The present invention relates to communications system and, more particularly, to methods and apparatus for implementing wireless communications systems.
Some link layers technologies (or mac technologies), like flash-OFDM developed by Flarion Technologies, Inc., provide a number (we use 16 for illustration but could be more or less) of mac_streams that each can hold a small number (e.g.: 1-10) of packets in them. In some systems of this type there is a potential for anything up to 160 packets or more that could be stored in mac_streams for a particular MN.
IP Packets (e.g.: 40-1500 bytes) are typically segmented into smaller mac_frames (e.g.: ˜24 bytes) within each mac_stream. For purposes of explaining the invention we assume that Link Layer Automatic Repeat Requests (ARQs) includes frames but that they also retain the knowledge of packet boundaries as occurs in some systems. Also, for purposes of discussion, note that a packet may start or end within the middle of a frame.
During a hand-off a Mobile Node (MN), e.g., portable receiver/transmitter, in a mobile (wireless) communication system connects to a new Access Router (AR) and an indication is sent to the old AR used by the MN about the movement. The old AR may then need to forward to the new AR any packets stored for that MN within mac_streams at the old AR, and any subsequently arriving packets for that MN, e.g., so that they can be transmitted by the new AR to the MN. A typical network scenario when using Mobile IP for general mobility management and inter-AR hand-off, is shown in
Note that in general;
In addition, note that whilst the flows described are delineated as packets, it is equally applicable and beneficial to be able to redirect partial packets delineated as mac_frames, when part of a packet as frames has been sent via the old link to the MN before the old link was lost, necessitating the remaining frames in that packet to be forwarded to the new AR for forwarding via the new link.
A range of hand-off scenarios can be envisaged in general, and specifically with this example network scenario, depending on the hand-off triggers from the link-layer and the IP layer, as well as the capabilities of the access links, ARs and MN.
In a L2 Break before Make (L2BBM), where the L3 can only be BBM by implication, the old link is lost, whether by accident or design, and the new link is only available after this event, and not before.
In a L2 Make Before Break (L2 MBB) with L3 BBM, the old and new links are available in parallel but the MN IP stack can only operate over one of these links at the same time and there is a finite time for moving to the new link and configuring the IP stack on the new link.
In a L2 Make Before Break (L2MBB) with L3 MBB, the old and new links are available in parallel as are the MN IP stacks to exploit those links.
Where possible, it is useful to send all packets A and B over the old AR link before hand-off (C=0). However, in hand-off cases, where the old link is lost prematurely, the remaining IP packets and mac_frames at the old AR should be transferred to the new AR so that they can be delivered to the MN over the new link, and hence effect a loss-less hand-off.
A first problem is apparent in that the old AR needs to know what packets/frames must be forwarded to the new AR but only the MN knows with certainty what was actually received from the old AR. In addition, the new AR needs to know what packets/frames need to be sent over the new link to the MN. This is particularly difficult in cases where the links use Automatic Repeat ReQuest (ARQ) techniques to deliver mac_frames reliably by managing repeat attempts over access links, and by receiving reception acknowledgments from the MN. In this case it is necessary for the old AR to send its ARQ state forward to the new AR along with the associated packet/frame data in a process known as Context Transfer. If too much data is sent forward (safety first) then duplicate information may be sent to the MN over the two links (inefficient, reduced throughput and increased latency to other packets for the MN) whilst if insufficient data is sent forward then frame and hence incomplete packet loss may result.
A second problem is then clear in that the forwarded packets/frames will take a finite time to reach the new AR link, via a core (e.g., land based) network used to connect ARs, during which time packets/frames can't be delivered to the MN even if the new AR mac-layer or IP stack are already configured. This appears to the MN as a lack of connectivity that will temporarily increase the latency of application packets in transit.
A third problem is then apparent if we try to forward over all packets/frames, for a specific MN undergoing hand-off, as fast as possible to the new AR to minimize this latency. The result will be a burst of packets on the backhaul uplink from the old AR, through the core network, and on the backhaul downlink to the new AR. This burst of packets can interfere with other packets on the links, potentially causing buffer overflows (losses), increased latency and/or reduced throughput for other application flows.
A fourth problem is that the competition between the forwarded packets/frames for the handed off MN, and other application flows for other MNs, normally occurs independent of the relative importance of the underlying application QoS (Quality of Service) requirements. If we forward all packets quickly to exploit the new link early, then we maximally affect other application flows, whilst if we smooth the forwarded flow over time then the new link is under-utilized.
In view of the above discussion it becomes apparent that there is a need for improved methods and apparatus for handling hand-offs, which address some or all of the above discussed hand-off problems.
In the old AR and new AR, the packets and frames under delivery to a MN will typically have been classified into QoS classes. Packets and frames of each class are then provided, based on the QoS class, the appropriate resources and scheduler behavior, for correct delivery over the access link. For example Voice packets will typically require much lower latency than bulk data and these relative performance bounds are well known. For discussion purposes, we will call the classified IP packets, and its constituent mac_frames after those packets have been fragmented at the link layer, a mac stream. In general, each packet can start at any point within a mac_frame, following completion of a previous packet from the same mac_stream, with the location of the packet start therefore being offset from the frame start. The scheduler will treat each mac_stream independently, from a QoS perspective, as each constitutes a sequence of packets that must be delivered in order, and the scheduler can allocate resources to each mac_stream to affect relative mac_stream performance. Note that whilst the packets in each mac_stream are delivered in order, the ARQ mechanism associated with each mac_stream should be able to deliver mac_frames out of order to support retransmissions of errored frames.
For ordered delivery of IP packets within each mac stream, obviously,
Flow E=remaining A, then B, then D flow packets
For reliable packet delivery within each mac_stream, obviously, all outstanding packets, frames and associated ARQ state should be forwarded to the new AR and the MN should internally transfer the ARQ state for the old AR link to the new AR link.
In accordance with a first feature of the invention the transfer of the said packets, frames and ARQ state is split up into multiple parallel and serial phases, with one parallel phase for each mac_stream, each parallel phase then being broken up into multiple serial phases. Said phases and associated data are marked in sympathy with the QoS marking of the associated mac_stream, and then treated during forwarding and Context transfer according to the relative QoS marking between other competing mac_streams from the same MN, and relative to forwarded mac_streams from competing MNs undergoing hand-off, and relative to other application flows to stationary MNs.
To facilitate this, the forwarded IP packets and frames are appropriately marked using the IP header Type of Service or Diff-serv Code Point (TOS/DSCP) QoS fields for the mac_streams, appropriate bandwidth is assigned on the affected links (old AR backhaul uplink, in the core, and on the new AR backhaul downlink) for forwarded packets as part of hand-off, and a sympathetic scheduling algorithm is employed on these links and in the old and new AR mac layer, to arbitrate between the multiple flows on any link to ensure timely and fair delivery of packets and frames on the links.
The parallel phases between each mac_stream enables each new AR mac_stream to be configured and populated with data at a rate suitable for that mac_stream for that MN and specifically enables different MNs to have differential hand-off performance according to their service level (e.g., gold, silver, bronze, etc.).
The serial phases enable the context transfer to be undertaken in multiple stages, as not all data in any mac_stream needs to get to the new AR early, immediately, or at all. Only sufficient data needs to be sent to avoid the new AR, and hence the MN, from being starved of critical packets. The appropriate set of serial phases for a particular MN and mac_stream depends on other techniques described below as well as scheduler QoS configuration in the old AR. In particular, given network conditions and by using packet timestamps and serial numbers, the schedulers used in a system implementing the invention can detect stale and unwanted mac_frames and discard them locally rather than fruitlessly forwarding them to the new AR where they will be dropped anyway.
Hand-off ‘aggressiveness’ in terms of advanced forwarding for any MN mac_stream is controlled by the number of competing hand-offs at any point in time, the amount of bandwidth/latency assigned to those hand-offs on a per service level basis, the service level and other QoS attributes of a particular MN, the QoS attributes of the mac_stream, the number of packets and frames in said mac_stream that need to be transferred to the new AR, the complexity of the associated ARQ state that itself is dependent on the retransmission status of frames within the mac_stream.
The advantages of this invention is that we get QoS control of hand-off providing early fair use of the new link, strict ordering of packets is maintained with minimal duplication over the air-link, hand-off burstiness (to this and other MN traffic) is prevented, packets can be intelligently dropped early, and hand-off aggressiveness is controlled and can be sympathetic to link/network conditions.
One feature of this invention is to break the old link state into different serial phases/packets/QoS markings so that QoS queuing can treat them correctly. Specifically, rather than waiting until the old link breaks, the ARQ state is sent gradually over from the moment the hand-off process starts and the new AR address is known. In accordance with the invention each mac_stream has an Identification (ID). Each packet in the stream has a Sequence Number (SN) and an Offset for example. This Offset indicates the ARQ frame number for the start of that packet and the byte offset of the packet start within that frame. The position of each byte in an IP packet is then known in its constituent frames. The packet sequence number increases the effective size of the frame numbering space from what is normally a relatively small ARQ retransmission window (typically 256). The packet sequence number can be used to detect packets that get lost/re-ordered during forwarding from old to new ARs, and the mac_frame number and offset from front of packet is used to locate any packet in a normalized ARQ space for each mac_stream at the new AR. The combination of the packet SN and the ARQ SN, hence referred to as the SN, uniquely identifies a mac_frame in the system, and the offset of the packet within a mac_frame enables the packet boundary to be tracked.
The complete ARQ window state for each mac-stream would be, for example;
An example packet forwarding strategy is one feature of this invention. The packet forwarding strategy sends all flow A packets down the old link whilst sending flow B packets (that arrive at the old AR after MIP hand-off commences) to the new AR. Flow D will arrive at the new AR directly of course as a result of the MIP hand-off signal at the HA.
We wish to ensure that the MN receives flow A, then flow B then flow C packets in order in each mac_stream. Therefore, at the new AR, flow D packets are queued behind flow B packets, and neither should be sent to the MN until flow A has been delivered to the MN and the old link dropped in a planned hand-off.
However, the old link can be dropped at any time as a result of radio effects, forcing the timing of the hand-off. In this case, all the flow A packets might not have been delivered over the old link, and no flow B packets might have arrived at the old AR for forwarding, or those that have arrived might not have arrived yet at the new AR for the new link. We therefore have to still ensure that flow A, B and D packets can still be delivered in order to the MN, and ensure that the right packets are available at the new link quickly so that the new link can be used immediately after the old link is dropped. Note that the arrival of flow B packets at the new AR does not help us here if there are still flow A packets unsent at the old AR . . . we still need to delay flow B packets until the unsent flow A packets (and constituent frames) have been sent over to the NAR . . . Note that this is not strictly necessary from a packet re-ordering perspective because many applications can cope with some re-ordering. However this may be beneficial for some applications that cannot cope with re-ordering and hence a characteristic of the specific mac_stream used by that applications for its packets.
An additional aspect of this invention is therefore to queue packets from flow A, B and D in such a way that ordered delivery is maintained. This means flow B packets are not sent to the mac_layer in the new AR, and fragmented or shredded into mac_frames until the remaining flow A mac_frames from the old AR are received at the new AR following the loss of the old link. In addition, flow D packets from the HA should be held up behind both flow A and flow B packets.
An alternative and preferred implementation of this invention is to instead keep flow B packets in the old AR and classify and queue them behind the flow A packets into the mac-streams, and then send remaining flow A packets (and frames) and then flow B packets to the new AR when the old link breaks. Note that this means that the hand-off signal from the new AR to the old AR does not affect flow B routing and it is only the breaking of the old link, with packets still outstanding in the old AR mac_streams that causes packets to be forwarded to the new AR as flow C. Thus, ordering is controlled at the old AR rather than at the new AR between flow A and flow B, whilst ordering between A/B and flow D is controlled at the new AR. Whichever strategy we use though, we still have potentially ARQ state and a lot of packets/frames to send to the new AR before the new link can be used (more to be sent if A/B ordering is controlled at the new AR of course).
The earlier we start the hand-off (higher signal to noise radio power threshold for hand-off) the longer the old link can be up to forwarding remaining packets, at an increased cost to the overall system because of transmit powers to/from two base stations and the occupancy of two sets of radio resources such as tones and session ids etc.
The faster we try to make the hand-off, increasing efficiency of the radio system, the less chance we will have to drain flow A (and optionally B) over the old link.
We cannot predict how quickly the S/N will drop on the old link in the future, so some unplanned (i.e. unexpected and premature) old link removal will still occur, independent of the above timings.
We cannot send remaining old link flow A frames instantaneously to the new AR, and the delay will depend on paths lengths, network buffer status, QoS settings on the transferred packets.
We can therefore increase radio costs to increase probability that flow A (and optionally B) is fully delivered over old link, then accept a certain level of unplanned drops and associated new link delays whilst flow A (and B) information is transferred, or put in new mechanisms that increase the probability that flow A will be delivered over the old link and increase the probability of the correct packet (by flow order per mac_stream) being available at the new link.
One aspect of this invention is for the QoS scheduler to assign increased priority to flow A packets from a MN undergoing hand-off, relative to other MNs flow A packets not undergoing hand-off, so that for a given hand-off time, the probability of flow A packets across all mac_streams being delivered is higher. In such an embodiment, even if all packets are not drained over the old link, we have given the MN earlier service than it would have got without a hand-off, which compensates for the future lack of service that the MN will experience as a result of the remaining flow A packets being forwarded to the new AR after the old link is dropped. The shorter the hand-off time, the more efficiently we use radio resources, so the greater the increase in priority we need to give to the flow A packets to make up for the potential flow A packet transfer delay between the old AR and the new AR.
Two limitations occur with this approach though. The first is that the increase in priority we can give the handing off MN is limited by the effect of that higher priority on competing traffic to other MNs at the old AR. In the same way that the forwarded flow A/B packets interfere with traffic on the uplink from the old AR and on the downlink to the new AR, so transferring some of those packets over the old AR downlink (in additional to normal QOS treatment for that MN) can interfere with other MNs traffic over the air-link. Conceptually, we would wish to bound the priority increase so that the interference from the hand-off of a MN is shared equally between the MNs at the old and new ARs.
The second limitation is that this oversending is still dependent on the characteristics of the air-link and at any moment we could find that, from a priority perspective, we would wish to preferentially schedule this MN during the hand-off, but that at the same time the radio channel to that MN is very bad due to instantaneous effects or planned transmit diversity cycle. Therefore whilst clearly useful this may not be a complete solution in some instances.
As well as sending the ARQ state early to the new AR, one aspect of this invention is for the old AR to send a copy of a flow A (then B) packet to the new AR when it reaches a mac_stream specific trigger depth away from the front of that mac_stream queue. The closer a packet gets to the front of a mac_stream queue, then the more likely it is to be approaching imminent service over the air-link, and hence the greater the need for it to be at the new AR when the old link breaks. In particular, when a packet is in the ARQ loop and has seen service for one of its associated mac_frames, then on loss of old link, that mac_stream cannot deliver any additional mac_frames (including repairs of the sent frame) to that MN from that mac_stream, until the packet under service is sent to the new AR. Therefore as well as delaying service to this packet, any subsequent packets will also be blocked.
As a flow A packet progresses in the mac, it is marked when it reaches the trigger point, causing it to be forwarded to the backhaul QoS scheduler along with an update of the ARQ progress since the last report, and the ARQ SN information for this packet. This information can be preferably combined into a single special packet (general ARQ plus packet ARQ state plus packet) or sent as two packets (general ARQ SN update+packet specific stuff) or even three individual packets. When combining the packet ARQ and actual packet into a single inter-AR packet (for example), the packet itself would be encapsulated before forwarding to the new AR as in MIP, but will now include the ARQ_synch information which could be implemented as a shim layer underneath the MIP header or as an IP custom option header or a custom encapsulation. Note that the new AR strips the ARQ_sync information from the frames/packets before it delivers to the mobile node.
At some point later, the data packet will be sent from the old AR, and eventually arrive at the new AR, as part of flow C. The packet sequence number information enables the new AR to detect packet drops and adjust the mac-stream and ARQ SN appropriately. As the ARQ SN updates are received at the old AR backhaul QoS scheduler, it can report on the completion or dropping of a previously delivered data packet copy from the old link mac_stream, and hence cancel the transmission of that data packet copy to the new AR. In addition, by sending that ARQ update onto the new AR, it can delete packets from the mac_stream queue at the new AR, that have been delivered successfully (or dropped) from the old link mac_stream queue.
Some examples of non-exhaustive Trigger policies, which can be used, are;
The QoS mechanisms will try to improve the chance of flow A packets being delivered over the old air-link. Meanwhile, packets of high QOS flows are duplicated and sent early to the new AR so that if the old link breaks, these high QoS/low latency flows will be unaffected by the hand-off.
The longer we can afford to maintain the old link in a planned hand-off, the less useful are the mechanisms described here, but they will always be useful for dealing with the unplanned loss of the old link. The longer the old link is up, the less efficient is our use of the overall radio system resources and so appropriate alternative use of backhaul resources to save those air-link resources may be preferable.
Note that the greater the bandwidth per MN, as a fraction of the cell resource, the greater the saving as the required time to drain flow A for that MN increases. In addition, the greater the service level of that MN then the less effective is the use of the increased priority over the air-link to drain flow A. Finally, the more bursty the packet arrivals are for MN, the harder it is for the scheduler to instantaneously provide sufficient air-link resources to drain flow A because those resources would normally be spread over a period of time (but this is potentially ok because the fact that it would be spread over time at the old AR means that it does not need to be instantaneously forwarded to the new AR).
In one implementation, packet copies and ARQ update signaling is minimized in the old AR, by using the old link mac_stream memory to hold the flow A packets that have reached the trigger point in any mac_stream, until they have been scheduled by the backhaul scheduler. In addition, but independently, to efficiently maintain flow A/B ordering and to enable flow B packets to also be delivered over the old link during BBM or MBB hand-offs, flow B packets should continue to be delivered to the old link mac_stream queues as this will ensure they are classified correctly behind the flow A packets, will only be forwarded to the new link when they reach the trigger point, and will avoid the need for the hand-off completion timing being tied in an artificial way by the amount of packets in flow A when the hand-off commenced (e.g., by the desire to drain flow A over the old link, and to send all flow B packets to the new link, which leads to the problem of old link under-utilization if there are no/few flow A packets when hand-off commences, and subsequently arriving flow B packets cannot make use of the old link).
Similarly, at the new AR, flow A and flow B packets arriving at the new AR should be placed in a single queue in delivery order. This means only arriving flow D packets to the new AR have to be kept in a separate queue so that they can be added after hand-off completion behind flow A and flow B packets, instead of having three queues which would be required if flow A and B weren't combined in order at the old AR, and if flow B was sent to the new AR in advance of flow A packets.
For the combined flow A/B queue at the new AR, this queue can be implemented at the IP layer so that ARQ updates can be efficiently processed without having to traverse the IP/mac interface. The IP A/B plus D queue would only be pulled into the layer 2 mac_streams at the new AR after the old link is lost and the MN sends a request to use the new link. This is because we wish to delay fragmentation of the IP packets into mac_frames until the final required packets (for delivery to the MN from the new AR) are known. This avoids the ongoing intermediate complications arising from the reformatting of the mac_frames (offset and ARQ SN start) due to the variable length of IP packets, as any packet that has been copied to the new AR is completed at the old AR, or as forwarded packets are reordered during forwarded or retransmitted or lost during forwarding. This also saves linked list manipulations if the ARQ SN space and associated mac_frames are implemented as a linked list of address locations pointing to the start and stop of each mac_frame of each packet in mac memory.
Note that the transfer of these packets into the mac-layer could still be made before or after the final ARQ state information is received from the old AR on loss of old link.
An additional aspect of this invention is to send the IP queue to the mac layer before the final ARQ information is received and restart sending from the front of the first packet in each mac_stream. This ensures that sending can commence immediately the new link is set-up, and is decoupled from the arrival of the final ARQ state because there is a chance of this state being either delayed or dropped in error. The effect of this is that the MN may receive duplicate ARQ SN transmissions, for mac_frames that it signaled that it had received over the old link. This looks to the MN like a false retransmission (a known ARQ possibility) that is caused by an ACK from the MN being converted to a NACK at the AR as a result of a transmission error. In this case though, it occurs with the addition of an increased delay (and SN evolution) between the original mac_frame being sent (and ACKed) and the unnecessary retransmission being made. This is because in one exemplary example of such an ARQ system, the receipt of the NACK blocks the sending of any new mac_frames until the retransmission processing is completed, which bounds the ARQ SN evolution that can occur between sending a mac_frame and receiving the associated NACK (propagation decode/encode and processing delays). This increased ARQ SN evolution can be handled provided the SN space cannot wrap in this time. It is therefore important that incremental ARQ SN progress information is sent from the old AR to the new AR to minimize this duplication and avoid the wrapping, or the information should include the packet SN information as well and be exchanged with the MN before sending over the new AR link (new AR to MN report). The MN can then simply and safely compare the SN of any mac_frame with its own SN state, during hand-off, and reACK the retransmissions even if the contents of the mac_frames have been corrupted because the contents have already been received at the MN via the old link. Additional protection from this ARQ SN wrap is provided of course if the IP packets are not sent to the new link mac layer until the final ARQ state is received. An additional aspect of this invention is therefore to send this final ARQ state information from the MN to the new AR. This is useful whether or not the IP queue is sent to the mac layer with or without old AR final ARQ state. If the old AR final ARQ state is not received, and the IP queue is sent to the mac_layer, then the MN final ARQ state report to the new AR will be delivering very similar (but more accurate) information and will therefore minimize unnecessary retransmission. Even if the IP queue is sent to the mac after reception and application of the final ARQ state from the old AR, the MN has more accurate information on the ARQ state for what was actually received by the MN (as opposed to what was sent by the old AR). Therefore, the MN sending its final ARQ state information to the MN will suppress additional unnecessary retransmissions.
In the next aspects of this invention, called ARQ correlation, the tracking of the ARQ SN state across multiple parallel AR links during L2, and the sending of packet copies to the new AR, enables the system to support the transmission of mac_frames from the same packets to the MN in parallel from multiple ARs. The MN will, for example, be able to select mac_frames from either (any) link for the same packet and will correlate ACKs and NACKs across the two links so that successful mac_frame reception over the new link will result in an ACK being sent over both the old and new links even though the mac_frame from the old link was received in error. ARQ update reports from the MN can then be sent uplink to the ARs (and exchanged between ARs) to prevent one AR attempting to send a mac_frame that has already been received from the old link, and has the additional benefit of avoiding the need for the two links to be synchronized (same mac_frame must be sent at the same time as in CDMA soft-hand-off). As the two links have different error characteristics, loads etc so their ARQ spaces will move relative to each other enabling both to contribute to the safe and efficient delivery of each packet. This mechanism will reduce the required S/N ratio required on a candidate new link and for dropping the old link because the combination of the two links S/N ratio is reducing the effective BER, as in CDMA soft-hand-off. This is likely to produce operational and efficiency benefits for MBB capability.
In accordance with additional features of this invention, covering the case of the uplink transmission direction, the old AR sends partially received packets, in the form of a subset of its component mac_frames, to the new AR along with its final AR state, after old link breaks. The scheduler is organized so that during hand-off whole packets are given higher priority than normal so that if a packet is started it is quickly completed in the uplink direction. These partially received packets potentially need to be sent as soon as they are received, rather than as a block for the QoS reasons covered for the downlink direction. With early forwarding we therefore also send updates to the old AR backhaul interface and the new AR as whole packets are completed. The new AR may or may not send any packets to the backhaul until the final ARQ report is received from the old AR (although the cost of sending a whole packet unnecessarily here is much less in terms of bandwidth although it does result in a duplicate packet (one from each AR)) towards the session peer of the MN undergoing hand-off. Therefore, the MN should send its uplink final ARQ state to the new AR because this provides faster delivery of this information to the new AR, as well as additional reliability, for the equivalent information from the old AR although in this case it is the old AR information that is more accurate. Finally, soft hand-off involves the correlation of the received frames from the old link, with the frames received over the new link, in the new AR, so that packets can be sent when fully formed irrespective of where each component frame was received. The ARQ state and packet sequence number information from the two links enables this correlation to be made. The potential loss of a mac_frame between the old AR and the new AR can be dealt with in one of two ways. They can be sent using TCP/SCTP retransmission capabilities or the new AR signals to the MN information on outstanding mac_frames for a packet so that the MN resending over either link can repair this frame loss. This acts as a very slow ARQ repair mechanism and requires the MN to keep packets until a packet completion event, containing a packet SN is received from the new AR.
The present invention is a continuation of U.S. application Ser. No. 10/266,244 filed Oct. 8, 2002 titled, “METHODS AND APPARATUS FOR QUICKLY EXPLOITING A NEW LINK DURING HAND-OFF IN A WIRELESS NETWORK,” which claims the benefit of U.S. Provisional Patent Application Ser. No. 60/328,468 filed Oct. 10, 2001 titled “METHODS AND APPARATUS FOR QUICKLY EXPLOITING A NEW LINK DURING HAND-OFF IN A WIRELESS NETWORK”, both of which are hereby expressly incorporated by reference.
Number | Date | Country | |
---|---|---|---|
60328468 | Oct 2001 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 10266244 | Oct 2002 | US |
Child | 12256336 | US |