CN117377111A - Distributed multi-cell dynamic resource allocation method based on perception assistance - Google Patents

Abstract

The invention relates to the field of wireless communication resource allocation, in particular to a distributed multi-cell dynamic resource allocation method based on perception assistance, which realizes the performance of approaching to centralized multi-cell resource allocation by a scheme with low interaction quantity between stations and low channel estimation cost. The technical scheme includes that the method comprises the steps of estimating user kinematic parameters and CSI based on double-function radar signal echoes; determining a user set scheduled by the local station based on the estimated CSI; the base station interactively schedules the motion parameters of the user to estimate the CSI from the base station to the adjacent station user; constructing a distributed beam forming optimization problem model; and optimizing the beam forming by combining fractional programming with semi-positive relaxation and solving a beam forming vector. The method and the device are suitable for multi-cell resource allocation.

Description

Distributed multi-cell dynamic resource allocation method based on perception assistance

Technical Field

The invention relates to the field of wireless communication resource allocation, in particular to a distributed multi-cell dynamic resource allocation method based on perception assistance.

Background

MCRA (Multi-cell Resource Allocation ) jointly improves network spectral efficiency by efficiently designing user scheduling sets and collaborative precoding. Theoretically, all base stations can transmit CSI (Channel State Information ) and the like to the central processing unit through the wireless backhaul link, and at this time, the multi-cell network can be regarded as a huge mimo system to realize resource allocation in a centralized manner, so as to fully exert the performance advantage of MCRA. However, as the network size increases, the base station transmits all information back to the central processor, and it becomes impractical to cooperatively implement resource allocation, mainly for the following two reasons. First, the computational complexity of centralized signal processing is high, and a central server capable of achieving such a powerful computational capability may not be available; second, sharing global channel state information can result in a significant communication burden and high network latency in situations where backhaul bandwidth is limited.

To solve the above two challenges, the prior art and the like propose a Signal-to-interference-plus-noise ratio (slnr) Signal measurement scheme instead of the conventional SINR (Signal-to-interference-plus-noise ratio). In this scheme, each base station considers intra-cell interference and power leakage to users in other cells based only on CSI derived from itself, and uses a traffic model to implicitly characterize cell-to-cell user scheduling, thereby achieving multi-cell resource allocation in a distributed manner. However, due to the difference between the SLINR and SINR, and in real life, the traffic model is often irregular, which directly results in an ineffective resource allocation. In addition, in a dynamic network, the frequent channel estimation of all users in the system by each base station not only results in huge signaling overhead, but also the limited number of orthogonal pilots restricts the accuracy of the CSI estimation.

Disclosure of Invention

The invention aims to overcome the defects of the prior art, provides a distributed multi-cell dynamic resource allocation method based on perception assistance, and realizes the performance of approaching to centralized multi-cell resource allocation by a scheme with low interaction quantity between stations and low channel estimation cost.

The invention adopts the following technical scheme to achieve the aim, and provides a distributed multi-cell dynamic resource allocation method based on perception assistance, which comprises the following steps:

step 1, estimating user kinematic parameters and CSI (channel state information) based on double-function radar signal echoes;

step 2, determining a user set scheduled by the local station based on the estimated CSI;

step 3, the base station interactively schedules the motion parameters of the user to estimate the CSI from the local station to the adjacent station user;

step 4, constructing a distributed beam forming optimization problem model;

and 5, optimizing beam forming by combining fractional programming with semi-positive relaxation and solving a beam forming vector.

Further, the step 1 specifically includes:

each base station transmits a dual function radar signal to the user terminal using the perceptually integrated beam, estimates a kinematic parameter of the user based on the reflected signal of the user terminal, and constructs a corresponding next-time channel state vector.

Estimating a kinematic parameter of a user based on a reflected signal of the user side, and constructing a corresponding next-time channel state vector specifically includes:

at time n-1, the base station l,first, the reflected signals of different users are distinguished by using a receiving filter, and then the scheduled user k is estimated based on a matched filtering method _l Time delay of->And Doppler shift->Finally estimate angle->Obtaining a path loss coefficient predicted at the moment n according to a kinematic evolution model>Doppler shift->Angle->To construct base station l and service user k _l Direct channel between: />Wherein N is _t Representing the number of transmit antennas of the base station>Representing the set of users served by base station l.

Further, step 2 specifically includes: and constructing a corresponding channel state vector at the next moment according to the base station, and determining a user set scheduled by the base station at the next moment.

A greedy proportional fair zero-forcing scheme is adopted to select a user set scheduled at n moments, and the method specifically comprises the following steps:

step 21, selecting a group of continuous T _s -1 time unscheduled user Solving +.>Sum rate of users in (1)

Step 22, calculatingWherein->Is the historical average rate of the time n-1 before the user i, and is fed back to the service base station through uplink, wherein +.>And->

Step 23, if R is greater than or equal to R', thenRepeating steps 21 and 22 until R < R'; output->As the set of users scheduled by base station l at time n.

Further, the step 3 specifically includes: the base stations interact kinematic parameters of users to be scheduled of the base station through a backhaul network to assist the base station in estimating channel state vectors of the base station and other adjacent base station service users.

The interaction between the base stations through the backhaul network specifically comprises:

base station/determines according to two-dimensional coordinates thereofTwo-dimensional coordinates of the user->Movement speed +.>Consider backhaul networksIntrinsic delay T of the network _d The base station l is T before the next information transmission moment _d Information is interacted with other base stations, then the base station l estimates itself to schedule the user t to the other base station m _m Is a direct channel of: />

Further, the step 4 specifically includes: and the base station utilizes the estimated channel state vector to construct a virtual proportional fairness and rate maximization distributed beam forming optimization problem model with power limitation and perception error limitation.

The virtual proportional fair and rate maximization problem model is expressed as:

wherein,a precoding vector representing a time node n of a kth user in a cell l; />Representing a user scheduling set of the cell l at the time node n; p (P) _l Indicating the maximum transmit power of base station l; />Is user k _l The estimated error for i (including angle, delay and doppler shift) and is required to be less than a tolerable constant c; in addition, a->Is user k _l Virtual rate function at time n, +.>Is user k _l The average rate at the first n-1 time is used as a weight.

Further, optimizing beamforming and solving beamforming vectors by fractional programming in combination with semi-definite relaxation specifically includes: by introducing auxiliary variablesAnd->The objective function is equivalent to the one described above,

when (when)Fixing (I)>Then->

The semi-positive problem is expressed as:

where G is the matched filter gain,and->

In addition, in the case of the optical fiber,is the antenna gain of the base station, a _i Is a perceptual error constant;

and if the constraint condition of rank 1 is not considered, adopting eigenvalue decomposition to obtain the optimal beam forming vector.

The beneficial effects of the invention are as follows:

the invention provides a low-interaction and low-cost distributed dynamic multi-cell collaborative user scheduling and precoding design scheme, the interaction level is ultra-low and far lower than the CSI level, the feedback bandwidth is not influenced, and the scheme is easier to expand to a future large-scale heterogeneous network system. Meanwhile, based on the sensing capability of the base station, the motion parameters of the user are estimated, and a dynamic network channel is constructed, so that the estimation cost of the time-varying channel is reduced.

Compared with the scene without perception assistance, the proportional fairness and rate performance of the invention can be greatly improved, and the performance of centralized multi-cell resource allocation is approximated.

Drawings

Fig. 1 is a flow chart of distributed multi-cell dynamic resource allocation based on perception assistance according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a system model with integrated sense of general provided in an embodiment of the present invention;

FIG. 3 is a schematic diagram of a resource allocation framework provided by an embodiment of the present invention;

fig. 4 is a schematic performance diagram of resource allocation in a dynamic scenario according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

The invention provides a distributed multi-cell dynamic resource allocation method based on perception assistance, which is shown in fig. 1 and comprises the following steps:

step 1, estimating user kinematic parameters and CSI (channel state information) based on double-function radar signal echoes;

step 2, determining a user set scheduled by the local station based on the estimated CSI;

step 3, the base station interactively schedules the motion parameters of the user to estimate the CSI from the local station to the adjacent station user;

step 4, constructing a distributed beam forming optimization problem model;

and 5, optimizing beam forming by combining fractional programming with semi-positive relaxation and solving a beam forming vector.

In one embodiment of the present invention, step 1 specifically includes:

each base station transmits a dual function radar signal to the user terminal using the perceptually integrated beam, estimates a kinematic parameter of the user based on the reflected signal of the user terminal, and constructs a corresponding next-time channel state vector. The sense of general integrated system model is shown in fig. 2.

The step 2 specifically comprises the following steps: and constructing a corresponding channel state vector at the next moment according to the base station, and determining a user set scheduled by the base station at the next moment.

The step 3 specifically comprises the following steps: the base stations interact kinematic parameters of users to be scheduled of the base station through a backhaul network to assist the base station in estimating channel state vectors of the base station and other adjacent base station service users.

The step 4 specifically comprises the following steps: and the base station utilizes the estimated channel state vector to construct a virtual proportional fairness and rate maximization distributed beam forming optimization problem model with power limitation and perception error limitation.

Taking downlink coordinated multipoint joint transmission as an example, as shown in fig. 2, consider a cellular millimeter wave system with L base stations, the set of which is denoted asEach base station is equipped with a base station consisting of N _t Root transmit antenna and N _r A uniform linear array of receive antennas. The receiving antenna array is used for receiving echo signals to actively sense the motion parameters of the user, and the number of the antennas is assumed to be enough. Let->Representing and base station/@>Associated user set, k _l Representation->Is the kth user of (a). Suppose that each user is fit with itThe LoS channel is always present between the serving base stations, while the cell-crossing channel between the base station and the users in the neighboring cells consists of LoS and multiple channels, e.g. rice fading channels. All base stations are interconnected by a limited capacity backhaul. The transmission time T is divided into N equally long intervals epoch, each having a time length Δt. Backhaul delay T _d And satisfy T _d ＜ΔT。

Will user k _l The transmitted signal in epoch n is expressed asAnd the corresponding transmit precoding vector is denoted +.>If user k _l When epoch n is scheduled, & gt>Otherwise->Thus, user k _l The received signal is +.>Wherein (1)>Andrepresenting from base station m to user k _l Downlink channel of (2), and->Is additive white gaussian noise.

Proportional fairness is one of the most important transmission metrics in dynamic networks, while PFR (Proportional Fairness Rate ) maximization problem can be solved by processing the following problem model at each epoch n:

wherein P is _l Representing the maximum transmit power of base station l, the cumulative average of the rates of the first (n-1) epochs isAnd->Is user k _l At the transmission rate of epoch n'.

The problem model described above requires a complex mixed integer nonlinear programming, which requires the exchange of instantaneous CSIHowever, this can lead to significant delays under limited backhaul bandwidth. In addition, for mobile users, CSI estimation needs to be performed frequently to achieve timely estimation, which results in significant signaling overhead.

In order to solve the problems, the invention adopts the radar sensing technology of the base station to develop a novel distributed MCRA framework. First, the present invention employs a SALINR signal metric that relies solely on CSI from the base station to distribute optimized beamforming vectors. Second, each base station l estimates user k using the reflected radar signal _l Kinematic parameters (including angle)Distance->And speed->). Under LoS channel, the acquired kinematic parameters can be used for BSl to estimate the user's channel +.>In addition, the cross-cell LoS channel can be obtained through interaction with other base stations. Then, the estimated channels are used for subsequent user scheduling, respectively +.>And beamforming +.>And (5) optimizing.

The echo signal received by base station l at epoch n can be modeled as:

wherein the method comprises the steps ofIs the antenna array gain,/">Respectively represent user j _m Reflection coefficient, doppler frequency and round trip delay for base station l, < >>Representing symmetric complex gaussian noise. Further, the transmission steering vector a (θ) and the reception steering vector b (θ) are respectively:

since the user angle change in the epoch interval is small, when k' _l ≠k _l When it is availableAnd->Thus, base station l is related to user k _l The echo signals of (2) are:

wherein,by means of a matched filter and an angle estimation method (such as MUSIC, etc.), the angle is +.>Delay->And Doppler frequency->The measurement noise of (a) is respectivelyAnd->And is also provided with

Wherein G is the matched filter gain, a _i I= { θ, τ, μ } is the AND systemThe relevant constants are configured.

The base station l performs one-step prediction by means of a state evolution model (such as an extended Kalman filter) to obtain a signal corresponding to the user k _l Associated angleDistance->Speed->And reflection coefficient->Further, the channel vector may be estimated as

Wherein,is the path loss coefficient, depending on +.>And carrier frequency f _c . In addition, in the case of the optical fiber,

SALINR is an alternative signal measurement indicator that can be used as a virtual objective function in distributed multi-cell resource allocation. Specifically, the arithmetic average of the interference power of base station l to users served by other base stations is introduced as:

thus, the virtual PFR maximization problem can be updated as:

wherein,and->Is a SALINR expression.

While global CSI is not required compared to the problem model, CSI for users scheduled from base station to other base stations is still required. Furthermore, inter-base station scheduling of coupling remains a challenge for interference coordination. To address the above challenges, the present invention contemplates resource allocation for the next several phases.

The invention separates the user scheduling and the beam forming design, adopts an enhanced low complexity Proportion Fair Zero Forcing Greedy (PFZFG) method to determine the scheduling set before the beam forming vector is optimizedThe core idea of PFZFG is to select users with weaker channel correlation, resulting in a spatially well separated scheduling set. This separation helps to reduce intra-cell interference in downlink transmissions, and in combination with salnr-based precoding, can mitigate inter-cell interference, thereby improving system throughput. In addition, it reduces the overlap of the beam main lobes in the respective intended directions, thus further improving the accuracy of active perception. Furthermore, the PFZFG approach also considers the fairness characteristics of users over successive periods. If some users are long-lived (e.g., T _s -1 epochs) are not scheduled, then the current epochs should be scheduled immediately to prevent large motion parameter tracking errors. The specific scheduling algorithm comprises the following steps:

(1) Selecting a group to be in succession T _s -1 time unscheduled user Solving +.>Sum rate of users in (1)

(2) Calculation ofWherein->Is the historical average rate of the time n-1 before the user i, and is fed back to the service base station through uplink, wherein +.>And->

(3) If R is greater than or equal to R', thenRepeating steps (2) - (3) until R < R'; output->As the set of users scheduled by base station l at time n.

Once the collection is determinedBased on the base station's own coordinates and the user's kinematic parameters, the base station/can determine the user +.>Coordinates of->And speed->The base station l obtains the user by transmitting back the informationAngle m +.l +.>Path loss coefficient->And Doppler frequency->So that the base station l to user t can be estimated _m Is a cross-cell channel of (c): />

Based on the above formula, the leakage value can be approximately calculated. Since only state information of scheduled users needs to be exchanged, the exchange overhead is low.

To this end, the beamforming vector optimization problem may be solved in a distributed manner across the various base stations. Clearly, the estimated noise variance of the echo signal plays a crucial role in channel estimation and will also directly influence the effectiveness of the solution of precoding. Especially in multi-cell resource allocation, the direction of beamforming is difficult to align with the target user for accurate sensing.

For this reason, for a base station, the present invention considers the virtual PFR maximization problem of being power constrained and perceived error constrained as follows:

in order to solve the non-convex problem, the method adopts a method of combining fractional programming with semi-definite relaxation to solve the problem. In detail, by introducing auxiliary variablesAnd->The objective function is equivalent to

When (when)Fixed, optimal->And->To address non-convex perceptual constraints, the problem of planning by semi-positive rules can be further expressed as:

wherein,and->

Andif the constraint of rank 1 is not considered, semi-deterministic planning is a convex problem and can be easily solved by CVX. Then, using the eigenvector approximation, we can get +.>

In simulation experiments, the numerical results presented verify the validity of the ISAC-SALINR framework proposed by the present invention. Consider a three base station scenario with coordinates (-100, 0), (0, 0), (100, 0), each serving 22 users at a speed of 15 m/s. The base station and the user are all in f _c =30 GHz and bandwidth b=30 kHz. In addition, Δt=20ms, T _d =4ms and T _s =3. The transmit power is 38dBm, the noise power spectral density is-174 dBm/Hz, and the rice factor ρ=10 dB. Furthermore, a _θ ＝0.1，a _τ ＝6.7×10 ^-7 ，a _μ =0.5, g=10, c=0.5 and path lossThe results of the present invention are based on the average of 10 simulations of different user locations and compare PFRs of different schemes running 10 iterations of the algorithm. The present invention also considers a centralized Weighted Minimum Mean Square Error (WMMSE) as a performance comparison scheme.

Figure 4 plots the PFR obtained over time. For the idea CSI case, it is assumed that neither cwmse nor SALINR algorithms need to be perceived, and that the schedule is embedded in the optimization of BF designs. The SALINR algorithm of the present invention implements a PFR comparable to CWMMSE. In contrast, over time, the performance of cwmse with initial CSI experiences a significant PFR degradation due to non-negligible CSI errors. Furthermore, if all users are scheduled, ISAC-SALINR (unscheduled) will be aware of using more resources, resulting in reduced communication performance. In contrast, the PFR obtained by the ISAC-SALINR algorithm provided by the invention approaches CWMMSE, and the effectiveness of our framework is verified. In summary, the method provided by the invention has good performance in aspects of system performance improvement, channel fading resistance and the like.

The foregoing is merely a preferred embodiment of the invention, and it is to be understood that the invention is not limited to the form disclosed herein but is not to be construed as excluding other embodiments, but is capable of numerous other combinations, modifications and environments and is capable of modifications within the scope of the inventive concept, either as taught or as a matter of routine skill or knowledge in the relevant art. And that modifications and variations which do not depart from the spirit and scope of the invention are intended to be within the scope of the appended claims.

Claims (10)

1. The distributed multi-cell dynamic resource allocation method based on perception assistance is characterized by comprising the following steps of:

step 1, estimating user kinematic parameters and CSI (channel state information) based on double-function radar signal echoes;

step 2, determining a user set scheduled by the local station based on the estimated CSI;

step 3, the base station interactively schedules the motion parameters of the user to estimate the CSI from the local station to the adjacent station user;

step 4, constructing a distributed beam forming optimization problem model;

and 5, optimizing beam forming by combining fractional programming with semi-positive relaxation and solving a beam forming vector.

2. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 1, wherein step 1 specifically comprises:

each base station transmits a dual function radar signal to the user terminal using the perceptually integrated beam, estimates a kinematic parameter of the user based on the reflected signal of the user terminal, and constructs a corresponding next-time channel state vector.

3. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 2, wherein estimating the kinematic parameters of the user based on the reflected signal of the user side, and constructing the corresponding next-time channel state vector specifically comprises:

at time n-1, the base stationFirst, the reflected signals of different users are distinguished by using a receiving filter, and then the scheduled user k is estimated based on a matched filtering method _l Time delay of->And Doppler shift->Finally estimate angle->Obtaining a path loss coefficient predicted at the moment n according to a kinematic evolution model>Doppler shift->Angle->To construct base station l and service user k _l Direct channel between: />Wherein N is _t Indicating the number of transmitting antennas of the base station, U _l Representing the set of users served by base station l.

4. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 1, wherein step 2 specifically comprises: and constructing a corresponding channel state vector at the next moment according to the base station, and determining a user set scheduled by the base station at the next moment.

5. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 4, wherein a greedy proportional fair zero forcing scheme is adopted to select the user set scheduled at the time of n, and the method specifically comprises:

step 21, selecting a group of continuous T _s -1 time unscheduled user K _s ＝{π ₁ ,...,π _j-1 }，Solving K with equal power zero-forcing beamforming _s Sum rate of users in (1)

Step 22, calculatingWherein->Is the historical average rate of the time n-1 before the user i, and is fed back to the service base station through uplink, wherein +.>And->

Step 23, if R is greater than or equal to R', thenRepeating steps 21 and 22 until R < R'; output->As the set of users scheduled by base station l at time n.

6. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 1, wherein step 3 specifically comprises: the base stations interact kinematic parameters of users to be scheduled of the base station through a backhaul network to assist the base station in estimating channel state vectors of the base station and other adjacent base station service users.

7. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 6, wherein the interaction between base stations through backhaul network specifically comprises:

base station/determines according to two-dimensional coordinates thereofTwo-dimensional coordinates of the user->Movement speed +.>Taking into account the inherent delay T of the backhaul network _d The base station l is T before the next information transmission moment _d Information is interacted with other base stations, then the base station l estimates itself to schedule the user t to the other base station m _m Is used for direct channel transmission.

8. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 1, wherein step 4 specifically comprises: and the base station utilizes the estimated channel state vector to construct a virtual proportional fairness and rate maximization distributed beam forming optimization problem model with power limitation and perception error limitation.

9. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 8, wherein the virtual proportional fairness and rate maximization problem model is expressed as:

wherein,a precoding vector representing a time node n of a kth user in a cell l; />Representing a user scheduling set of the cell l at the time node n; p (P) _l Indicating the maximum transmit power of base station l; />Is user k _l The estimated error for i (including angle, delay and doppler shift) and is required to be less than a tolerable constant c; in addition, a->Is user k _l Virtual rate function at time n, +.>Is user k _l The average rate at the first n-1 time is used as a weight.

10. The distributed multi-cell dynamic resource allocation method based on perception assistance according to claim 1, wherein optimizing beamforming by fractional programming in combination with semi-positive relaxation and solving beamforming vectors specifically comprises: by introducing auxiliary variablesAnd->The objective function is equivalent to the one described above,

when (when)Fixing (I)>Then->

The semi-positive problem is expressed as:

where G is the matched filter gain,and->

In addition, in the case of the optical fiber,is the antenna gain of the base station, a _i Is a perceptual error constant;

and if the constraint condition of rank 1 is not considered, adopting eigenvalue decomposition to obtain the optimal beam forming vector.