CN112788645A

CN112788645A - Distributed neighbor node distribution estimation method based on adaptive compressed sensing

Info

Publication number: CN112788645A
Application number: CN202110197393.XA
Authority: CN
Inventors: 朱弘恣; 蔡云翔; 王潇; 常姗; 沈剑钢; 过敏意
Original assignee: Shanghai Jiaotong University
Current assignee: Shanghai Jiaotong University
Priority date: 2021-02-22
Filing date: 2021-02-22
Publication date: 2021-05-11
Anticipated expiration: 2041-02-22
Also published as: CN112788645B

Abstract

A distributed neighbor node distribution estimation method based on adaptive compressed sensing comprises the steps that firstly, all vehicle nodes use synchronous network time as seeds, a random transmitting power sequence is generated by a pseudo-random number generation algorithm and randomly selected to be a transmitting end or a receiving end, and identity information subjected to Hash processing is collected to conduct neighbor node distribution estimation; the vehicle node serving as a transmitting end constructs a local bitmap, and the bitmap containing the identity information of the vehicle node is transmitted to a receiving end in an OFDM mode through OOK modulation; the vehicle node serving as a receiving end receives and demodulates the OFDM signal superposed in the air, the process is repeated for multiple rounds to obtain enough information, a bitmap containing identity information is obtained, the number of the vehicle nodes under the current communication radius is estimated, and then the distribution condition of neighbor nodes in the whole communication radius range is recovered through a self-adaptive compressed sensing method.

Description

Distributed neighbor node distribution estimation method based on adaptive compressed sensing

Technical Field

The invention relates to a technology in the field of wireless communication, in particular to a distributed neighbor node distribution estimation method based on adaptive compressed sensing.

Background

The Internet of vehicles is a novel wireless mobile self-organizing network, vehicles, roads and people which are originally isolated from each other are organically cooperated through message broadcasting, safety and danger early warning of traveling are achieved, services such as vehicle-mounted navigation and entertainment are provided, and safety and comfort of drivers and passengers are improved. The existing car networking technology can be mainly divided into a method for installing additional equipment and a method for not installing additional equipment. The method for adding additional equipment needs other equipment besides the basic unit for vehicle networking communication, for example, the method based on computer vision needs to rely on a camera and a monitoring system, and the vision-based mode has poor effect at night or in poor weather conditions; besides, there are methods based on a communication base station or a roadside communication unit RSU, methods based on a self-induction coil, methods based on a wireless vehicle sensor, methods based on acoustics, methods based on ETC toll station statistics, and the like. These methods require additional installation of some equipment and have the disadvantages of small coverage, inaccurate measurements, high laying or maintenance costs, etc. Another broad category of ways that do not rely on additional devices is often not suitable for different vehicle distribution states when the spatial distribution of the vehicles has some a priori knowledge, or occupies a large amount of communication resources to accomplish the grouping or aggregation.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides a distributed neighbor node distribution estimation method based on self-adaptive compressed sensing, which utilizes the sparsity of the distribution of vehicles in space, recovers the distribution of all neighbor nodes in the maximum communication radius range by randomly measuring a small number of radii in the maximum communication radius range and greatly reduces the measurement cost.

The invention is realized by the following technical scheme:

the invention relates to a distributed neighbor node distribution estimation method, firstly all vehicle nodes use synchronous network time as seeds, generate random transmitting power sequences by a pseudo-random number generation algorithm, randomly select the random transmitting power sequences to be a transmitting end or a receiving end, and collect identity information subjected to hash processing to carry out neighbor node distribution estimation; the vehicle node serving as a transmitting end constructs a local bitmap, and the bitmap containing the identity information of the vehicle node is transmitted to a receiving end in an OFDM mode through OOK modulation; the vehicle node serving as a receiving end receives and demodulates the OFDM signal superposed in the air, the process is repeated for multiple rounds to obtain enough information, a bitmap containing identity information is obtained, the number of the vehicle nodes under the current communication radius is estimated, and then the distribution condition of neighbor nodes in the whole communication radius range is recovered through a self-adaptive compressed sensing method.

The transmission to the receiving end is carried out in an OFDM mode and is realized by adopting broadcasting, non-unicast or multicast.

The random transmitting power sequence is as follows: all vehicle nodes have synchronized network time, randomly generate random numbers n by using the time as a seed, change the random numbers into serial numbers 0 to n-1 by using a modulus operation i ═ rand ()% n, and correspond to communication radiuses { R [₀，R₁，…，R_n-1The transmit power sequence of { P }₀，P₁，…，P_n-1Selecting corresponding transmitting power P_i。

The identity information adopts but is not limited to: a MAC address, or license plate number or frame number, etc. can indicate a unique identifier for the identity of the vehicle node.

The random role selection specific operation is as follows: in order to communicate from radius R, due to possible lack of central facilities in the network and the devices being half-duplex communication_iObtain enough neighbor nodes inThe method using probability solves the problem. That is, the node randomly generates a random number, and becomes the receiving end with a probability of p and becomes the transmitting end with a probability of 1-p based on the random number. When v becomes the transmitting end, the identity information is used as the transmitting power P_iBroadcasting to surrounding neighbor nodes; when v becomes the receiving end, it receives the radius R_iInformation sent by the inner neighbors.

The local bitmap comprises: a bloom filter field of m bits as carrying information and an indication field of one bit.

The constructing of the local bitmap is as follows: when the vehicle becomes a transmitting end, the identity information is loaded into the OFDM symbols for broadcasting, and other nodes at the receiving end are helped to construct a bit map based on the bloom filter. Initially, all bit maps are set to 0, then the transmitting end node processes its own identity information through a hash function (maps other information to a number) to map the identity information to a number, and the number is used as a sequence number to set a certain position of a 01 sequence to 1. For example, the initial state bitmap is 000 … 000 (48 0 s), the value obtained by hash-out processing of the MAC address is 10, the 10 th 0 is set to 1, and the obtained 01 sequence can be used to represent the identity information. The l bits indicate that the information of the field is used as identification bits of the adaptive compressed sensing process.

The OFDM mode is as follows: the information of the local bitmap is embedded into the subcarrier of the OFDM symbol by OOK modulation, bit 1 and bit 0 are respectively mapped to (1, 0) and (0, 0) in the constellation diagram, and the signal processing flow of the transmitting end is compatible with the 802.11p protocol.

Specifically, the 802.11 protocol has 64 sub-carriers, wherein the sequence numbers are [ -26: -22, -20: -8, -6: -1,1:6,8:20,22:26]The remaining 4 bits are used as pilot frequency, i.e. identification bits of the adaptive compressed sensing process, and the remaining 11 bits are set to 0 as channel estimation, so as to determine a threshold value for distinguishing 01. When the length of the local bitmap is greater than 48, the local bitmap needs to be carried using multiple OFDM symbols. The vehicle node at the transmitting end generates OFDM symbols according to the currently selected power level P_iLaunching。

The estimation of the number of the vehicle nodes under the current communication radius refers to: when the vehicle node is selected to become a receiving end, an OFDM receiving process is used, the number of 1 in the received bitmap is calculated after the bitmap sent by the surrounding nodes is collected and overlapped, and the current neighbor number obtained by estimation is calculated by adopting an accumulation-based method.

The accumulation-based method is as follows: number of current neighbors

Wherein:

to transmit power P_iCorresponding communication radius R_iThe number of neighbor nodes is estimated, m is the number of bits of the bitmap, c (Z) is the number of 0's therein, k is the number of hash functions used,

is the estimated proportion of the number of neighbor nodes currently acquiring information.

The self-adaptive compressed sensing method comprises the following steps: the method for estimating the distribution of the neighbor nodes by utilizing the sparse characteristic of the distribution of the vehicle nodes and using a compressed sensing method specifically comprises the following steps:

1) determining bases for conversion to sparse representation: obtaining a sparse representation of the signal using a differential transformation basis, treating the distribution of neighboring nodes in space as a series of doublets { (P)₀，N(P₀))，(P₁，N(P₁))，…，(P_n-1，N(P_n-1) In which P is present)₀＜P₁＜…＜P_n-1(ii) a The vector of N is converted into a sparse vector S by a basis transformation, i.e. S- Ψ N or N- Ψ^-1S, where Ψ is an n × n differential transformation base, and Ψ^-1Is a lower triangular matrix, i.e.:

for example, whenWhen N is (2, 2, 6), S is (2, 0, 4).

2) Constructing a measurement matrix: the measurement matrix Φ is a matrix of dimensions m × n, where m is equal to the number of measurements, each row of which is associated with a size P₀To P_n-1Corresponds to the transmit power (or current radius) from which each vehicle randomly derives a transmit power P based on the current synchronized network time_i∈{P₀，P₁，…，P_n-1Selecting one of the nodes and finishing the estimation of the number of the nodes of the vehicle under the radius when P is_jIs selected in the ith measurement, then_ijIs set to 1. Let Y represent the result of m measurements, then: Y-N- Ψ^-1S, because of the non-negligible error in the measurement of the number of nodes within the radius R, each measurement is made with a non-negligible error

All have noise η, so that Y ═ Φ N + η ═ Φ Ψ^-1s+η。

3) Recovering signals distributed by neighbor nodes: the problem of recovering the distributed signal of the neighbor node can be translated into the following l₀And (3) optimization problem:

s.t.y ═ AS, where: y and a ═ Φ Ψ^-1Are known; the optimization problem can be solved by, but not limited to, algorithms such as Orthogonal Matching Pursuit (OMP), compressed sample matching pursuit (CoSaMP), and the like.

Preferably, the signal with the sparsity K can be reconstructed through m times of measurement, and when m is more than or equal to b.mu²(Φ, Ψ) · K · logn, wherein: b is a normal number and μ (Φ, Ψ) is the correlation between the measurement matrix Φ and the sparse representation basis Ψ. This correlation is defined as

It can be seen that the smaller the correlation between the measurement matrix Φ and the sparse representation base Ψ, the fewer measurements are required to recover the signal. Since the measurement matrix Φ is generated randomly, the measurement matrix Φ is fixedThe correlation between the sparse representation bases Ψ is very small. So satisfying m 3K to 4K can restore the signal perfectly.

4) Carrying out self-adaptive adjustment according to the sparsity: different vehicle nodes have different neighbor node distributions, and the required measurement times are different according to different sparsity of the surrounding neighbor nodes: after m measurements, the vehicle arranges the m measurements in order of the magnitude of the transmission power (or communication radius). Since the measurement result under the communication radius has non-negligible error, M is first subjected to full variation regularization (TV-regularization). The noise reduced z is then used for signal recovery and is recorded as

When the vehicle node continues to perform z +1 th measurement, the distribution estimation can be obtained

The vehicle may calculate an indicator characterizing the discrepancy between the two distribution estimates

When the indicated quantity is less than the threshold e, the neighbor node distribution estimation process for the vehicle node may terminate. In this case, the vehicle clears the i-bit indication field portion of the local bitmap (sets it to 0) in future broadcast procedures, but still performs the role selection procedure. Until no signal indicating the domain exists in the received signals, that is, all the neighbor nodes complete the neighbor node estimation process, no more measurement information is needed.

Technical effects

The invention integrally solves the defects of the prior art that extra equipment needs to be counterfeited, the vehicle distribution mode needs to be matched, the consumed communication resource is large, the response is slow and the cost is high.

Compared with the prior art, the method and the device can be compatible with the existing equipment, can efficiently and quickly carry out node distribution estimation, and can carry out self-adaptive adjustment according to the sparsity. The method does not depend on the prior knowledge of the vehicle node distribution form, does not need to be additionally provided with additional hardware equipment such as a camera and a laser radar, and has the advantages of high estimation precision, low time delay, quick response and less communication resource consumption.

Drawings

FIG. 1 is a schematic diagram of the principles of the present invention;

FIG. 2 is a bitmap of an embodiment in which a transmitting node hashes its own identity information and inserts it into the bitmap;

FIG. 3 is a schematic diagram of signal processing of a transmitting node in an embodiment;

FIG. 4 is a diagram illustrating signal processing at a receiving node in an embodiment;

FIG. 5 shows an embodiment in which the receiving module solves the bitmap by thresholding 01;

FIG. 6 is a schematic diagram of an overlay bitmap obtained by overlaying local bitmaps of neighbor nodes in the embodiment in the air;

FIG. 7 is a schematic diagram of a series of overlay bitmaps synthesized by a logical OR operation in an embodiment;

FIG. 8 is a schematic diagram illustrating a neighbor node distribution solution by compressed sensing in an embodiment;

FIG. 9 is a schematic photograph of the prototype setup in the example;

FIG. 10 shows the test results of the prototype in the example;

FIG. 11 is a diagram illustrating simulation results of an embodiment of the present disclosure for investigating the influence of different OFDM symbol numbers on the estimation accuracy of the number of neighboring nodes under a certain radius;

FIG. 12 is a simulation result diagram for studying the influence of the number of different hash functions on the estimation accuracy of the number of neighbor nodes under a certain radius in the embodiment;

FIG. 13 is a simulation result diagram for investigating the influence of different neighbor node densities on the neighbor point distribution estimation accuracy in the embodiment;

FIG. 14 is a CDF graph illustrating the accuracy of various methods under different statistical distributions in an example;

FIG. 15 is a CDF chart showing the number of measurements in different distributions in the example;

FIG. 16 is a diagram illustrating recovery effects of different methods according to an embodiment.

Detailed Description

As shown in fig. 1, the present embodiment relates to a distributed neighbor node distribution estimation system for implementing the method, including: random transmission power and role selection module, transmission module, receiving module and self-adapting neighbor distribution estimation module, wherein: and the random transmitting power and role selection module generates a random transmitting power sequence according to the current synchronous network time and randomly distributes roles to all nodes. The transmitting module carries out Hash processing according to the identity information of the transmitting module, constructs a local bitmap and transmits the bitmap, and the receiving module carries out decoding according to the collected overlapped OFDM symbols to obtain the overlapped bitmap, so that the number of neighbor nodes under the current communication radius is estimated. The self-adaptive neighbor distribution estimation module estimates the neighbor node distribution through a compressed sensing algorithm according to a series of collected neighbor node number information under different radiuses, and the process is self-adaptively terminated under the influence of neighbor node density.

As shown in fig. 2, the random transmit power and role selection module constructs a local bitmap by: the deadline initializes an all 0-bit bitmap, where m bits are used as bloom filter data bits and l bits are used as an indication field for adaptive compressed sensing termination. The transmitting end node processes the identity information (such as the MAC address) through a plurality of hash functions to map to certain bits on the bitmap of bits. The number of mapped bits is 1. The resulting sequence containing 01 is a bitmap containing identity information.

As shown in fig. 3, the transmitting module is compatible with the common 802.11 protocol. Mapping the local bitmap onto a star map through an OOK modulation module to obtain a complex signal; then converting the time domain signal into a frequency domain signal through an inverse Fourier transform (IFFT) module; then adding a guard interval module to add a cyclic prefix or a postfix shape; and finally, transmitting the upper carrier through hardware after beamforming and IQ sampling.

As shown in fig. 4, the receiving module is compatible with the 802.11 protocol, and the reverse of the transmitting process described above. Firstly, a high-frequency signal is changed into a baseband signal by a lower carrier, and then IQ sampling is carried out, then a guard interval is removed, OOK decoding is carried out, and a superposed bitmap is obtained.

As shown in fig. 5, the existing 802.11p protocol specifies 64 sub-carriers, wherein 48 sub-carriers are used to transmit bloom filter data fields, 4 sub-carriers are pilot fields used to transmit adaptive compressed sensing, and 12 sub-carriers are set to 0. The zeroed 12 subcarriers can just be used as a threshold to decide on a 01 boundary. That is, the threshold of 0 (for example, the largest module may be taken as the threshold, that is, the circle with black being centered at 0 in the figure, and the threshold is used to determine 01 in the data field, that is, greater than the threshold is 1 (the circle in the figure), and less than the threshold is 0 (the rectangular frame in the figure) according to the size of 12 0-set subcarriers (the solid circle in the figure).

As shown in fig. 6, the superposition is done over the air for all bit maps. And the three vehicle nodes respectively construct and send respective local bitmaps. And the node at the receiving end receives the superposed bitmap which is obtained by logical OR operation of the bitmap of each neighbor node. This process may introduce errors since the channel is not a perfect channel.

As shown in fig. 7, for each time the superimposed bitmap collected by the process of fig. 6 only contains information of a part of neighbor nodes, the vehicle node superimposes a series of bitmaps collected by itself through a logical or operation to obtain enough information of the neighbor nodes. And estimating the number of the neighbor nodes under the communication radius according to the number of 1 in the integrated superposed bitmap.

As shown in FIG. 8, the measurement matrix Φ is a matrix of dimension m n, where m is equal to the number of measurements, each row of which is associated with a dimension P₀To P_n-1Corresponds to the transmit power (or current radius). Each vehicle randomly derives a transmission power P from the current synchronized network time_i∈{P₀，P₁，…，P_n-1And selecting one of the nodes and finishing the estimation of the number of the nodes of the vehicle under the radius. If P is_jIs selected in the ith measurement, then_ijIs set to 1. Let Y denote the result of m measurements, we have Y ═ Φ N ═ Φ Ψ^-1And S. Because at the radiusThe measurement of the number of nodes in the range of R has non-negligible error, and each measurement has one error

There is noise η, so Y can be expressed as: n + η phi psi^-1And S + eta. The problem of recovering the distributed signal of the neighbor node can be translated into the following l₀Optimization problem

s.t.y ═ AS, where Y and a ═ Φ Ψ^-1Are known. The optimization problem can be solved by algorithms such as Orthogonal Matching Pursuit (OMP), compressed sample matching pursuit (CoSaMP), and the like. In addition, there are many algorithms for solving the optimization problem, which can be used to solve the problem.

The prototype validation is specifically achieved in this example by: the prototype of the system was implemented using 4 USRP N210s and a PC computer. Each USRP uses the CBX daughter board and is clocked with the GPSDO module (synchronization error 0.01 ppm). The PC computer runs an Ubuntu16.04 LTS 64bit operating system and is provided with an Intel i7-8700CPU and a 16GB memory. All four USRPs are divided into two groups and connected to PC computers using 1Gbps ethernet for better transmission. The software for signal processing uses GNU Radio.

As shown in fig. 9, 3 USRPs serve as transmitting terminals and 1 USRP serves as receiving terminals. The new center frequency is 5.89GHz and the bandwidth is 10 MHz. The sampling rate of the USRP is set to 10M/s, i.e., since each sample point is a complex number of 8 bits, the data acquisition efficiency is 80 Mbps. In order to verify the feasibility of solving the superposition bitmap, the present embodiment performs simulation (emulation) on the channels for vehicle-to-vehicle communication, and uses 7 typical channels as shown in the following table.

These channels are based on the standard Tapped Delay Line (TDL) channel model, and contain 7 typical scenarios for vehicle communication. For each vehicle channel, one million bits are transmitted per transmitting end. The receiving end decodes the superposed OFDM symbols after receiving the signals, and calculates the 0 bit error rate, the 1 bit error rate and the total bit error rate. Fig. 10 shows experimental results at different signal-to-noise ratios. From this experimental result, the following experience can be obtained in this example: firstly, based on the existing decoding method, the 1 bit error rate is usually higher than the 0 bit error rate, and secondly, the 1 bit error rate is sharply reduced and the 0 bit error rate is rapidly increased along with the improvement of the signal-to-noise ratio. The reason for this is that the decision threshold becomes smaller as the signal-to-noise ratio rises, which means that it is easier to have 0 solved to 1. thirdly, the total error rate is below 2.5% even in the worst case of channel conditions. This embodiment will not improve the solution method in the future to expect better performance under different channel conditions.

In the embodiment, a large number of simulation experiments are carried out to test the performance of the method under different vehicle distribution conditions and different channel states. Specifically, the present embodiment randomly generates a distribution of neighboring nodes (with the common three distributions, uniform distribution, gaussian distribution, and poisson distribution). For the medium type of distribution, the present embodiment changes the number of all vehicles from 20 to 100.

In this embodiment, different channel states are represented by different bit error rates, and the accuracy index includes: a Neighbor Number Estimation Error Rate (NEER) and a Neighbor Distribution Estimation Error (NDEE).

The estimated error rate of the number of neighbors is as follows: under a certain communication radius, the ratio of the estimated absolute error of the neighbor node number to the real result

The neighbor distribution estimation error is as follows: the difference between the result of the neighbor node distribution estimation and the real distribution situation is represented by the RMS error of the two signals,

when the superposed bitmap is solved, the influence of the failed parameters on the indexes is verified, two compressed sensing recovery algorithms of CoSaMP and OMP are used, and nearest neighbor interpolation and linear interpolation are used for comparison.

The nearest neighbor interpolation means that: nearest neighbor interpolation estimates unknown points using nearest known points.

The linear interpolation refers to: linear interpolation estimates an unknown point using the value of the line between the nearest two points.

The embodiment specifically evaluates the accuracy of the distribution condition of the neighbor nodes in the whole communication radius range by the adaptive compressed sensing method through the following steps:

1) effect of bitmap length: the embodiment randomly generates the neighbor nodes in uniform distribution, the number of the nodes is from 20 to 100, and the step value is 20. In this embodiment, the number of times of the role selection process is set to 9, and the number of hash functions used is set to 2. in this embodiment, the length of the local bitmap is changed from 1(48 bits) to 4(192 bits). NEER was studied for its performance in 6 different channel states (BER error rates) of 0.1%, 0.25, 0.5%, 1%, 2.5% and 5%. For each experimental setup, this example repeated the experiment 500 times to observe statistical results.

As shown in fig. 11(a) to (f), the results of experiments of NEER in different channel states are shown. The abscissa is the number of OFDM symbols. As can be seen from the experimental results, NEER becomes smaller as the length of the bitmap increases. In addition to this, it can be seen that the singing bitmap can cope with a greater bit error rate (worse channel conditions) and denser vehicle distribution. In practical applications, applications are very sensitive to response time, and different upper layer applications have different response time requirements. The strategy for selecting the length of the bitmap is to select as many OFDM symbols (i.e. larger bitmaps) as possible on the basis of satisfying the response time of the upper layer application so as to reduce the error of the data estimation of the neighbor nodes.

2) Influence of the number of hash functions: the experiment is similar to the above experiment parameters except that the number of OFDM symbols is fixed to 2 (i.e. the length of the bitmap is fixed to 96). the number of hash functions is changed to 1, 3, 5, 7, resulting in NEER in six channel states. Each channel state experiment was still repeated 500 times.

As shown in fig. 12(a) to (f), the results of experiments of NEER in different channel states are shown. The abscissa is the number of hash functions. It can be seen that as the hash function increases, more channel errors can be tolerated, i.e. initially the NEER decreases as the hash function increases. On the other hand, more hash functions may overload the bitmap, resulting in increased error in the estimation result. The number of hash functions to be selected is based on the length of the bitmap.

a) influence of neighbor node density: based on the above experiment of estimating the number of neighbor nodes, the number of OFDM symbols selected in this embodiment is 2, the number of hash functions is 3, and the number of rounds of the role selection process is 9. In the embodiment, the controllable transmission power gear number is set to be 100, the termination threshold value is set to be 0.03, the uniformly distributed neighbor nodes are randomly generated, the node numbers are respectively 20, 40, 60, 80 and 100, and the performance of the N DEE under six channel states is researched. For each experimental setup, this example ran 500 replicates to observe statistical results.

Fig. 13(a) to (f) show the experimental results of the N DEE under different channel conditions. The abscissa is the number of neighbor nodes. It can be seen that all methods give stable results under normal conditions. The channel state has a large influence on the node distribution estimation.

b) Influence of vehicle distribution type: the experimental setup was similar to the above experiment. The difference lies in that three different node distribution forms are adopted, namely uniform distribution, Poisson distribution and Gaussian distribution. Specifically, vehicle nodes are generated in the above-described 3 distributions on four-lane roads having a length of 10 km. For even distribution, the probability of a vehicle appearing anywhere on the road is the same. For the gaussian distribution, the sum parameter is set to 5000 meters and 2500 meters, respectively. For a cedar distribution, the distribution parameter is set to 0.2. in a given distribution situation, this embodiment has all nodes estimate their own neighbor node distribution.

As shown in fig. 14(a) to (c), a Cumulative Density Function (CDF) is obtained for all vehicles in three different node distributions. It can be seen that the PeerProbe has the best effect.

As shown in fig. 15 and 16, the CDF is a graph of the number of measurements required under different distribution conditions. It can be seen that more than ninety percent of the vehicle nodes need only not exceed 40 measurements at different radii.

The invention has the technical effects that:

firstly, the node under the distributed self-organizing condition can efficiently and accurately estimate the number of neighbor nodes under a certain communication radius. Due to the absence of a central facility (such as a base station or a road side communication unit (RSU)), the information transmission and reception of the nodes cannot be coordinated. To address this challenge, the present invention has all vehicles build Bloom filters (Bloom filters) in a distributed manner. Vehicles equipped with half-duplex communication devices randomly select either a transmit state or a receive state. The node in the sending state hashes (hash) the identity information of the node and inserts the identity information into the bit table (namely, the corresponding position is 1, and the rest positions are 0). And the vehicle node in the receiving state collects the distributed bit table sent by the neighbor node, and estimates the number of the neighbor node under the communication radius according to the number of 0 and 1 in the table. The process needs to be repeated for multiple rounds until all vehicle nodes collect enough information of different neighbor nodes, and the number of the neighbor nodes under the communication radius can be estimated with a high probability.

Second, the distributed bit table is efficiently and reliably transmitted between vehicles. Due to limited channel resources, it is not feasible that all nodes contend for the channel or transmit in turn by means of packet (packet). To solve this challenge, the present invention uses OOK modulation to embed the information of the bit table into OFDM symbols, where one subcarrier of each OFDM symbol conveys 1-bit information. In addition, vehicles in the transmitting state can transmit simultaneously, OFDM symbols are superposed in the air, and nodes in the receiving state can solve all the bit tables after the logic OR operation of the distributed bit tables and can still be used for estimating the number of neighbors.

Thirdly, highly dynamic variations of the distribution of the vehicle in space are dealt with. Due to the fact that distribution is sometimes sparse and sometimes dense, various different situations cannot be handled by the traditional compressed sensing method. In order to solve the problem, the invention provides a self-adaptive compressed sensing method, the measurement times can be self-adaptively adjusted according to the sparsity of the neighbor nodes, and in addition, the invention adds a Total Variation regularization (Total Variation) process before compressed sensing so as to improve the recovery precision.

The frame is a completely distributed frame, and the vehicle does not need to be additionally provided with any additional hardware equipment. According to the invention, through a large number of simulations, although a large error (even more than 10% may exist) may exist in the estimation of the number of nodes within a certain radius, the error can be well suppressed by the adaptive compressed sensing framework. Compared with the traditional compressed sensing framework, the framework needs fewer measurement times and can achieve higher estimation accuracy.

The foregoing embodiments may be modified in many different ways by those skilled in the art without departing from the spirit and scope of the invention, which is defined by the appended claims and all changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

Claims

1. A distributed neighbor node distribution estimation method is characterized in that firstly all vehicle nodes use synchronous network time as seeds, a random transmitting power sequence is generated by a pseudo-random number generation algorithm and randomly selected to become a transmitting end or a receiving end, and identity information subjected to Hash processing is collected to carry out neighbor node distribution estimation; the vehicle node serving as a transmitting end constructs a local bitmap, and the bitmap containing the identity information of the vehicle node is transmitted to a receiving end in an OFDM mode through OOK modulation; the vehicle node serving as a receiving end receives and demodulates the OFDM signal superposed in the air, the process is repeated for multiple rounds to obtain enough information, a bitmap containing identity information is obtained, the number of the vehicle nodes under the current communication radius is estimated, and then the distribution condition of neighbor nodes in the whole communication radius range is recovered through a self-adaptive compressed sensing method.

2. The distributed neighbor node distribution estimation method according to claim 1, wherein said random transmit power sequence is: all vehicle nodes have synchronized network time, randomly generate random numbers n by using the time as a seed, change the random numbers into serial numbers 0 to n-1 by using a modulus operation i ═ rand ()% n, and correspond to communication radiuses { R [₀，R₁，…，R_n-1The transmit power sequence of { P }₀，P₁，…，P_n-1Selecting corresponding transmitting power P_i。

3. The distributed neighbor node distribution estimation method according to claim 1, wherein the randomly selecting role specific operation is: the node randomly generates a random number, and becomes a receiving end with the probability of p and becomes a transmitting end with the probability of 1-p according to the random number; when v becomes the transmitting end, the identity information is used as the transmitting power P_iBroadcasting to surrounding neighbor nodes; when v becomes the receiving end, it receives the radius R_iInformation sent by the inner neighbors.

4. The distributed neighbor node distribution estimation method according to claim 1, wherein said constructing the local bitmap is: when the vehicle becomes a transmitting end, the identity information is loaded into the OFDM symbol for broadcasting, and other nodes at the receiving end are helped to construct a bit map based on the bloom filter; initially setting all bit bitmaps to 0, then enabling a transmitting end node to pass through a hash function on own identity information and map other information into a number, and setting a certain position of a 01 sequence to 1 by using the number as a serial number; the l bits indicate that the information of the field is used as identification bits of the adaptive compressed sensing process.

5. The distributed neighbor node distribution estimation method according to claim 1, wherein the estimation of the number of vehicle nodes under the current communication radius is: when the vehicle node is selected to become a receiving end, an OFDM receiving process is used, the number of 1 in the received bitmap is calculated after the bitmap sent by the surrounding nodes is collected and overlapped, and the current neighbor number obtained by estimation is calculated by adopting an accumulation-based method.

6. The distributed neighbor node distribution estimation method according to claim 1, wherein said accumulation-based method is: number of current neighbors

Wherein:

7. The distributed neighbor node distribution estimation method according to claim 1, wherein said adaptive compressed sensing method is: the method for estimating the distribution of the neighbor nodes by utilizing the sparse characteristic of the distribution of the vehicle nodes and using a compressed sensing method specifically comprises the following steps:

1) determining bases for conversion to sparse representation: obtaining a sparse representation of the signal using a differential transformation basis, treating the distribution of neighboring nodes in space as a series of doublets { (P)₀，N(P₀))，(P₁，N(P₁))，…，(P_n-1，N(P_n-1) In which P is present)₀＜P₁＜…＜P_n-1(ii) a The vector of N is converted into a sparse vector S by a basis transformation, i.e. S- Ψ N or N- Ψ^-1S, wherein: Ψ isAn n x n differential transformation base, and Ψ^-1Is a lower triangular matrix, i.e.:

2) constructing a measurement matrix: the measurement matrix Φ is a matrix of dimensions m × n, where m is equal to the number of measurements, each row of which is associated with a size P₀To P_n-1Corresponds to the current radius, each vehicle randomly derives from the transmission power P according to the current synchronized network time_i∈{P₀，P₁，…，P_n-1Selecting one of the nodes and finishing the estimation of the number of the nodes of the vehicle under the radius when P is_jIs selected in the ith measurement, then_ijIs set to 1; let r represent the result of m measurements, then: Y-N- Ψ^-1S, each measurement

All have noise η, so that Y ═ Φ N + η ═ Φ Ψ^-1S+η；

s.t.y ═ AS, where: y and a ═ Φ Ψ^-1Are known;

4) carrying out self-adaptive adjustment according to the sparsity: different vehicle nodes have different neighbor node distributions, and the required measurement times are different according to different sparsity of the surrounding neighbor nodes: after m times of measurement, the vehicle arranges the m times of measurement results according to the size sequence of the transmitting power or the communication radius; because the measurement result under the communication radius has non-negligible error, M carries on the regularization of the total variation first; the noise reduced z is then used for signal recovery and is recorded as

8. The distributed neighbor node distribution estimation method according to claim 7, wherein the signal with sparsity K can be reconstructed by m measurements when m ≧ b- μ²(Φ, Ψ) · K · logn, wherein: b is a normal number, μ (Φ, Ψ) is the correlation between the measurement matrix Φ and the sparse representation base Ψ; this correlation is defined as

9. The distributed method of estimating neighboring node distribution according to claim 7, wherein the distribution estimate is obtained when the vehicle node continues to perform z +1 th measurement

And when the indicated quantity is less than the threshold value epsilon, the estimation process of the distribution of the neighbor nodes of the vehicle node is terminated.