US-12627400-B2 - Size-based neural network selection for autoencoder-based communication

Abstract

Methods, systems, and devices for wireless communications are described. In some wireless communications systems, devices may implement multiple autoencoders for communications. A wireless device may select an autoencoder to use for communications based on a size parameter for a message. For example, a user equipment (UE) may receive a grant from a base station indicating a size parameter for communicating a message. The UE and base station may determine, from a set of neural network (NN)-based encoders configured at the UE, an NN-based encoder corresponding to the size parameter. The UE may communicate the message with the base station according to the grant and based on the determined NN-encoder. In some examples, the UE and base station may determine a number of resource segments from a set of resources allocated for communication and may determine respective NN-based encoders for the different resource segments.

Inventors

• Qiaoyu Li
• Yu Zhang
• Hao Xu
• Chao Wei
• Liangming WU
• Rui Hu

Assignees

• QUALCOMM INCORPORATED

Dates

Publication Date

20260512

Application Date

20200911

Claims (20)

1. 1 . A method for wireless communications at a user equipment (UE), comprising: receiving, from a network device, a grant indicating a size parameter for communicating a message; determining, from a subset of a plurality of neural network-based encoders configured at the UE, a neural network-based encoder based at least in part on a neural network-based encoder index indicated by the grant, a transmission parameter, or both, the subset of the plurality of neural network-based encoders corresponding to the size parameter; and communicating the message with the network device according to the grant and based at least in part on the determined neural network-based encoder.
2. 2 . The method of claim 1 , wherein the communicating comprises: communicating the message in a non-coherent transmission, the non-coherent transmission comprising the message and no reference signals associated with the message for channel estimation.
3. 3 . The method of claim 1 , further comprising: determining the subset of the plurality of neural network-based encoders corresponding to the size parameter.
4. 4 . The method of claim 1 , further comprising: determining the subset of the plurality of neural network-based encoders corresponding to the size parameter.
5. 5 . The method of claim 4 , wherein the transmission parameter comprises a bandwidth indicator, a modulation scheme, a channel coding rate, or a combination thereof.
6. 6 . The method of claim 1 , further comprising: receiving, from the network device, a configuration message indicating the plurality of neural network-based encoders and one or more size parameters corresponding to each neural network-based encoder of the plurality of neural network-based encoders.
7. 7 . The method of claim 1 , wherein the UE is pre-configured with the plurality of neural network-based encoders and one or more size parameters corresponding to each neural network-based encoder of the plurality of neural network-based encoders.
8. 8 . The method of claim 1 , wherein the size parameter corresponds to a transport block size for the message, a number of resource elements for transmitting the message, a number of physical resource blocks for transmitting the message, a number of orthogonal frequency domain multiplexing symbols for transmitting the message, a number of coded bits for transmitting the message, or a combination thereof.
9. 9 . The method of claim 1 , wherein the communicating comprises: modulating the message using the determined neural network-based encoder; and transmitting, to the network device, the modulated message.
10. 10 . The method of claim 1 , further comprising: determining a neural network-based decoder corresponding to the determined neural network-based encoder, wherein the communicating comprises: receiving, from the network device, the message; and demodulating the message using the determined neural network-based decoder.
11. 11 . The method of claim 10 , wherein the neural network-based decoder is determined based at least in part on a neural network-based decoder index indicated by the grant.
12. 12 . The method of claim 1 , wherein the grant comprises a dynamic uplink grant, a configured grant, a dynamic downlink grant, a semi-persistent scheduling configuration, or a combination thereof.
13. 13 . A method for wireless communications, comprising: determining a plurality of segments for a set of resources allocated for communicating a message; determining, for each segment of the plurality of segments, a respective neural network-based encoder from a respective subset of a plurality of neural network-based encoders based at least in part on a neural network-based encoder index, a transmission parameter, or both, the respective subset of the plurality of neural network-based encoders corresponding to a respective size parameter for each segment, wherein a first neural network-based encoder determined for a first segment of the plurality of segments is different from a second neural network-based encoder determined for a second segment of the plurality of segments; and communicating the message in the set of resources based at least in part on the determined respective neural network-based encoder for each segment of the plurality of segments.
14. 14 . The method of claim 13 , further comprising: determining that a transport block size for the message satisfies a threshold size, wherein the plurality of segments is determined based at least in part on the transport block size for the message satisfying the threshold size.
15. 15 . The method of claim 13 , wherein the communicating comprises: modulating the message in each segment of the plurality of segments using the determined respective neural network-based encoder for each segment of the plurality of segments; and transmitting the modulated message.
16. 16 . The method of claim 13 , further comprising: determining, for each segment of the plurality of segments, a respective neural network-based decoder corresponding to the determined respective neural network-based encoder for each segment of the plurality of segments, wherein the communicating comprises: receiving the message; and demodulating the message in each segment of the plurality of segments using the determined respective neural network-based decoder for each segment of the plurality of segments.
17. 17 . The method of claim 13 , wherein the plurality of segments is determined based at least in part on one or more resource elements of the set of resources, one or more physical resource blocks of the set of resources, one or more sub-bands of the set of resources, one or more orthogonal frequency domain multiplexing symbols of the set of resources, one or more orthogonal frequency domain multiplexing symbol groups of the set of resources, or a combination thereof.
18. 18 . The method of claim 13 , wherein the respective size parameter for a segment of the plurality of segments corresponds to a number of resource elements for the segment, a number of physical resource blocks for the segment, a number of orthogonal frequency domain multiplexing symbols for the segment, a number of coded bits for the segment, or a combination thereof.
19. 19 . A user equipment (UE), comprising: at least one processor; at least one memory coupled with the at least one processor; and instructions stored in the at least one memory and executable by the at least one processor to cause the UE to: receive, from a network device, a grant indicating a size parameter for communicating a message; determine, from a subset of a plurality of neural network-based encoders configured at the UE, a neural network-based encoder based at least in part on a neural network-based encoder index indicated by the grant, a transmission parameter, or both, the subset of the plurality of neural network-based encoders corresponding to the size parameter; and communicate the message with the network device according to the grant and based at least in part on the determined neural network-based encoder.
20. 20 . The UE of claim 19 , wherein the instructions to communicate are executable by the at least one processor to cause the UE to: communicate the message in a non-coherent transmission, the non-coherent transmission comprising the message and no reference signals associated with the message for channel estimation.

Description

CROSS REFERENCE The present application is a 371 national stage filing of International PCT Application No. PCT/CN2020/114677 by LI et al. entitled “SIZE-BASED NEURAL NETWORK SELECTION FOR AUTOENCODER-BASED COMMUNICATION,” filed Sep. 11, 2020, which is assigned to the assignee hereof, and which is expressly incorporated by reference in its entirety herein. FIELD OF TECHNOLOGY The following relates to wireless communications, including size-based neural network (NN) selection for autoencoder-based communication. BACKGROUND Wireless communications systems are widely deployed to provide various types of communication content such as voice, video, packet data, messaging, broadcast, and so on. These systems may be capable of supporting communication with multiple users by sharing the available system resources (e.g., time, frequency, and power). Examples of such multiple-access systems include fourth generation (4G) systems such as Long Term Evolution (LTE) systems, LTE-Advanced (LTE-A) systems, or LTE-A Pro systems, and fifth generation (5G) systems which may be referred to as New Radio (NR) systems. These systems may employ technologies such as code division multiple access (CDMA), time division multiple access (TDMA), frequency division multiple access (FDMA), orthogonal frequency division multiple access (OFDMA), or discrete Fourier transform spread orthogonal frequency division multiplexing (DFT-S-OFDM). A wireless multiple-access communications system may include one or more base stations or one or more network access nodes, each simultaneously supporting communication for multiple communication devices, which may be otherwise known as user equipment (UE). In some wireless communications systems, a wireless device (e.g., a UE or a base station) may utilize an autoencoder for encoding transmissions, decoding transmissions, or both. However, an autoencoder may be optimized or otherwise trained based on a specific message size for communication and may perform relatively poorly for other message sizes. As such, a wireless device implementing an autoencoder may experience inefficient modulation techniques, large signaling overhead, or both depending on the size of a communicated message. Furthermore, in cases where a message is communicated in a non-coherent transmission, channel estimation may be unavailable to help with demodulation of the message, resulting in potentially unreliable demodulation using an autoencoder. SUMMARY The described techniques relate to improved methods, systems, devices, and apparatuses that support size-based neural network (NN) selection for autoencoder-based communication. Generally, the described techniques provide for a user equipment (UE) and a base station to implement autoencoder-based communications by selecting a specific NN to use for communicating a message from a set of multiple configured NNs based on a size parameter for the message. In some cases, the message may be communicated in a non-coherent transmission (e.g., a transmission lacking reference signals, such as demodulation reference signals (DMRSs), for channel estimation). The UE may be configured or preconfigured with the set of NN-based encoders. The UE may receive, from the base station, a grant indicating the size parameter (e.g., a transport block (TB) size or another size parameter) for communicating a message. Based on the size parameter, the UE and base station may each determine an NN-based encoder (i.e., the same NN-based encoder from the configured set of NN-based encoders). In some cases, multiple NN-based encoders may be associated with a single size parameter. In some such cases, the base station may explicitly indicate one NN-based encoder using an NN-based encoder index included in the grant. In some other such cases, the UE and base station may implicitly determine one NN-based encoder based on one or more transmission parameters. The UE and base station may communicate the message according to the grant and based on the determined NN-based encoder. For example, one wireless device may encode the message for transmission using the NN-based encoder, and the other wireless device may determine an NN-based decoder for decoding the message. In some cases, the NN-based decoder may correspond to the determined NN-based encoder. In some other cases, the NN-based decoder may be determined based on an NN-based decoder index indicated by the grant. In some examples, separate NN-based encoders and/or decoders may be determined for different segments of allocated resources. For example, a wireless device (e.g., a UE or a base station) may determine the different segments of allocated resources based on a respective size parameter of the segments. A method for wireless communications at a UE is described. The method may include receiving, from a base station, a grant indicating a size parameter for communicating a message, determining, from a set of NN-based encoders configured at the UE, an NN-based encoder