GB2447431A - Allocating channel bandwidth dependent on predicted variable bit rate - Google Patents

Abstract

An access controller on a wireless communications medium distributes frame information encoded by one of first and second encoding schemes, for example I, P or B frames transmitted as a Group of Pictures (GOP) as part of an MPEG video. To allocate sufficient channel bandwidth to ensure a desired Quality of Service a predicted traffic demand for a forthcoming encoded frame is determined, by retrieving pre-stored prediction configuration information, and using said prediction configuration information to determine said predicted traffic demand, and determining a prediction error for a previous prediction carried out by that frame prediction means and to modify said pre-stored prediction configuration information accordingly. Following this, the predicted traffic demand can be employed in the allocation of frames to the medium, more specifically to transmission opportunities defined in the medium.

Description

The present invention is concerned with wireless communication in a

wireless communication network. The invention has applicability, in particular, but not exclusively, in wireless audio/video products based on the 802.11 (or 802. lie) protocol.

The IEEE 802. lie protocol is a Quality of Service (QoS) extension to the original 802.11 protocol. It is based on the Hybrid Coordination Function (HCF) that contains two medium access mechanisms: HCF Controlled Channel Access (HCCA), and Enhanced Distributed Channel Access (EDCA). This will be familiar to any person skilled in the field of the IEEE 802. lie Standard.

One of the main features of the 802.11 e HCF is TXOP (transmission opportunity), which refers to a time period during which a station is allowed to transmit a burst of data frames. A TXOP is called an EDCA-TXOP when it is obtained by. winning a successful EDCA contention, or an HCCA-TXOP when it is obtained by receiving a Q0S poll frame from the QAP (Q0S-enhanced access point). In order to control delay, the maximum value of a TXOP is bounded by a value called TXOPmit, which is determined by the QAP. A QSTA (Q0S-enhanced station) can transmit multiple frames within its TXOP allocation. This new feature also tends to provide time-based fairness between QSTAs, which can help to remedy the performance anomaly of the legacy 802.11 MAC when different STAs operate at different data rates, and slow STAs may starve fast ones.

Figure 1 shows an example of an 802.11 e beacon interval. During a beacon interval, a QAP is allowed to start several contention-free bursts called controlled access periods (CAPs) using HCCA at any time after detecting a channel as being idle for a time interval of PIFS (PCF inter-frame space). PIFS is defined as being shorter than DIFS and AIFS and, because of this, a QAP is given a greater opportunity to start HCCA than EDCA. HCCA is more flexible than the legacy PCF (point coordination function) because the latter is only allowed in a CFP period, while a QAP can initiate HCCA whenever it wishes during the whole beacon interval. To leave enough space for EDCA, the maximum duration of HCCA in a beacon interval is limited by a variable TCAPUIL. On receiving QoS request frames (containing Traffic Specifications or TSPECs) from QSTAs, a QAP scheduler residing on the QAP first deteimines the selected service interval (Si), which should be the highest sub-multiple value of the beacon interval and a number lower than the minimum value of all the maximum service intervals for all the admitted flows. Then an 802.11 e beacon interval is divided into an integer number of SIs, and QSTAs are polled sequentially during each selected SI.

In the HCF scheduling algorithm specified by the 802.1 le standard, TXOP is allocated per-stream and calculated using a simple formula based on the assumption of constant bit rate (CBR) traffic with a fixed mean data rate: TXOP=ISI*p/L]*L/R o, where SI is the service interval (the interval between successful polls sent to a stream), p is the mean data rate, L is the nominal MSDU (MAC Service Data Unit) size, R is the physical transmission rate, and 0 is the overhead including ACKs, polls and inter-frame spaces. Each station is then polled in a round- robin manner and allocated its corresponding TXOP.

It can be seen that the above 802. lie reference scheduler is simplistic in that it allocates a constant TXOP at every service interval for a stream. However, it has been widely recognized that VBR (variable bit rate) video traffic such as MPEG-4 is highly bursty (i.e. with a high peak-to-mean ratio), highly correlated and would often have heavy tail distributions. An example of such a traffic stream is shown in Figure 2. This shows the traffic profile, with respect to time, of the data to represent a stream of the movie "Jurassic Park". This stream can be seen to be a \BR stream, as identifiable from the wide variations in traffic rates throughout the course of the stream.

If bandwidth were statically allocated according to the mean data rate of the traffic stream, large queues, large delays, and excessive packet loss could arise. In particular, if the time required to send an entire frame is longer than the TXOP, the remaining fragments of the frame have to be transmitted in the next opportunity. This can cause excessive delays. Packet drops may arise if the next opportunity is after the deadline for transmission of the packet, and thus transmission of the next packet needs to commence.

Consequently this can result in deterioration in received video quality. This means that, in order to guarantee a particular Q0S, the number of VBR flows admitted in a network must be restricted to avoid such problems arising. This is achieved by peak rate allocation, whereby allocate channel time is allocated according to the maximum frame rate to tackle the variable rate problem.

However, peak-rate based allocation often results in network under-utilization and bandwidth wastage because fewer VI3R. flows can be admitted than in the case where the mean rates are used. Therefore it is desirable to devise an effective and intelligent approach to the support of multimedia applications with stringent bandwidth and delay requirements. In a practical context, for example in an 802.1 le-based home network with multiple high-bandwidth devices (e.g. HDTV, DVD), it is necessary to have an adaptive bandwidth allocation scheme in place to meet the QoS requirements of all the applications, such as lower job failure rate, lower delay jitter and higher throughput.

As an example, Figure 3 compares the performance characteristics of a VBR MPEG-4 stream and its CBR equivalent. The CBR stream has the same bit rate and mean packet size as the MPEQ-4 trace, but is transmitted via constant MSDIJs. It will be clear to the skilled reader that the substantial delay arising for MPEG video is a direct consequence of its VBR characteristics. V.BR sources are highly variable in both arrival rate and packet size. An accepted stream hence either underutilizes an awarded TXOP, or is allocated a TXOP not sufficiently proportioned to its queue length, resulting in poor performance from the reference scheduler. On the other hand, the CBR video delay remains below 30 ms because CBR sources, with fixed arrival patterns and constant packet sizes, are well catered for by the fixed TXOP scheduler, and resources can be allocated reliably.

Similarly, Figure 4 shows a transmission delay cumulative probability distribution for both CBR and VBR video, where the cause of the high mean delay for VBR video is apparent from the large surge in delay around 300 ms. Indeed, the contrast between CBR and VBR is striking in that CBR can be guaranteed to give rise to a delay below a relatively low maximum, whereas VBR delay can be anything up to 6 or 7 times the maximum delay for a CBR stream.

Various publications have offered teachings on improving on the above situation.

However, no source has provided a ready and evident way forward. For instance, Taiwanese patents TW520589B and TW561748B deal with bandwidth allocation in wireless ATM networks, which transmit information but without many of the constraints discussed above.

International patent application W020040043 10 describes a video compression scheme which is not directly related to wireless networks.

US2005094558 offers a description of a method of predicting traffic load in a WLAN and uses this information for access control, for example in adapting contention windows.

According to an aspect of the invention, therefore, there is provided an access controller for a wireless communications network, the access controller being operable to allocate access to a wireless communications medium for distribution of packet based information to other stations in the network in use, the access controller comprising traffic prediction means operable to process variable bit rate information arranged as frames, each from being encoded according to one of first and second encoding schemes, the frames being arranged in a known or ascertainable sequence with regard to said encoding schemes, the traffic prediction means comprising first and second frame prediction means, respectively operable to determine a predicted traffic demand for forthcoming frames encoded with said first and second encoding schemes, in accordance with said known or ascertainable sequence, wherein each said first and second frame prediction means comprises memory means operable to store prediction configuration information which is used by the respective prediction means to determine a predicted traffic demand for a forthcoming frame of that encoding scheme, and prediction configuration information setting means operable to determine a prediction error for a previous prediction carried out by that frame prediction means and to modify said prediction configuration means accordingly.

By operating in this way, an access controller in accordance with the invention is able to target VBR video traffic, characterised to a large extent by variable frame sizes. In previous arrangements, such as exemplified by US2005094558,traffic prediction in this context is not provided.

According to another aspect of the invention, traffic is allocated to a channel by use of a simple adaptive linear prediction algorithm that a station employs to predict the bandwidth requirements for future frames. This prediction, in turn, can be used to allocate TXOPs according to traffic demand.

It will be appreciated that aspects of the invention can also be implemented by means of hardware configured by suitable software, which can be provided to the hardware by any suitable medium. Such a medium could be an optical or magnetic storage device (such as a disk), a non-volatile memory device, such as a Flash memory, or a computer receivable signal carrying a file of instructions whether directly executable or encoded to allow a user to cause installation of a suitable executable program. For instance, a codec could be provided to a device by a received Short Message Service (SMS) message.

Specialised hardware devices could alternatively be provided for integration into a general purpose computer. By this, an ASIC or an FPGA could be provided with sufficient functionality such that the thereby configured computer results in a device in accordance with an aspect of the invention or operable to perform a method in accordance with an aspect of the invention.

A specific embodiment of the invention will now be described with reference to the accompanying drawings, in which: Figure 1 is a schematic illustration of a communications channel compliant with the IEEE 802.lle Standard, through the progress of one beacon interval; Figure 2 is a graph representing a traffic stream over time for the movie Jurassic Park'; Figure 3 is a graph comparing perfonnance characteristics of a VBR MPEG-4 traffic stream and its CBR equivalent, in accordance with a prior art example; Figure 4 is a graph of a cumulative probability density function for delay of single CBR and VBR video streams in accordance with a prior art example; Figure 5 is a graph comparing actual and predicted sequences of I frames for a first example of use of a specific embodiment of the invention; Figure 6 is a graph of autocorrelation of a prediction error generated in accordance with

the first example;

Figure 7 is a three dimensional graph of relative mean squared error for a suitable linear prediction algorithm employed in the specific embodiment, against step size and order; Figure 8 is a graph of relative mean squared error for the specific embodiment of the invention, on variation of the step size of the linear predictor; Figure 9 is a schematic diagram of a wireless communications network in accordance with the specific embodiment of the invention; Figure 10 is a schematic diagram of an access point in the wireless communications network illustrated in figure 9; and Figure 11 is a schematic diagram of a transmitter module of the access point illustrated in figure 10.

Given the potentially high data rate of future 802.1 1-based WLANs, one of the major applications in such networks is real-time multimedia traffic, e.g. VBR video traffic. A VBR traffic stream can be generated by an MPEG encoder that compresses a video signal at a constant picture rate (e.g. 25 frames/sec) to produce a coded bit stream with a highly variable bit rate. An MPEG video is divided into units called group of pictures (GOP). Each GOP consists of a combination of!, P, B frames, which have different statistical characteristics. In the usual fashion, I designates an initiation frame which is relatively uncompressed and for which decompression is not dependent on any other frame, P designates a predictive frame which is decoded with reference to a preceding frame, and B designates a bi-predictive frame, decoding of which is achieved with reference to both preceding and following frames. A typical GOP pattern consisting of 12 frames is:

IBBPBBPBBPBB

However, it will be understood that the actual pattern used in a GOP is dependent on several factors, such as the use to which the GOP is to be put. For instance, a video-conferencing facility will generally accept a relatively degraded picture quality, and so may involve very infrequent transmission of I-frames.

The specific embodiment described as follows is intended to provide a facility for the allocation of channel bandwidth to VBR MPEG-4 traffic in 802. lie networks, based on traffic prediction. Referring now to Figure 9, in an embodiment of the present invention, a wireless communications system 10 comprises an access point 100 with hard wired connection to outside communications facilities (such as a cable or satellite entertaimnents feed and/or the internet) and a plurality of wireless stations 200. These wireless stations are configured to request services from the access point, and to receive data in response to such requests.

In general, the wireless stations 200 operate in accordance with existing communications standards, in particular IEEE8O2. lie, and so require no further explanation. The access point is enhanced in comparison with 802.1 le and so will be described further.

The access point 100 comprises a processor 124 operable to execute machine code instructions stored in a working memory 126 and/or retrievable from a mass storage device 122. By means of a general-purpose bus 125, user operable input devices 130 (if supplied) are in communication with the processor 124. The user operable input devices 130 comprise any means by which an input action can be interpreted and converted into data signals, for example, DIP switches.

Audio/video output devices 132 are further connected to the general-purpose bus 125, for the output of information to a user. Audio/video output devices 132 include any device capable of presenting information to a user, for example, status LEDs. The user could be an install/service engineer or a home user.

A commurncations unit 140 is connected to the general-purpose bus 125, and further connected to a first antenna or set of antennas 150. By means of the communications unit 140 and said antenna(s) 150, the access point 100 is capable of establishing wireless communication with wireless stations within its coverage area. The communications unit 140 is operable to convert data passed thereto on the bus 125 to an RF signal carrier in accordance with the 802.11 e Standard.

In the access point 100 of Figure 9, the working memory 126 stores applications 128 which, when executed by the processor 124, cause the establishment of an interface to enable communication of data to and from wireless stations. The applications 128 thus establish general purpose or specific computer implemented utilities and facilities that are used in supplying wireless stations within the coverage area with requested data, using the available transmission resources.

To do this, the access point 100 implements a transmitter 160 as illustrated in figure 11.

As described in this example, the transmitter 160 is implemented in software, but there would be no obstacle to the skilled person in implementing it in hardware also; the function of the transmitter is entirely technical and tangible.

The transmitter 160 captures data to be transmitted in accordance with requests, the data being arranged as identified above in either I-, P-or B-frames.

A frame type separator 162 develops three streams of frames and then passes these three streams for independent traffic prediction in a linear traffic predictor 164. The function of the linear traffic predictor 164 follows.

It is acknowledged that simple prediction techniques can predict VBR video traffic successfully (in terms of prediction error) in wired networks such as ATM and IP networks (A. Adas, Using adaptive linear prediction to support real-time VBR video under RCBR network service model, IEEE/ACM Transactions on Networking, vol. 6, no. 5, 1998). Adaptive linear prediction does not require prior knowledge of the video statistics, nor does it assume stationarity, and is thus very suitable for on-line real-time applications. Assume the rate of the nth frame of a VBR traffic stream is s(n). A standard Mth-order one-step linear predictor has the form s'(n + 1) = w(l)s(n -1) = WTS(n), where M is the order of the linear predictor, w(1),1 = 0,...,M -1, are the prediction filter coefficients, W = [w(0), w(l),..., w(M -1)]T and S(n) = [s(n),s(n -l),...,s(n -M + l)]T.

The prediction error is e(n) = s(n 1)-s'(n + 1).

The normalized linear mean square (LMS) predictor 164 thus operates as follows: 1. Start with an initial estimate of the filter coefficients W(O).

2. For each new data point, update W(n) using the following equation: W(n + 1) = W(n) + pe(n -l)S(n -1)/fJ S(n -1) I2, where u is the step size.

If 0< p <2, the algorithm will converge towards the mean. Compared to the standard LMS algorithm, the normalized LMS algorithm is less sensitive to step size.

For MPEG-4 streams, because I, P, and B frames have different statistical characteristics, these are separated by the frame type separator 162 and the linear traffic predictor predicts each frame type separately.

Simulations using real video traces (presented later) have shown that the prediction error resembles white noise which is uncorrelated and therefore requires only small buffers.

Once traffic prediction has taken place, the prediction parameters are passed to a TXOP allocator 166. In 802. lie HCF, TXOPs are allocated every service interval (SI).

Without loss of generality, it is assumed that the SI is the same as the inter-frame time of the MPEG traffic (e.g. 40 ms or 25 fraines/s). The dynamic channel time allocation algorithm for an MPEG-4 flow i operates as follows: 1) Predict the size of the next frame s(n +1).

2) Then the TXOP for the next SI is given by: TX0J =[s,(n 1)/L,]*L IR. +0, where all the parameters are as defined previously.

Meanwhile, the MPEG traffic stored in a frame transmission queue 168 is retrieved by a transmission manager 170, in accordance with the TXOP allocations made by the TXOP allocation manager 166.

As described in this embodiment, traffic prediction is implemented at the QAP which monitors all the MPEG flows.

In an alternative to this embodiment, each wireless station 200 can predict its own traffic and pass this information to the QAP via some signalling mechanism before its TXOP is allocated. In this way, network resources can be better matched to the traffic demand, and hence guaranteeing the timely delivery of every frame. As a result both job failure rate and delay variation can be minimized.

TXOP allocation in the 802.11 e standard is based on the assumption of simple, constant bit rate traffic. Because VBR traffic can have very different frame sizes from one frame to another, fixed channel time allocation (e.g. according to fixed-size I, P, B frames) results in less optimal performance, e.g. higher job failure rate (the rate at which packets are dropped due to missing their deadlines) and higher delay. In contrast, in the proposed scheme the QAP allocates TXOPs based on accurate traffic predictions, therefore matching network resources to the traffic demand. Application QoS is maintained while network utilization is kept high. The simple, adaptive linear predictor described above does not incur much computation overhead. Further, if implemented at the QAP, there is no extra signalling overhead (between QAP and QSTAs).

The proposed TXOP allocation framework is flexible in that various traffic prediction techniques can be applied here according to the trade-off between performance improvement and implementation complexity. For example, instead of predicting each frame type separately, the rate of the next GOP (the sum of all the frame rates in a GOP) can be predicted. Further, it has been observed that I frames are the most important MPEG frames that contain the bulk of the video frame data. When I frame size changes significantly, P and B frame sizes also change greatly. Thus, the LMS predictor can be used to predict merely the I frame of the next GOP, which is often the largest in a GOP.

The dynamic TXOP assignment scheme can also be applied to other types of wireless networks, e.g. time slot allocation in future IEEE 802.16 networks.

Experiments To test the performance of the prediction algorithm, experiments have been carried out using real MPEG-4 traffic traces. These are available from Arizona State University Video Traces Research Group (http://trace.eas.asu.edu/). Six MPEG-4 encoded, medium quality video traces of popular movies were selected, each with a length of 60 minutes and a frame rate of 25 frames/s: Jurassic Park I (hereinafter designated Park); Silence of the Lambs (Lambs); Star Wars LV (Star); Star Trek First Contact (First); Die Hard Ill (Die); and Aladdin.

Table I summarizes their statistics: mean, minimum and maximum frame sizes (in bytes). It is evident that MPEG-4 traces have highly variable frame sizes (and bit rates) and are very bursty.

Park Lambs Star First Die Aladdin Mean frame size 1300 880 390 540 1200 770 Minimum frame size 26 28 26 26 26 26 Maximwnframesjze 8511 11915 4690 5945 8161 6735 -Table 1: Trace statistics It has been found that M = 12 is a good choice according to the Akaike information criterion (Adas). By experiments, it was found that,u = 0.2 provides a sufficiently good performance for most cases. Figure 5 shows the actual and predicted sequences of the I frames of Silence of the Lambs. Figure 6 presents the autocorrelation function of the prediction error.

It is observed that in contrast to the original, highly correlated video trace, the error process resembles an uncorrelated white noise, hence requiring only small buffer spaces. Table 2 tabulates the prediction results for all the six traces, with I, P, B sub-sequences listed separately. The performance metric is relative mean squared error (RMSE) or inverse signal-to-noise ratio (SNR'): >e(n)2 / RMSE= / 2 / s(n) -Park Lambs Star First Die Aladdin 1 0.0160 0.0499 0.0220 0.0354 0.0412 0.0383 P 0.0912 0.1737 0.7968 0.1531 0.1211 0.2189 B 0.0415 0.0406 0.2861 0.0592 0.0397 0.1721 Table 2: Prediction results From Table 2 it can be seen that, in absolute terms, the linear predictor is effective for MPEG-4 traces, although RMSE values vary depending on the video sequence predicted. Predictions of I frames generally have the best performance, while P frames are much more difficult to predict. RMSE is an inverse indicator of prediction quality -the smaller the RMSE value, the better prediction quality can be achieved. As will be seen from the table, the Star Wars trace appears to be the most challenging to predict, with quite high RMSE for P frame prediction. One possible explanation is that, because the movie content comprises a number of highly dynamic "action" sequences, there are many sudden scene changes, resulting in large prediction errors.. This would tend to suggest that genres in which content is somewhat less dynamic, or the result of long "takes", will be easier to predict. However, the present embodiment offers performance advantages, whatever the content.

The two predictor parameters p and Mhave been tuned to study their impacts on prediction performance. Figure 7 shows that for this particular trace, RMSE increases as Mdecreases or p increases. When Mis small, p has a strong influence on the performance. On the other hand, the larger p becomes, the more obvious the impact of M. With p fixed, RMSE does not change significantly beyond M= 12, indicating 12 is a suitable choice for M. Figure 8 demonstrates that, forM fixed at 12, RMSE drops almost linearly as p decreases, with its minimum reaching 0.2245 when p is 0.01. So when the step size is properly chosen, Star Wars can be predicted with reasonable accuracy. To this end, variable step size LMS algorithms can be used to overcome the difficulty of scene change and optimize the convergence performance of the LMS algorithm. In summary, the normalized LMS algorithm is able to predict a wide range of MPEG-4 traffic accurately enough to be used for bandwidth allocation in 802.1 le networks.

Claims (19)

1. CLAIMS: 1. An access controller for a wireless communications network,

   the access controller being operable to allocate access to a wireless communications medium for distribution of packet based information to other stations in the network in use, the access controller comprising traffic prediction means operable to process variable bit rate information arranged as frames, each from being encoded according to one of first and second encoding schemes, the frames being arranged in a known or ascertainable sequence with regard to said encoding schemes, the traffic prediction means comprising first and second frame prediction means, respectively operable to determine a predicted traffic demand for forthcoming frames encoded with said first and second encoding schemes, in accordance with said known or ascertainable sequence, wherein each said first and second frame prediction means comprises memory means operable to store prediction configuration information which is used by the respective prediction means to determine a predicted traffic demand for a forthcoming frame of that encoding scheme, and prediction configuration information setting means operable to determine a prediction error for a previous prediction carried out by that frame prediction means and to modif' stored prediction configuration information accordingly.
2. 2. An access controller in accordance with claim 1 wherein the first and second frame prediction means are operable independently of each other.
3. 3. An access controller in accordance with claim 1 or claim 2 and further comprising a third frame prediction means corresponding to a third frame encoding scheme, the third frame prediction means also comprising memory means operable to store prediction configuration information which is used by the third prediction means to determine a predicted traffic demand for a forthcoming frame of the third encoding scheme, and prediction configuration information setting means operable to determine a prediction error for a pre'ious prediction carried out by the third frame prediction means and to modify stored prediction configuration information accordingly.
4. 4. An access controller in accordance with claim 3 wherein said first, second and third encoding schemes comprise respectively a stand-alone scheme wherein a frame is encoded without reference to any other frame in the stream, a predictive scheme wherein a frame is encoded with reference to a previously processed frame, and a bi-directional scheme wherein a frame is encoded with reference to a previously processed frame and a frame yet to be processed.
5. 5. An access controller in accordance with claim 4 wherein said sequence is known as a sequence of twelve frames, commencing with a stand-alone frame and followed by a sequence of predictive and bi-directional frames.
6. 6. An access controller in accordance with claim 5 wherein said sequence comprises -IBBPBBPBBPBB wherein I represents a stand-alone frame, B represents a bi-directional frame, and P represents a predictive frame.
7. 7. An access controller in accordance with any preceding claim wherein each frame prediction means comprises a linear predictor.
8. 8. An access controller in accordance with claim 7 wherein the linear predictor is a one step linear predictor.
9. 9. An access controller in accordance with claim 7 or claim 8 wherein each frame prediction means is operable in accordance to determine predicted traffic s'(n +1) as: s'(n + 1) = Ew(l)s(n -1) = WTS(n), where Mis the order of the linear predictor, w(1),1 = O,...,M -1, are the prediction filter coefficients, W =[w(O),w(l),...,w(M _1)]T, and 8(n) = {s(n),s(n -l),...,s(n -M +
10. 10. An access controller in accordance with claim 9 wherein said respective prediction configuration information setting means is operable to determine an error e(n) = s(n + 1)-s'(n +1) and thus to modify stored prediction configuration information in accordance with: W(n + 1) = W(n)+ ,e(n -l)S(n -1)! HS(n -1) 112,where p is a step size.
11. 11. An access controller in accordance with any preceding claim and further comprising traffic control means operable to allocate traffic to designated transmission opportunities in the medium in accordance with predicted traffic demand information determined by the respective prediction means.
12. 12. A method of controlling access to a wireless communications medium for distribution of frame based information to other stations in use, said frame information being encoded by one of first and second encoding schemes, the method comprising: determining with which of first and second encoding schemes frame information to be distributed is encoded, determining a predicted traffic demand for a forthcoming frame encoded with said determined encoding scheme, by retrieving pre-stored prediction configuration information, arid using said prediction configuration information to determine said predicted traffic demand, and determining a prediction error for a previous prediction carried out by that frame prediction means and to modify said pre-stored prediction configuration information accordingly.
13. 13. A method in accordance with claim 12 wherein the step of determining with which encoding scheme said frame information is encoded, comprises noting the position of the frame represented by said frame information in a known sequence of encoding schemes employed to encode a sequence of frames of which the frame represented by said frame information is a member.
14. 14. A method in accordance with claim 12 or claim 13 wherein the frame information can be encoded by said first encoding scheme, said second encoding scheme, or a third encoding scheme, said step of determining further determining if said frame information is encoded with said third scheme as opposed to the first and second encoding schemes.
15. 15. A method in accordance with claim 14 wherein the frame information is part of an MPEG compliant GOP stream.
16. 16. A method in accordance with any one of claims 12 to 15 and further comprising allocating frame information to transmission opportunities defined on said channel in accordance with said predicted traffic demand.
17. 17, A method in accordance with claim 16 and further comprising transmitting in accordance with said allocation.
18. 18. A computer program product comprising a storage medium storing computer executable instructions operable to cause a general purpose computer apparatus to operate as an access controller in accordance with any of claims I to ii, or to perform the method of any of claims 12 to 17.
19. 19. A computer receivable signal carrying computer executable instructions operable to cause a general purpose computer apparatus to operate as an access controller in accordance with any of claims 1 to 11, or to perform the method of any of claims 12 to 17.