Multi-cell interference coordination method based on channel knowledge map

Abstract

The invention relates to a multi-cell interference coordination method based on a channel knowledge map, which comprises the following steps of: deploying a channel gain map (CGM) at a base station end and extracting the CSI from the CGM under the condition of considering inter-cell interference so as to calculate a signal to interference plus noise ratio (SINR) of a user on each sub-channel; performing sub-channel allocation of interference coordination in a multi-cell scene based on a maximized SINR algorithm; and after the CSI of each sub-channel is updated through a channel gain map, multi-cell power distribution is carried out through a minimum weighted mean square error (WMMSE) algorithm. In a multi-cell and multi-user scene, the complete CSI is obtained by constructing the channel gain map to carry out multi-cell sub-channel allocation and combined power allocation, so that the overhead of channel estimation is remarkably reduced while the system and rate gains are ensured.

Description

Technical Field

[0001] The present invention relates to a technology in the field of wireless communications, in particular to a multi-cell interference coordination method based on a channel knowledge map. Background Art

[0002] In 5G, with the complexity of the wireless environment, in addition to the acquisition of channel state information (CSI), there are also complex interference scenarios, such as integrated air-ground-sea networks. In such an environment, there will be serious co-channel interference between multiple base stations, resulting in a decrease in communication capabilities, which has a very negative impact on the user experience. The traditional network optimization method based on road test requires a lot of overhead and cannot be intelligently optimized with the wireless environment and user service distribution. In interference coordination without considering channel conditions, the base station directly allocates wireless resources equally to each user, ensuring fairness among users. However, in the multi-cell scenario, due to the huge co-channel interference between base stations, such interference coordination methods do not consider specific channel conditions, and thus cannot obtain good throughput; and the acquisition of perfect CSI is almost impossible in practice. Under the dense deployment of 5G base stations, the time overhead caused by channel estimation and the estimation error of the channel matrix are almost unavoidable. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention proposes a multi-cell interference coordination method based on a channel knowledge map. In a multi-cell and multi-user scenario, a channel gain map is constructed to obtain complete CSI to perform sub-channel allocation and joint power allocation of multiple cells, thereby significantly reducing the overhead of channel estimation while ensuring system and rate gains.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a multi-cell interference coordination method based on a channel knowledge map. A channel gain map (CGM) is deployed at a base station end. Under the condition of considering the interference between cells, after CSI is extracted from the CGM and the signal to interference plus noise ratio (SINR) of the user on each sub-channel is calculated, sub-channel allocation of interference coordination in a multi-cell scenario is performed based on a maximum SINR algorithm. After the CSI of each sub-channel is updated through the channel gain map, multi-cell power allocation is performed through a minimum weighted mean square error algorithm (WMMSE).

[0006] The channel gain map is constructed based on the DNN trained with simulation data.

[0007] The simulation data is obtained using but not limited to Wireless Insite.

[0008] The CSI is obtained based on the base station-user position pairs in the channel gain map.

[0009] The present invention relates to a system for implementing the above method, comprising: a channel knowledge map construction and CSI acquisition unit, a maximum SINR multi-cell sub-channel allocation unit and a WMMSE power allocation unit, wherein: the channel knowledge map construction and CSI acquisition unit is deployed at the base station end, and obtains the global CSI by inputting the location information of the base station-user into the CGM without any channel estimation and backhaul bandwidth overhead; the maximum SINR multi-cell sub-channel allocation unit allocates a suitable sub-channel to each user according to the SINR sorting list based on the acquired CSI information through a maximum SINR heuristic algorithm; The WMMSE power allocation unit updates the CSI of each user through CGM according to the result of subchannel allocation, and performs multi-cell power allocation based on the weighted minimum mean square error algorithm to reduce the inter-cell interference of the system and obtain system and rate gains. Technical Effects

[0010] The present invention uses CGM-based multi-cell channel estimation-free interference coordination to achieve multi-cell inter-cell interference coordination without any channel training overhead on the basis of building CGM. Compared with other channel estimation algorithms, it significantly reduces the CSI acquisition overhead and the cost of CSI exchange between base stations, reduces inter-cell interference, and obtains certain system and rate gains. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a flow chart of the present invention;

[0012] Figure 2 A flowchart for sub-channel joint power allocation;

[0013] Figure 3 This is a schematic diagram of the 5G multi-cell interference network scenario and CKM deployment;

[0014] Figure 4 It is a 3D real environment map;

[0015] Figure 5 is the channel gain map;

[0016] Figure 6 It is a real physical map;

[0017] Figure 7 It is a real CSI map;

[0018] Figure 8 Schematic diagram of CGM built based on DNN;

[0019] Fig. 9It is a comparison diagram of different interference coordination schemes and rates;

[0020] Fig.10 is a schematic diagram of average SINR in an embodiment;

[0021] Fig.11 Comparison of channel training overhead of different methods. DETAILED DESCRIPTION

[0022] like Figure 1 As shown, this embodiment relates to a multi-cell interference coordination method based on a channel knowledge map for a non-convex optimization type interference coordination problem caused by co-channel interference between multiple cells. Through the two stages of sub-channel allocation and power allocation, after multi-cell sub-channel joint power allocation is performed based on a channel gain map, a certain channel gain is obtained while reducing the channel estimation overhead. Specifically, it includes:

[0023] Step 1: Construct a channel gain map (CGM) based on DNN, including:

[0024] Step 1.1: According to the location of the base station and the user, the channel gain information on multiple sub-channels is predicted by a deep neural network (DNN).

[0025] The DNN adopts a fully connected neural network (FCNN), and commercial ray tracing software Wireless Insite is used to generate training data and test data for training the FCNN.

[0026] like Figure 4 As shown in the figure, the park environment map and its corresponding ground truth CSI data are simulated and generated by Wireless Insite using the built-in ray tracing algorithm, which is used to train the DNN to minimize the mean square error (MSE) between the real CSI and the predicted CSI.

[0027] Step 1.2: Based on the location information of the base station (BS) and the terminal device (UE), without any channel estimation, use Figure 3 and Figure 5 The CGM shown approximates the wireless channel gain at the UE location, thereby recovering the complete CSI, specifically: the channel gain h at each user location k =f(ε k ,β k ), where: β k is the wireless propagation environment, the base station-user location pair ε k ∈U, U is the location space of all potential UEs of interest.

[0028] Step 2: Figure 2As shown, subchannel allocation in interference coordination in a multi-cell scenario is performed based on a channel gain map, where each subchannel of a base station can be allocated to at most one user in its cell, and each user obtains at most one subchannel, specifically including:

[0029] Step 2.1: When the transmit power of each base station is known, allocate subchannels to each user to maximize the system sum rate R(l,p) = ∑ m∈M ∑ n∈N l m,k,n r m,k,n , where: binary variable l m,k,n =1 means that base station m allocates subchannel n to user k, otherwise it means that subchannel n is not allocated to user k, and the rate r obtained by k on subchannel n m,k,n =Blog2(1+SINR m,k,n ), the signal interference noise ratio received by user k on subchannel n h m,k,n is the channel gain from base station m to user k on subchannel n, σ 2 is the noise power, M is the number of cells, and N is the number of subchannels waiting to be allocated for each base station.

[0030] Step 2.2: Calculate the SINR of each user on each subchannel and sort them to find the maximum SINR for each base station, and assign the corresponding subchannel n to the corresponding user k. K m ={1,2,...,K m} is the set of service users in m that have not been allocated sub-channels;

[0031] Step 2.3: After the allocation is completed, m Eliminate user k from the subchannel set, eliminate subchannel n from the subchannel set N, and then update the SINR sorted list of the remaining users. Repeat this allocation process until all users are allocated to subchannels.

[0032] Step 3: Multi-cell power allocation based on the channel gain map, specifically including:

[0033] Step 3.1: After completing the subchannel allocation step of multiple cells based on CGM, the base station updates the local CSI of each user on the corresponding subchannel by querying CGM to allocate the transmission power to each user without frequent channel estimation.

[0034] Step 3.2: Use the weighted minimum mean square error (WMMSE) algorithm to find the appropriate downlink power allocation coefficient p for each user. m,k , That is, the maximum system and rate are obtained when each user meets the following power constraints: Where: m∈M,P max is the maximum transmit power of each base station.

[0035] The weighted minimum mean square error algorithm, for a given transmit power and noise interference, calculates the optimal linear combination weights of the channel observation variables to maximize the total transmission rate of the system. It can be regarded as a non-convex optimization problem, but after system transformation, it can be transformed into a convex optimization problem, specifically including:

[0036] a) Initialize the power allocation coefficient for each user, and each base station evenly allocates power to serve users.

[0037] b) Update the transmit weight factor w according to the power allocation and the receiving beam of each user k =(1-u k h m,k p m,k ) -1 , where: u k is the receiving beam factor of user k, h m,k is the channel gain of base station m at user k, p m,k is the power allocation coefficient of base station m to user k, and the receiving beam update formula of user k to serving base station m is Where: p j,k is the interference power sent by base station j to user k, h j,k is the channel gain of interfering base station j to user k, σ 2 is the noise power;

[0038] c) Using the updated receiving beam and weight factors, a weighted linear least squares problem is solved based on the weighted MMSE criterion to obtain the new power allocation for user k. Where: p max is the constraint of the maximum transmit power of the base station:

[0039] d) By iteratively calculating step b and step c, the optimal linear combination weight of the received signal and the optimal power allocation for each user are obtained according to the termination condition.

[0040] The termination condition is: the WMMSE algorithm converts the weighted system sum rate maximization problem into the weighted sum mean square error minimization problem. Therefore, after the three optimization variables are optimized for multiple iterations, when When , that is, in each iteration, the weighted mean square error increment is less than ε, the algorithm stops, where: and are the weight factors of user k at time t and time t-1 respectively.

[0041] After specific practical experiments, for multi-cell cellular network scenarios, under different maximum transmission powers of base stations, the gains of the CGM-based interference coordination algorithm and other interference coordination algorithms in terms of system and rate are compared, and the time overhead required for CGM and other channel estimation algorithms to obtain CSI is compared. Consider a 1km×1km urban scenario with some buildings and vegetation, and use Wireless Insite to simulate such a real scenario. There are 4 macro base stations and multiple users distributed in the scenario. The user distribution conforms to the Poisson Point Process (PPP) and is evenly distributed throughout the scenario. The transmit power of each base station varies from 40dBm to 46dBm.

[0042] like Figure 6 As shown in the figure, the real scene map constructed by Wireless Insite is shown in the figure. The simulation data of channel gain at different positions is obtained by running the ray tracing algorithm of Wireless Insite, which is used to train DNN. The real CSI map drawn based on the simulation data is shown in the figure below. Figure 7 As shown, the training data is obtained through simulation to train DNN to construct CGM. The constructed CGM is as follows Figure 8 As shown in the figure, it can be seen intuitively that the DNN-based construction method can better match the real CSI map, which shows that the channel gain prediction effect of CGM is better. Figure 6 The real layout of the physical environment.

[0043] The average SINR of all users is used to verify the effectiveness of this method in reducing interference. The number of users is set to change from 50 to 60 to simulate a larger inter-cell interference situation, and the average SINR of all users is compared. Fig.10 As shown in the figure, the CGM-based method also has certain performance gains in terms of interference. The average user SINR of this method is about 2.5dB higher than that of the LS channel estimation algorithm and about 0.4dB higher than that of the SVD-based channel estimation algorithm, which shows that the proposed interference coordination scheme is also effective in reducing inter-cell interference.

[0044] like Fig.11As shown in the figure, the time overhead difference between the CGM-based CSI acquisition method and the two channel estimation methods is as follows. It can be seen that compared with the traditional LS channel estimation and SVD channel estimation algorithms, the CGM-based method can significantly reduce the required training time slots. For example, when the number of users is 50, the LS-based estimation algorithm requires about 70 time slots, while the SVD-based algorithm requires 125 time slots to complete the channel estimation. Compared with these two algorithms, the CGM-based method consumes almost no time overhead and can obtain similar or even better sum rate performance.

[0045] Compared with the existing technology, this method constructs a channel knowledge map and deploys it at the base station end, without any channel estimation overhead and channel information exchange overhead between base stations, and designs a sub-channel joint power allocation algorithm in a multi-cell scenario. This method improves the sum rate gain of the wireless network system, while significantly reducing the time overhead required for channel estimation and interference between cells.

[0046] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A multi-cell interference coordination method based on a channel knowledge map, characterized in that: The channel gain map (CGM) is deployed at the base station end. Taking into account the interference between cells, the CSI is extracted from the CGM to calculate the signal to interference plus noise ratio (SINR) of the user on each subchannel. Then, the subchannel allocation of interference coordination in the multi-cell scenario is performed based on the maximum SINR algorithm. After updating the CSI of each subchannel through the channel gain map, the multi-cell power allocation is performed through the minimum weighted mean square error algorithm (WMMSE).

2. The multi-cell interference coordination method based on channel knowledge map according to claim 1, characterized in that: The extracting of CSI specifically includes: Step 1.1, according to the location of the base station-user, predict the channel gain information on multiple sub-channels through a deep neural network (DNN); Step 1.2: Based on the location information of the base station (BS) and the terminal equipment (UE), the wireless channel gain at the UE location is approximated using CGM to restore the complete CSI, specifically: the channel gain h at each user location k =f(ε k ,β k ), where: β k is the wireless propagation environment, the base station-user location pair ε k ∈U, U is the location space of all potential UEs of interest.

3. The multi-cell interference coordination method based on channel knowledge map according to claim 1, characterized in that: The subchannel allocation described above, where each subchannel of the base station can be allocated to at most one user in its cell, and each user obtains at most one subchannel, specifically includes: Step 2.1: When the transmit power of each base station is known, allocate subchannels to each user to maximize the system sum rate R(l,p) = ∑ m∈M ∑ n∈N l m,k,n r m,k,n , where: binary variable l m,k,n =1 means that base station m allocates subchannel n to user k, otherwise it means that subchannel n is not allocated to user k, and the rate r obtained by k on subchannel n m,k,n =Blog2(1+SINR m,k,n ), the signal interference noise ratio received by user k on subchannel n h m,k,n is the channel gain from base station m to user k on subchannel n, σ 2 is the noise power, M is the number of cells, and N is the number of subchannels waiting to be allocated for each base station; Step 2.2: Calculate the SINR of each user on each subchannel and sort them to find the maximum SINR for each base station, and assign the corresponding subchannel n to the corresponding user. K m ={1,2,...,K m } is the set of service users in m that have not been allocated sub-channels; Step 2.3: After the allocation is completed, m Eliminate user k from the subchannel set, eliminate subchannel n from the subchannel set N, and then update the SINR sorted list of the remaining users. Repeat this allocation process until all users are allocated to subchannels.

4. The multi-cell interference coordination method based on channel knowledge map according to claim 1, characterized in that: The multi-cell power allocation specifically includes: Step 3.1, after completing the subchannel allocation step of multiple cells based on CGM, the base station updates the local CSI of each user on the corresponding subchannel by querying CGM, so as to allocate the transmission power to each user without frequent channel estimation; Step 3.2: Use the weighted minimum mean square error (WMMSE) algorithm to find the appropriate downlink power allocation coefficient for each user. That is, the maximum system and rate are obtained when each user meets the following power constraints: Where: m∈M,P max is the maximum transmit power of each base station.

5. The multi-cell interference coordination method based on channel knowledge map according to claim 1 or 4, characterized in that: The weighted minimum mean square error algorithm calculates the optimal linear combination weights of the channel observation variables for the transmission power and noise interference, and converts the problem of maximizing the total transmission rate of the system into a convex optimization problem for solution, which specifically includes: a) Initialize the power allocation coefficient for each user, and each base station evenly allocates power to the service user; b) Update the transmit weight factor w according to the power allocation and the receiving beam of each user k = ( 1-u k h m,k p m,k) -1 , where: u k is the receiving beam factor of user k, h m,k is the channel gain of base station m at user k, p m,k is the power allocation coefficient of base station m to user k, and the receiving beam update formula of user k to serving base station m is Where: p j,k is the interference power sent by base station j to user k, h j,k is the channel gain of interfering base station j to user k, σ 2 is the noise power; c) Using the updated receiving beam and weight factors, a weighted linear least squares problem is solved based on the weighted MMSE criterion to obtain the new power allocation for user k. Where: p max is the constraint of the maximum transmit power of the base station: d) By iteratively calculating step b and step c, the optimal linear combination weight of the received signal and the optimal power allocation for each user are obtained according to the termination condition.

6. The multi-cell interference coordination method based on channel knowledge map according to claim 5, characterized in that: The termination condition is: That is, in each iteration, the iteration is stopped when the weighted mean square error increment is less than ε, where: and are the weight factors of user k at time t and time t-1 respectively.

7. A multi-cell interference coordination system based on a channel knowledge map for implementing any of the methods described in claims 1-6, characterized in that: include: Channel knowledge map construction and CSI acquisition unit, maximum SINR multi-cell sub-channel allocation unit and WMMSE power allocation unit, wherein: the channel knowledge map construction and CSI acquisition unit is deployed at the base station end, and obtains the global CSI by inputting the base station-user location information into the CGM without any channel estimation and backhaul bandwidth overhead; the maximum SINR multi-cell sub-channel allocation unit allocates appropriate sub-channels to each user according to the SINR sorting list based on the acquired CSI information through the maximum SINR heuristic algorithm; the WMMSE power allocation unit updates the CSI of each user through the CGM according to the result of sub-channel allocation, and performs multi-cell power allocation based on the weighted minimum mean square error algorithm to reduce the system's inter-cell interference and obtain system and rate gains.

Images