Packet fault-tolerant bounding method based on protocol natural redundancy

Abstract

The present invention provides a packet fault-tolerant delimitation method based on protocol natural redundancy. The method comprises: receiving an aggregated packet sequence; performing rough delimitation on the aggregated packet sequence to obtain a rough delimitation result sequence, including multiple suspected starting boundaries; treating each suspected starting boundary as a node on a path, starting from the first suspected starting boundary, using protocol natural redundancy to construct a path metric for path search, including: for the current suspected starting boundary g b , g b Set as downstream node, search for g b All possible upstream nodes g a ; For each upstream node g a , assuming g a to g b There is only g between a As the starting boundary of the group, calculate the distance from g1 to g a Then to g b The path cumulative metric of the constructed path; the path with the largest path cumulative metric is taken as the current best path, and the search continues for the next suspected starting boundary until the search is completed for the last suspected starting boundary, and the search ends; after the search is completed, the current best path is output as the group boundary sequence estimated by the present invention.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communications, and in particular to a packet fault-tolerant delimitation method based on protocol natural redundancy. Background Art

[0002] In many communication systems, the transmitter aggregates short packets (or data packets) at a specific protocol layer into longer packets (or bursts) before transmitting them downstream. The receiver then performs the reverse process at the corresponding protocol layer to locate the starting boundaries of multiple consecutive short packets, thus completing packet delimitation (or burst segmentation). This approach reduces header encapsulation overhead for lower-layer protocols and improves throughput. However, the introduction of packet delimitation also makes the system more sensitive to transmission errors. Typically, the starting boundary of a packet is determined by sliding a search for a synchronization marker on a sequence of hard-decision results in the received data, such as the 0x7E boundary marker used by High-level Data Link Control (HDLC) and Point-to-Point (PPP) protocols. Bit errors in the received data can lead to missed packet boundary detections or false alarms, resulting in incorrect delimitation and segmentation of the aggregated packet. Therefore, packets are often equipped with an error check field. If a segmentation error causes the check to fail, the corresponding packet will be required to be retransmitted. Under low signal-to-noise ratio conditions, the frequency with which error retransmission mechanisms are triggered increases significantly, significantly reducing system throughput. In some services with high real-time requirements and one-way broadcast services, this mechanism becomes very difficult or even impossible to implement.

[0003] In order to deal with this situation, we can only continue to mine other available natural redundant information from the protocol. References [1] (Ali U, Kieffer M, Duhamel P. Joint protocol-channel decoding for robust aggregated packet recovery at wimax MAC layer [C]. IEEE 10th Workshop on SignalProcessing Advances in Wireless Communications, 2009: 672-676) and [2] (Fan Liang, Wang Xiaomei, Yang Dongyu. A PDU fault-tolerant delimitation algorithm using optimal path search [J]. Journal of Xidian University, 2016, 43 (5): 160-166) understand packet delimitation from the perspective of channel decoding, and provide specific algorithms from the perspectives of symbol-by-symbol decoding and sequence decoding. Reference [1] considers the frame aggregation process as a Markov process and constructs a Trellis diagram with an indefinite number of steps that can represent all possible aggregation schemes. Finally, the BCJR algorithm is used to estimate the number of packets and the boundary, and an 8dB gain is achieved in the MAC packet delimitation of the Wimax protocol compared with the traditional hard segmentation method. Although this method achieves near-optimality in theory, the BCJR algorithm often uses normalization processing at each moment in the actual forward and backward recursion to prevent precision overflow, which makes the given packet number estimation criterion difficult to implement. Reference [2], from the perspective of sequence decoding, regards the delimitation problem of the protocol data unit as a path search problem, first performs a rough delimitation based on the Hamming distance, and then performs the best path search on the rough delimitation result to obtain the final boundary judgment result. The rough delimitation step adopted by this method can effectively reduce the subsequent search space, but the path metric criterion given is an empirical approach and is bound to have a gap with the theoretical performance. Both approaches divide protocol fields into fixed fields, length fields, data fields, checksum fields, and other fields, and prioritize redundancy within fixed, length, and checksum fields to achieve packet delimitation. However, cross-layer consideration of redundant fields is relatively limited, and other types of protocol redundancy, such as repeated field redundancy and packet length distribution redundancy, have not yet been considered. Summary of the Invention

[0004] Aiming at the problem of insufficient performance of packet delimitation under low signal-to-noise ratio, the present invention provides a packet fault-tolerant delimitation method based on natural redundancy of the protocol.

[0005] The present invention provides a packet fault-tolerant delimitation method based on natural redundancy of a protocol, comprising:

[0006] Step 1: Receive an aggregated packet sequence;

[0007] Step 2: performing similarity matching on the fixed field and all fixed field observation vectors on the aggregated packet sequence to obtain a rough delimitation result sequence, wherein the rough delimitation result sequence includes multiple suspected start boundaries arranged in order;

[0008] Step 3: Treat each suspected starting edge as a node on the path, and starting from the first suspected starting edge, use the natural redundancy of the protocol to construct a path metric for path search. The path search process includes:

[0009] For the current suspected starting boundary g b , g b Set as the downstream node, which will meet the packet length constraint and be located at g b All previous suspected starting boundaries are taken as g b Upstream node g a ;

[0010] For each upstream node g a , assuming g a to g b There is only g between a As the starting boundary of the group, calculate the distance from g1 to g a Then to g b The cumulative path metric of the constructed paths based on the natural redundancy of the protocol;

[0011] The path with the largest cumulative metric is taken as the current best path, and the search continues for the next suspected starting boundary until the last suspected starting boundary is found.

[0012] Step 4: After the search is completed, the current best path is output as the final group boundary sequence.

[0013] Furthermore, step 2 specifically includes:

[0014] The correlation coefficient between the fixed field and the current fixed field observation vector is calculated. When the correlation coefficient is greater than the set threshold, the position of the last bit of the current fixed field observation vector is taken as a suspected starting boundary and added to the coarse delimitation result sequence.

[0015] Further, calculate from g1 to g a Then to g b The path cumulative metrics based on the natural redundancy of the protocol for the constructed paths specifically include:

[0016]

[0017]

[0018] in, k is a fixed field, u a Pkt a→b The length field is calculated, and its value is determined by g b -g a Size determination, R k and R u The fixed field k and length field u are a Random fields of equal probability 0 / 1 with the same dimension, is a single-step metric, M 1→a From node g1 to node g a The cumulative metric of the path.

[0019] Furthermore, it also includes:

[0020] If the redundant field types in the protocol stack used by the transmitting and receiving ends include repeated fields, then based on the packet boundary sequence output in step 4, the repeated field observation vectors of each packet are extracted to form a repeated field observation vector sequence;

[0021] Estimating a hard decision result of the repeated field according to the repeated field observation vector sequence;

[0022] Add the hard decision result of the repeated field to the original fixed field sequence to obtain a new fixed field sequence;

[0023] The path search is performed again according to the new fixed field sequence, and the group boundary sequence obtained by this path search is used as the final group boundary sequence.

[0024] Beneficial effects of the present invention:

[0025] (1) For downstream node g b , by the upstream node g a Jump to it and determine the packet Pkt a→b In this event, the existing path search method only uses the “identification node g a The starting boundary of a potential group is ignored. a to g b The other nodes between them are not the starting boundaries” is another hidden information, which causes the path metric in the existing path search method to be inaccurate. To solve this problem, the present invention considers the above two pieces of information at the same time, that is, in “g a to g b There is only g between a The calculation from g1 to g is based on the condition of "starting boundary of grouping" a Then to g b The path cumulative metric of the constructed paths can significantly improve the group delimitation performance.

[0026] (2) In order to address the shortcoming of the existing methods that the protocol field redundancy is not fully utilized, the present invention further introduces a decision feedback algorithm for repeated fields. That is, when repeated fields are available in the protocol stack, the estimated repeated fields are added to the original fixed fields and the path search is performed again, which can further improve the packet delimitation performance.

[0027] (3) The test results on the IPv4 / MAC / HDLC protocol stack show that the method of the present invention has significantly improved performance compared with the existing path search method and is close to the theoretical performance limit. It can also be extended and applied to other different protocol standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a group aggregation model;

[0029] Figure 2 A flow chart of a packet fault-tolerant delimitation method based on protocol natural redundancy provided by an embodiment of the present invention;

[0030] Figure 3 A schematic diagram of an improved path search process provided by an embodiment of the present invention;

[0031] Figure 4 This figure compares the performance of group delimitation using different methods provided in the embodiments of the present invention. DETAILED DESCRIPTION

[0032] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0033] Before introducing the technical solution of the present invention, in order to facilitate understanding of the technical solution of the present invention, a brief introduction to the aggregation group sequence and group delimitation is given here. Figure 1 As shown, n packets are sequentially spliced ​​into a sequence v, and the header of the lower layer is added in front of the sequence v to form a complete aggregate packet sequence.

[0034] For each packet, it consists of the payload source data and the header and tail of the multi-layer protocol. According to the role played in packet delimitation, these fields can be roughly divided into fixed field k, length field u, repeated field r, header check field c1, other header fields o, payload data field b and check field c2 covering the entire packet, etc. The corresponding observation information is represented as y k 、y u 、y r 、yo 、y c1 、y b and y c2 .

[0035] Since the length of the lower layer header is usually fixed, the receiving end only needs to consider the received sequence v corresponding to n packets when separating the packets. Therefore, the lower layer header is no longer taken into consideration in the aggregation and segmentation of data involved below. Assume that the length of the i-th packet is ξ i bits, the total length of n packets is bits, then the real starting boundaries of n packets ζ1,ζ2,...,ζ η They are Packet delimitation is to find these boundaries based on the received sequence y and the known natural redundancy information of the protocol when the number of packets η is unknown.

[0036] Example 1

[0037] like Figure 2 As shown, an embodiment of the present invention provides a packet fault-tolerant delimiting method based on protocol natural redundancy, comprising the following steps:

[0038] S101: receiving an aggregated packet sequence;

[0039] S102: performing similarity matching on the fixed field and all fixed field observation vectors on the aggregated packet sequence to obtain a rough delimitation result sequence, wherein the rough delimitation result sequence includes a plurality of suspected start boundaries arranged in a sequential order;

[0040] S103: Treat each suspected starting boundary as a node on the path, start from the first suspected starting boundary, and use the natural redundancy of the protocol to construct a path metric for path search; Figure 3 As shown, the path search process includes:

[0041] For the current suspected starting boundary g b , g b Set as the downstream node, which will meet the packet length constraint and be located at g b All previous suspected starting boundaries are taken as g b Upstream node g a ;

[0042] For each upstream node g a , assuming g a to g b There is only g between a As the starting boundary of the group, calculate the distance from g1 to g a Then to g b The cumulative path metric of the constructed paths based on the natural redundancy of the protocol;

[0043] The path with the largest cumulative metric is taken as the current best path, and the search continues for the next suspected starting boundary until the last suspected starting boundary is found.

[0044] S104: After the search is completed, the current best path is output. The sequence formed by all nodes on the current best path arranged in order is the final group boundary sequence.

[0045] Specifically, for the downstream node g b , by the upstream node g a Jump to it and determine the packet Pkt a→b This event is equivalent to identifying node g a is the starting boundary of a potential group, and g a to g b The other nodes between them are not the starting boundaries. The branch metric defined in the literature [2] only uses the former information and ignores the latter, which results in the segmentation performance under the path metric criterion not being able to achieve the expected effect. In order to improve the segmentation performance, the embodiment of the present invention considers the above two pieces of information at the same time when calculating the path cumulative metric, that is, in "g a to g b There is only g between a The calculation from g1 to g is based on the condition of "starting boundary of grouping" a Then to g b The cumulative path metric of the constructed paths.

[0046] Furthermore, as an implementation method, in the embodiment of the present invention, when performing coarse delimitation, the correlation coefficient between the fixed field and the fixed field observation vector is used as a similarity measurement indicator between the two. Correspondingly, the above-mentioned step S102 specifically includes: calculating the correlation coefficient between the fixed field and the current fixed field observation vector, as shown in formula (1); when the correlation coefficient is greater than a set threshold, the position of the last bit of the current fixed field observation vector is regarded as a suspected starting boundary and added to the coarse delimitation result sequence.

[0047]

[0048] Among them, k represents a fixed field, Represents the current fixed field observation vector, the numerator represents the inner product of two vectors, and the denominator represents the product of their respective two norms.

[0049] Specifically, because the length of each packet is an integer multiple of the byte size, that is, the actual packet boundary is at an integer multiple of the byte position, so when estimating the packet boundary, starting from the first bit, it is only necessary to slide the relevant matching pointer on the received sequence y in units of bytes to obtain the current fixed field observation vector

[0050] Assume that m possible starting boundaries {g1,g2,...,g m}, on this basis, add the termination boundary of the aggregation group sequence Finally, we get the rough delimited result sequence g={g1,g2,...,g m ,g m+1}.

[0051] Example 2

[0052] Based on the above embodiment 1, as an implementation method, the embodiment of the present invention uses the likelihood of the node to calculate the path cumulative metric. On this basis, for each upstream node g a , assuming g a to g b There is only g between a As the starting boundary of the group, calculate the distance from g1 to g a Then to g b The cumulative path metrics of the constructed paths based on the natural redundancy of the protocol, specifically including:

[0053] Using g a to g b All nodes between (except g b The single-step metric is defined based on the likelihood of the packet (which will be calculated later when it serves as an upstream node). In other words, the present invention no longer calculates likelihood per packet but per node. The new single-step metric based on the natural redundancy of the protocol is defined as follows:

[0054]

[0055] in, k is a fixed field, u a Pkt a→b The length field is calculated, and its value is determined by g b -g a Size determination, R k and R u The fixed field k and length field u are a A random field with equal probability of 0 or 1 and the same dimension.

[0056] From the above formula (2), it can be seen that when g aWhen it is regarded as the starting boundary, the likelihood calculation is the same as that in reference [2]. When other nodes are regarded as non-starting boundaries, the data sent at the corresponding positions are treated as random fields with equal probability of 0 / 1.

[0057] by Taking the likelihood at the field as an example, there are:

[0058]

[0059] Among them, l k is a fixed field length. For the bit-level likelihood information within the field,

[0060]

[0061] Among them, “≡” means normalizing each likelihood on the right side of the above formula, that is, uniformly subtracting After that, it does not affect the ratio and measurement properties between each other.

[0062] Similarly, the length field metric can be calculated and the complete single-step metric Furthermore, since the single-step metric proposed in this paper takes into account the upstream node g a To downstream node g b The likelihood information corresponding to the positions of all roughly delimited nodes between , so the cumulative metric of the path can be defined as the sum of all single-step metrics on the path. Then the new cumulative metric satisfies the following recursive equation:

[0063]

[0064] Among them, M 1→a From node g1 to node g a The cumulative path metric, The path continues to extend to node g b The cumulative metric of the new path formed.

[0065] Thus, when measuring the quality of different paths between any two nodes, even if the number of steps is different, they can still be directly compared because the number of times the field likelihood is calculated is the same. Find the path to node g according to the following formula b The path with the largest cumulative metric among all the paths is recorded as the node reaching g b The best path, and can be used to complete the delimitation and segmentation of the group

[0066]

[0067] Based on the above calculation method and embodiment, the specific details of the packet fault tolerance delimitation method based on protocol natural redundancy provided by the embodiment of the present invention are shown in the following Algorithm 1.

[0068]

[0069]

[0070] Example 3

[0071] Although the repeated fields r of different packets are unknown, they have the same value. If the repeated fields of N frames can be jointly demodulated, it is equivalent to improving the signal-to-noise ratio by 10lg(N)dB, and the field demodulation error rate will be greatly reduced. In the packet delimitation problem, since the packet boundaries have not yet been determined, the repeated fields of different packets are difficult to be accurately extracted and directly used or added to the path search metric. Therefore, when there are repeated fields r in the protocol stack, the embodiment of the present invention proposes a repeated field decision feedback mechanism to improve the performance of packet delimitation. Consider first using path search to complete the first round of delimitation, and then extracting the repeated fields for use. When the accuracy of the first round of delimitation is high and the number of packets is large enough, the demodulation error rate of the repeated fields will be low enough. The hard decision results of the repeated fields are then fed back to the packet delimitation task, added to the fixed field sequence, and the path search is performed here to complete the second round of delimitation, in order to improve the final delimitation performance. That is, based on the above embodiments, the embodiment of the present invention also includes the following steps:

[0072] If the redundant field types in the protocol stack used for communication between the transmitting and receiving ends include repeated fields, then based on the packet boundary sequence output in step S104, repeated field observation vectors of each packet are extracted to form a repeated field observation vector sequence;

[0073] Estimating a hard decision result of the repeated field according to the repeated field observation vector sequence;

[0074] Add the hard decision result of the repeated field to the original fixed field sequence to obtain a new fixed field sequence;

[0075] The path search is performed again according to the new fixed field sequence, and the group boundary sequence obtained by this path search is used as the final group boundary sequence.

[0076] The specific details of the packet fault-tolerant delimitation method based on protocol natural redundancy in an embodiment of the present invention are shown in Algorithm 2.

[0077]

[0078]

[0079] Example 4

[0080] The embodiment of the present invention takes the frame aggregation and segmentation involved in the HDLC protocol widely used in the field of satellite communications as an example to further illustrate the technical solution of the present invention and verify its effectiveness. The embodiment of the present invention simplifies the receiving condition into a one-way transmission scenario between two satellite transponders. Since it is impossible to distinguish the upper layer protocol type in each packet before the packet delimitation is completed, the data is generated with reference to the IPv4 / MAC / HDLC protocol stack format in this experiment, that is, only the HDLC is considered to continue to encapsulate the MAC header and the IPv4 protocol packet of variable length. At the same time, it is assumed that the IP layer does not fragment, does not use differentiated services, and the IP header length is fixed at 20 bytes. The IPv4 packet length is assumed to obey a bimodal distribution model, with the maximum and minimum values ​​set to 1500 and 40 bytes respectively. At the same time, the number of packets η in a single aggregated packet is taken as 20, and the relevant threshold is set to ρ th =0.3. Based on the above assumptions, the three available redundant protocol fields are summarized as follows:

[0081] Fixed field k (61 bits in total): IPv4 version number, header length and differentiated services field (0x4500), flag and fragment offset field (0x4000), MAC protocol number field (0x0800), HDLC flag field (0x7E), and the upper 5 bits of the IP packet total length field (0b00000);

[0082] Length field u (11 bits in total): the lower 11 bits of the total length field of the IP packet;

[0083] Repeating field r (104 bits in total): HDLC address field (8 bits), MAC address field (96 bits).

[0084] To verify the performance of the technical solution of the present invention, the packet delimitation error rates of different algorithms were compared. In the present invention, correct packet delimitation is defined as requiring that both the start and end boundaries are accurately found. That is, for any transmitted packet in an aggregated packet, if its start and end positions cannot be uniquely determined by two adjacent elements in the predicted boundary sequence, the packet is considered to have been incorrectly delimited.

[0085] Figure 4 The comparison algorithms include: the path search algorithm in reference [2] (the Hamming distance threshold is set to 20), the BCJR method in reference [1] with the related coarse delimiter and assuming that the number of packets in a single aggregate group is known, further taking the prior information of packet length distribution into consideration in the BCJR algorithm, the method of the present invention (without repeated field decision feedback), the method of the present invention (with repeated field decision feedback) and the method of the present invention (assuming that the repeated field is known, that is, directly treating the repeated field as a fixed field).

[0086] from Figure 4As can be seen, the method of the present application greatly improves the performance of the original path search method in document [2] due to the improved path metric, and can approximate the BCJR algorithm under the condition that the number of data packets (or groups) is known, and the gap to the BCJR algorithm further using the prior information of the data packet length distribution is also smaller, only about 0.4 dB. When considering that there is a redundant available repetition field in the protocol stack, the decision feedback mechanism makes the algorithm performance further greatly improved, and approximates the algorithm performance under the condition that the repetition field is known.

[0087] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A packet fault-tolerant delimitation method based on natural redundancy of a protocol, characterized in that: include: Step 1: Receive an aggregated packet sequence; Step 2: performing similarity matching on the fixed field and all fixed field observation vectors on the aggregated packet sequence to obtain a rough delimitation result sequence, wherein the rough delimitation result sequence includes multiple suspected start boundaries arranged in order; Step 3: Treat each suspected starting edge as a node on the path, start from the first suspected starting edge g1, and use the natural redundancy of the protocol to construct a path metric for path search. The path search process includes: For the current suspected starting boundary g b , g b Set as the downstream node, which will meet the packet length constraint and be located at g b All previous suspected starting boundaries are taken as g b Upstream node g a ; For each upstream node g a , assuming g a to g b There is only g between a As the starting boundary of the group, calculate the distance from g1 to g a Then to g b The cumulative path metrics of the constructed paths based on the natural redundancy of the protocol, specifically including: in, k is a fixed field, u is a length field, u a Pkt a→b The length field is calculated, and its value is determined by g b -g a Size determination, R k and R u The fixed field k and length field u are a Random fields of equal probability 0 / 1 with the same dimension, is a single-step metric, M 1→a From node g1 to node g a The cumulative path metric of y k and y u are the observation information corresponding to the fixed field k and the length field u respectively; The path with the largest cumulative metric is taken as the current best path, and the search continues for the next suspected starting boundary until the last suspected starting boundary is found. Step 4: After the search is completed, the current best path is output as the final group boundary sequence.

2. A method for packet fault tolerance delimitation based on protocol natural redundancy according to claim 1, characterized in that: Step 2 specifically includes: The correlation coefficient between the fixed field and the current fixed field observation vector is calculated. When the correlation coefficient is greater than the set threshold, the position of the last bit of the current fixed field observation vector is taken as a suspected starting boundary and added to the coarse delimitation result sequence.

3. The method for packet fault tolerance delimitation based on protocol natural redundancy according to claim 1, characterized in that: Also includes: If the redundant field types in the protocol stack used by the transmitting and receiving ends include repeated fields, then based on the packet boundary sequence output in step 4, the repeated field observation vectors of each packet are extracted to form a repeated field observation vector sequence; Estimating a hard decision result of the repeated field according to the repeated field observation vector sequence; Add the hard decision result of the repeated field to the original fixed field sequence to obtain a new fixed field sequence; The path search is performed again according to the new fixed field sequence, and the group boundary sequence obtained by this path search is used as the final group boundary sequence.

Images