data than actual required. In this case the scheduler can have a self limiting behavior depending on the history of the goodput R_actual in the following way:
  • [0147] If R_actual(t)≦R_actual(t−1), then use the following modification of equation 8:
    P_limit(intra,inter)=P_current(intra,inter)−ΔP_dec(intra,inter),  (11)
  • [0148] where ΔP_dec(intra) and ΔP_dec(inter) are certain decreases f intra and inter Power, respectively.
  • [0149] if R_actual(t)>R_actual(t−1), then use equatin 8 as it is.

[0150] This allows equalization of the total traffic in terms of traffic shaping. In order to keep the allocated resources available for the scheduling process RRM will not be informed about this self limiting. Nonetheless there will be a noticeable lower variation f power consumption which is beneficial for neighboring cells and DCH power control fluctuation in this cell.

[0151] FIG. 6 illustrates an example of handling P_limit(intra,inter) in the P_current(intra,inter) plane. As mentioned above, the limits P_limit(intra,inter) can vary dynamically. For each dimension there are the following three possibilities:

• • [0152] If there is an increase of P_current(intra,inter)>P_limit(intra,inter), then P_limit(intra,inter) is increased by ΔP_inc(intra,inter). The upper limits are given by P_max(intra,inter) that are allocated by RRM. This is the representation of equation 8.
  • [0153] If the interference decreases to P_current(intra,inter)<<P_limit(intra,inter), then P_limit(intra,inter) can be decreased by ΔP_dec(intra,inter). This is the representation of equation 11.
  • [0154] If P_current(intra,inter)≈P_limit(intra,inter), then P_limit(intra,inter) does not change.

[0155] These variations are done in both dimensions, and preferably separately. Thus there is an area of possible variations around P_limit(intra,inter) that is marked by a shaded area in FIG. 6. Since all of these parameters can be adjusted independently, the scheduler offers a great flexibility.

Improvement of MAC-scheduling Decision

[0156] According to a preferred refinement, an improvement can be made that is similar to the downlink as described by the referenced European Patent Application 00 310 343.9. Especially for NRT services it is not necessary to always attempt to pack the whole PDU into one TBS to be scheduled in one MAC-scheduling interval. It might be desirable to spread the transmission in time over several scheduling intervals. Hence, the invention proposes and uses the following enhancement to section “Main Function of the MAC Scheduler”.

[0157] Fr NRT-services for each PDU the maximum number N_schedule of MAC-scheduling intervals T_schedule that are allowed for an initial PDU transmissin is determined. The term initial means that this value does nt include potential retransmissions. The value N_schedule is determined by:
N_schedule=PDU_len/(R_Bmin·T_schedule).  (12)

[0158] Here, a system is assumed that obeys no other restrictions (e.g. ARQ . . . , see assumption mentioned above that if automatic repeat request (ARQ) is applied, the number of retransmissions is assumed to be significantly smaller than the regular traffic). Given this value for each PDU, the MAC scheduler can schedule less TB at a time. This may be motivated by several reasons. Firstly, the efficiency can be improved by reducing the padding. This is possible if there are several sized TF's defined that allow for the transmission in the next interval using smaller TBs, so less padding is needed. The variation of the created interference is also lowered when the data transmission is spread over several scheduling intervals instead of an on-off source behavior. This phenomenon is now investigated for inter and intra cell impact.

[0159] In contrast to the downlink, in the Uplink the scheduling decision is based on a two-dimensional optimization problem resulting from the limitations for intracell and intercell interference. FIG. 7 illustrates the problem using the P_current(intra,inter) plane, again. From the power control/interference perspective there ideally would be no power variation at all, i.e. the same target point should be reached in every scheduling cycle. For practical applications this means, this point should move in a defined manner.

[0160] For example, looking at two intervals and two UEs equally sized packets is compared. Once they are transferred serially, i.e. the whole packet is sent from the UE which is close to the BTS first and then the whole packet from the UE which is far away from the BTS. This means the P_current(intra) and P_current(inter) is first low and then substantially higher, because more power is needed for the transmission of the far UE data.

[0161] In the alternative scheduling discipline described here, the data is split into half packets and both UEs or MS transmit their division of their data in both intervals. This means the P_current(intra) and P_current(inter) has a medium value for both intervals, i.e. the point in the graph according to FIG. 7 remains the same. Intuitively this is beneficial for PC algorithms. The basic idea is to select TF's in such a way that the summation of P_current(intra) and P_current(inter) are approximated to the value from the last scheduling interval.

[0162] For a more realistic example some traces for the vector of P_current(intra, inter) as a result of MAC scheduling method described in the section “Main Function of the MAC Scheduler” is shown in FIG. 7. Each segment of the trace represents the increase of P_current(intra, inter) after a certain number of TB's for one UE has been scheduled. From the interference point of view the optimum would be at the intersection of the lines for P_limit(intra) and P_limit(inter). From FIG. 7 trace #1 obviously offers not the optimal solution, because the limit of P_limit(inter) is reached while P_current(intra) is far away from its P_limit(intra). However, trace #2 reaches the optimum regarding the interference. Therefore optimizing the scheduler against interference is equivalent to finding a trace within the plane that reaches the optimum point. Since a person skilled in the art is able to find several methods for this optimization, the individual methods are not described in detail.

[0163] However optimizing the scheduling against the interference alone as mentioned above, may contradict the optimization against the QoS. To overcome this problem, a parameter β=(0, . . . ,1) should introduced that indicates for:

• • [0164] β=0 an optimization against interference, only; e.g. trace #2 in FIG. 7;
  • [0165] β=1 an optimization against QoS, only; e.g. trace #1 in FIG. 7; and
  • [0166] β between 0 and 1 an optimization on both having a trade-off between interference and QoS and the trace would be somewhere between trace #1 and trace#2 in FIG. 7.

[0167] The use of such a parameter offers a free choice on the preferred amount of interference optimization with regard to individual flow's QoS requirements. However, it has to be mentioned, that by use of the limitation rules in section “Handling of the power limits P_limit(intra, inter)”, the limits P_limit(intra, inter) are dynamic, i.e., their intersection shifts towards the computed end points of the traces (cf. FIG. 6). In the non-optimal case (regarding interference), however, the limits could be far away from the allocated maximums, i.e., P_limit(intra, inter)<<P_max, (intra, inter).

[0168] This principle is depicted in more detail in FIG. 8. As can be seen therefrom, the current setting of the limits is given by the point P_limit(inter, intra). Based on the aforementioned, the allowed changes of that limit within the next period as described in the section “Handling of the power limits P_limit(intra, inter)” gives a certain area, where the new limit can be moved. This area is depicted as the rectangle in FIG. 8. Optimization against only the interference (β=0) means, to set the end of the trace within this rectangular area. The point P_target(inter, intra) is the final target of the optimization. All other points within the area are sufficient solutions of the optimization problem against interference and sub-optimal for the optimization problem of cell-capacity usage.

[0169] On the other side, optimization only against QoS (β=1) would lead to a trace that could be far away from the rectangular area. Therefore, the goal of combined optimization is to find an optimal β resulting in a trace as close as to the rectangle while also optimizing the QoS constraints of those data flows where required.

[0170] This invention is not restricted to the specific embodiments described above. A person skilled in the art will recognize that based on a rate conserving strategy the scheduler guarantees the required data rates.

[0171] Even if a delay is not explicitly addressed by the scheduling principle, if each of the respective data flows is in compliance with its required quality of service and given admission control the scheduler guarantees that there will be no additional delay due to congestion within the scheduling system.

[0172] Preferably, bit error rate (BER) requirements are guaranteed in addition via properly forward error correction (FEC) and automatic repeat request (ARQ) functions.

[0173] A most preferred application of the improved scheduler is the handling of data flows in the uplink and the uplink scheduling was described in detail in the above description. It may be applied for controlling an uplink shared channel (USCH) and/or for coordinating the transmission on several dedicated transport channels to (DCHs). However, the inventive method of QoS scheduling is not limited to the uplink but also can be applied to downlimk channels.