Integrated sensing and communication using deep neural networks

Abstract

End-to-end integrated radar sensing and communication (JSC) functionality is provided between a first device (101) and a second device (112) using a deep neural network (DNN). The DNN structure (103) of the first device generates JSC signal waveforms (107, 107a~107n) representing input communication data (102) and radar signals. The second device (1112) processes the JSC signal waveforms in its DNN structure (114) to generate reconstructed communication data (115). The DNN structure of the second device generates radar sensing feedback (RFB) signal waveforms (117) and reflections (111a~111n) from one or more objects (109a~109n) for transmission to the first device. The first device uses a DNN structure to process the reflections (110a-110n) of either the received RFB signal waveform and / or the JSC signal waveform transmitted from one or more objects (109a-109n) in order to generate radar sensing information (105) of one or more objects.

Description

[Background Technology]

[0001] The concept of wireless communication systems coexisting with positioning / sensing or radar systems is known, for example, as Joint Communication Radar (JCR), Joint Sensing and Communications (JSC), Joint Communications and Sensing (JCS), Integrated Communication and Sensing (ICS), Joint Radar and Communication Systems (JRCS), and Joint Radar Sensing and Communication (JRSC). Hereafter, the acronym JSC will be used to refer collectively to these types of dual communication / radar systems. Current JSC systems aim to share common radar sensing and communication hardware and air interface resources to enable simultaneous communication and radar functions of a particular JSC system type. JSC system types include, but are not limited to, bistatic JSC, multistatic JSC, and monostatic JSC. However, wireless communication system designers typically design and optimize wireless communication systems to provide high data rates with a specified quality of service, while radar system designers typically design and optimize radar systems for target object detection and range / velocity estimation. As a result of these different design objectives, wireless communication systems have historically used different waveforms than radar systems. For example, 5G advanced wireless communication systems typically use orthogonal frequency division multiplexing (OFDM) waveforms, while 6G wireless communication systems may use orthogonal time frequency space (OTFS) waveforms.In radar systems, the waveforms used can vary depending on the JSC system type, and may include, for example, frequency-modulated continuous wave (FMCW), linear frequency modulation (LFM), or nonlinear frequency modulation (NLFM) waveforms. The differences between these waveforms make it difficult to efficiently implement JSC in one or more JSC system types in 5G or 6G and beyond wireless communication systems.

[0002] Many fifth-generation (5G) or sixth-generation (6G) and later wireless communication systems have complex transmitter and receiver processing chains and standard waveforms (e.g., 5G OFDM waveforms) designed to meet the performance requirements of the corresponding communication standards. The 5G or 6G transmitter processing chain of a first device may include the arrangement of encoding, interleaving, scrambling, precoding, modulation, and radio frequency (RF) analog transmission components to process the input communication data for transmission using the OFDM waveform. The 5G or 6G receiver processing chain of a second device may include the arrangement of RF analog receiving, channel estimation, demodulation, descrambling, deinterleaving, and decoding components for receiving the OFDM waveform transmitted by the first device and recovering the transmitted input communication data. For one or more JSC system types, combining such hardware and corresponding waveforms (e.g., OFDM and OTFS) with radar sensing transmitter and receiver chains and their corresponding waveforms (e.g., FMCW, LFM, and NLFM) while simultaneously meeting the requirements of the corresponding communication standards and accurately performing target object detection and range / velocity estimation is a complex task. [Overview of the project]

[0003] This specification relates to methods, apparatus, and systems that extend wireless communication systems (e.g., current 5G Advanced or 6G and beyond) with transmitter / receiver deep neural network (DNN) structures that can efficiently perform integrated sensing and communication (JSC) under various communication and / or radar conditions / environment / performance requirements (e.g., beyond line of sight (NLOS) communication, within line of sight (LOS) communication, multipath interference, multiple access interference, narrowband interference, changes in meteorological or atmospheric conditions, changes in radar channel conditions, target characteristics, different target sizes and types, channel throughput, channel frequency or frequency band, channel bandwidth, channel delay spread, channel angle spread, or any other type of communication signal interference or radar conditions / performance requirements, etc.), while simultaneously performing accurate detection of target objects and / or range / velocity estimation, and meeting the communication performance requirements of wireless communication systems (e.g., corresponding 5G Advanced or 6G and beyond communication standards).

[0004] In a first aspect, the Disclosure provides a method performed by a first device, the method comprising establishing a JSC DNN operation for one or more time slots of a communication session with a second device, wherein for each time slot of one or more time slots, the method further comprises: acquiring input communication data for transmission to the second device for each time slot; processing the input communication data by the JSC DNN structure of the first device to generate an output JSC signal representing the input communication data and radar signals; transmitting the output JSC signal to the second device as a JSC signal waveform for each time slot via a communication channel; receiving one or more radar sensing signals for each time slot based on reflections of transmitted JSC signal waveforms from one or more objects; processing the received radar sensing signals by the JSC DNN structure of the first device to generate radar sensing information / measurements for one or more objects in the communication channel; and transmitting the radar sensing information / measurements for one or more objects for each time slot to one or more higher-layer protocols of the protocol stack of the first device.

[0005] In a second aspect, the Disclosure provides a method performed by a second device, the method comprising establishing a JSC DNN operation for one or more time slots of a communication session with a first device according to the first aspect, wherein for each time slot of one or more time slots, the method performs the following: receiving JSC signal waveforms transmitted from the first device for each time slot via a communication channel, the JSC signal waveforms representing input communication data and radar signals; the method further performs the following: receiving one or more further JSC signal waveforms for each time slot based on reflections of transmissions of JSC signal waveforms from one or more objects in the communication channel; processing the received JSC signal waveforms and further JSC signal waveforms in a JSC DNN structure in the second device to generate reconstructed communication data corresponding to the input communication data and radar sensing feedback signals associated with one or more objects; and transmitting the reconstructed communication data to a data sink of the second device or to one or more higher protocol layers of the protocol stack of the second device.

[0006] In a third aspect, the Disclosure provides a method performed by a second device, the method comprising establishing a JSC DNN operation for one or more time slots of a communication session with a first device according to the first aspect, wherein for each time slot of one or more time slots, the method performs the function of receiving JSC signal waveforms transmitted from the first device for each time slot via a communication channel, the JSC signal waveforms representing input communication data and radar signals, the method further performs the function of receiving one or more further JSC signal waveforms for each time slot based on reflections of transmissions of JSC signal waveforms from one or more objects in the communication channel, and processing the received JSC signal waveforms and further JSC signal waveforms by a communication DNN structure in the second device, the communication DNN structure configured to generate reconstructed communication data corresponding to the input communication data transmitted from the first device, the method further performs the function of transmitting the reconstructed communication data to a data sink of the second device, or transmitting the reconstructed communication data to one or more upper protocol layers of the protocol stack of the second device.

[0007] Further embodiments provide computer-readable media, apparatus, and systems for carrying out the methods of the first, second, third, and fourth embodiments.

[0008] The methods, apparatus, and system embodiments offer numerous advantages, including, for example, performing efficient JSC during a communication session between a first device and a second device, such as a 5G or 6G or later wireless communication system, for various communication channel conditions, while simultaneously performing accurate radar sensing (e.g., target object detection and / or range / velocity estimation) and meeting the performance requirements of a communication standard corresponding to one or more JSC system types. An additional advantage is the efficient, rapid, and dynamic reconfiguration of JSC DNN operation (e.g., bistatic, multistatic, or monostatic JSC) during a communication session between a first device and a second device while maintaining accurate radar sensing and communication performance. Further advantages include efficiently designing JSC waveforms for use in 5G or 6G or later wireless communication systems and for various radar sensing and communication channel conditions, while simultaneously performing accurate radar sensing and meeting the requirements of the corresponding communication standard.

[0009] Embodiments will be described as examples with reference to the following drawings. [Brief explanation of the drawing]

[0010] [Figure 1] This is a schematic diagram illustrating an exemplary wireless communication system for performing integrated sensing and communication (JSC) using a deep neural network (DNN). [Figure 2] This schematic diagram illustrates another exemplary wireless communication system for performing JSC using a DNN that highlights the hardware components of a device. [Figure 3a] Figure 1 or Figure 2 is a schematic diagram illustrating an example of a bistatic JSC in a wireless communication system. [Figure 3b] This is a schematic diagram showing an example of a multistatic JSC in the wireless communication system shown in Figure 1 or Figure 2. [Figure 3c]Figure 1 or Figure 2 is a schematic diagram showing an exemplary monostatic JSC of a wireless communication system. [Figure 3d] This is a schematic diagram illustrating an exemplary bistatic or multistatic JSC using the first, second, and third devices in the wireless communication system shown in Figure 1 or Figure 2. [Figure 4] This is a signal flow diagram showing the bistatic JSC DNN operation for one or more time slots of a communication session between the first device and the second device. [Figure 5] This is a signal flow diagram showing the multistatic JSC DNN operation for one or more time slots of a communication session between the first device and the second device. [Figure 6] This is a signal flow diagram showing the monostatic JSC DNN operation for one or more time slots of a communication session between the first device and the second device. [Figure 7a] This is a signal flow diagram showing the bistatic JSC DNN operation for one or more time slots of a downlink (DL) communication session between a base station (BS) and user equipment (UE). [Figure 7b] This is a signal flow diagram showing the multi-static JSC DNN operation for one or more time slots of a DL communication session between BS and UE. [Figure 7c] This is a signal flow diagram showing the monostatic JSC DNN operation for one or more time slots of a DL communication session between BS and UE. [Figure 8a] This flowchart illustrates an exemplary JSC DNN process for a first device performing JSC DNN operations within one or more time slots of a communication session with a second device. [Figure 8b] This flowchart illustrates the exemplary establishment of JSC DNN operation by the first device for one or more time slots of a communication session with the second device. [Figure 9a]A flowchart showing an exemplary JSC DNN process of a second device that performs bistatic or multistatic JSC DNN operations within one or more time slots of a communication session with a first device. [Figure 9b] A flowchart showing another exemplary JSC DNN process of a second device that performs monostatic JSC DNN operations within one or more time slots of a communication session with a first device. [Figure 9c] A flowchart showing an exemplary establishment of JSC DNN operations by a second device within one or more time slots of a communication session with a first device. [Figure 10a] A schematic diagram showing an exemplary master neural network table for selecting or configuring a corresponding JSC DNN structure for use in JSC DNN operations within one or more time slots of a communication session between a first device and a second device, and optionally a third device. [Figure 10b] A schematic diagram showing an exemplary first neural network table used by a first device to select and configure a corresponding JSC DNN structure for use in JSC DNN operations within one or more time slots of a communication session between the first device and the second device, and optionally a third device. [Figure 10c] A schematic diagram showing an exemplary second neural network table used by a second device to select and configure a corresponding JSC DNN structure for use in JSC DNN operations within one or more time slots of a communication session between the first device and the second device. [Figure 10d] A schematic diagram showing an exemplary third neural network table used by a third device to select and configure a corresponding JSC DNN structure for use in JSC DNN operations within one or more time slots of a communication session between a first device and a second device when the third device assists the first device. [Figure 10e] This flowchart illustrates the exemplary JSC DNN structure selection process for the first device during the establishment of JSC DNN operation with the second device. [Figure 11] This schematic diagram illustrates an exemplary co-training process for training the first and second JSC DNN structures for use by the first and second devices, respectively. [Figure 12] This is a schematic diagram of an example computer-readable medium. [Modes for carrying out the invention]

[0011] Common reference numbers are used throughout the drawings to indicate similar features.

[0012] Figure 1 shows an example of a wireless communication system 100 including a first device 101 and a second device 112 configured to perform joint sensing and communication (JSC) using a deep neural network (DNN). In the wireless communication system 100, the first device 101 and the second device 112 establish a communication session. During the communication session, the first device 101 and the second device 112 perform JSC DNN operations in one or more time slots of the communication session. The first device 101 and the second device 112 include corresponding configurable JSC DNN structures 103 and 114 that perform the JSC DNN operations. The JSC DNN operations include the first device 101 and the second device 112 each configuring their corresponding JSC DNN structures 103 and 114 according to a selected JSC system type (e.g., bistatic, multistatic, or monostatic JSC), communication performance requirements, radar sensing requirements, and channel parameters. After configuration, further JSC DNN operation includes the first device 101 and the second device 112 using their respective JSC DNN structures 103 and 114 to perform JSC according to a selected JSC system type within one or more time slots of the communication session.

[0013] The first device 101 includes a JSC DNN structure 103 coupled to a radio frequency (RF) front-end transmitter (Tx) / receiver (Rx) subsystem (RF front-end Tx / Rx subsystem) 104. The first device 101 acquires input communication data 102 from a data source for transmission to a second device 112 within one or more time slots. The input communication data 102 for a time slot is input to the JSC DNN structure 103. The JSC DNN structure 103 processes the input communication data 102 for a time slot to generate an output JSC signal 106a for transmission within that time slot. The output JSC signal 106a integrates the input communication data 102 with the radar signal and / or radar signal characteristics generated by the JSC DNN structure 103. That is, the output JSC signal 106a represents a digitally integrated communication and radar signal in which the input communication data 102 is incorporated together with the radar signal characteristics. For example, the JSC DNN structure 103 generates an output JSC signal 106a from input communication data 102, the output JSC signal 106a representing the input communication data 102 for transmission to a second device 112 within its time slot, but is digitally converted (or adapted) to also function as a radar signal. For example, the JSC DNN structure 103 processes the input communication data 102 to generate an output JSC signal 106a representing the input communication data 102 converted to function as a radar signal at the time of transmission. The RF front-end Tx / Rx subsystem 104 processes the output JSC signal 106a (e.g., digital-to-analog (DAC) conversion and frequency upconversion to radio frequency) and transmits it to the second device 112 via the wireless communication / radar channel 108 (also called the wireless communication channel or communication channel) as a JSC signal waveform 107 within the corresponding time slot. The JSC signal waveform 107 for the corresponding time slot represents the input communication data 102 that is transmitted to a second device 112, which is adapted to also function as a radar signal within that time slot.For example, the JSC signal waveform 107 for a corresponding time slot is a waveform representing the input communication data 102 transmitted to the second device 112 within that time slot and operates jointly as a radar signal within that time slot. This JSC signal waveform 107 for a time slot is an integrated radar sensing and communication signal waveform for a time slot, in which the input communication data 102 of the time slot is integrated / embedded. The JSC waveform 107 has communication signal characteristics and radar signal characteristics that enable it to be used for JSC applications. Within the environment of the wireless communication channel 108, there are one or more objects 109a, 109b, and 109c~109n (e.g., people, animals, obstacles, trees, buildings, vehicles and / or any other objects that can reflect the transmitted JSC signal waveform 107).

[0014] Depending on the JSC system type, the first device 101 receives, for each time slot, one or more reflections 110a, 110b, and 110c-110n (e.g., reflected radar signals) of the transmitted JSC signal waveform 107 from one or more corresponding objects 109a, 109b, and 109c-109n, and / or radar sensing feedback (RFB) signal waveform 117 from the second device 112. The RF front-end Tx / Rx subsystem 104 receives one or more reflections 110a-110n of the JSC signal waveform 107 from one or more objects 109a-109n, and / or the RFB signal waveform 117 from the second device 112, performs baseband processing (e.g., analog-to-digital conversion and frequency down-conversion to baseband) to obtain one or more radar sensing signals 106b. Since each of the radar sensing signals (or multiple signals) 106b is derived from the reflections 110a-110n or 1111a-111n of the JSC signal waveform 107 from one or more objects 109a-109n, each of the radar sensing signals (or multiple signals) 106b includes one or more components and / or radar signal characteristics of the radar signal incorporated within the transmitted JSC signal waveform 107, resulting from the transmission of the time slot output JSC signal 106a. The radar signal components / radar signal characteristics of the reflections 110a-110n or 111a-111n are associated with the corresponding objects 109a-109n. In other words, each radar sensing signal (or multiple signals) 106b for each time slot includes data representing one or more component radar signals (or multiple signals) or radar signal characteristics (or multiple signals) associated with the reflections 110a-110n or 111a-111n of the JSC signal waveform 107 transmitted from one or more objects 109a-109n in the time slot. The radar sensing signals (or multiple signals) 106b are input to and processed by the JSC DNN structure 103 to generate radar sensing information 105 (e.g., Doppler, velocity, positioning, etc.) for one or more objects 109a-109n.The radar sensing information 105 may also be called radar sensing measured values ​​or estimates, but will be referred to as radar sensing information 105 hereafter in this specification. The first device 101 detects one or more radar sensing signals 106b and these objects 109a to 109n and / or radar sensing information 105 about those detected objects using a JSC DNN structure 103.

[0015] The second device 112 includes an RF front-end Tx / Rx subsystem 113 coupled to another JSC DNN structure 114. The RF front-end Tx / Rx subsystem 113 of the second device 112 receives, for a corresponding time slot, at least component 107a of the transmitted JSC signal waveform 107 and one or more further reflected JSC signal waveforms 111a, 111b, and 111c-111n based on reflections of the JSC signal waveform 107 transmitted from one or more corresponding objects 109a, 109b, 109c-109n in the environment of the wireless communication channel 108. The JSC signal waveform 107 for a corresponding time slot represents input communication data 102 transmitted to the second device 112, which has been converted / adapted to also function as a radar signal within that time slot. The RF front-end Tx / Rx subsystem 113 receives at least component 107a of the transmitted JSC signal waveform 107 and one or more further reflected JSC signal waveforms 111a~111n, processes them (e.g., analog-to-digital (ADC) conversion and frequency down-conversion to baseband) to obtain the JSC signal 116a received in the corresponding time slot. The received JSC signal 116a in the corresponding time slot is input to the JSC DNN structure 114. The JSC DNN structure 114 processes the received JSC signal 116a to generate reconstructed communication data 115 corresponding to the input communication data 102 transmitted within the corresponding time slot.

[0016] Depending on the JSC system type (e.g., bistatic, multistatic, or monostatic JSC), the JSC DNN structure 114 of the second device 112 can be configured to either a) reconstruct communication data from the received JSC signal 116a (e.g., monostatic JSC system type), or b) reconstruct communication data from the received JSC signal 116a for transmission to the first device 101 as an RFB signal waveform 117 and generate an RFB signal 116b associated with one or more objects 109a~109n (e.g., bistatic or multistatic JSC system type). If the second device 112 is configured for b), the RF front-end Tx / Rx subsystem 113 transmits the RFB signal 116b to the first device 101 as an RFB signal waveform 117 for the corresponding time slot for processing by the JSC DNN structure 103 of the first device 101.

[0017] Subsequently, the first device 101 receives one or more radar sensing signals 106b for the corresponding time slot based on a) reflections 110a-110n of the transmitted JSC signal waveform for the corresponding time slot (e.g., monostatic JSC system type), b) RFB signal waveform 117 for the corresponding time slot (e.g., bistatic JSC system type), or c) reflections 110a-110n of the transmitted JSC signal waveform and RFB signal waveform 117 for the corresponding time slot (e.g., multistatic JSC system type). The first device 101 inputs one or more radar sensing signals 106b for the corresponding time slot to the JSC DNN structure 103. The JSC DNN structure 103 processes one or more radar sensing signals 106b and generates radar sensing information 105 (e.g., range, Doppler, velocity, position, or other radar measurements) for one or more objects 109a-109n in the environment of the wireless communication channel 108. The radar sensing requirements can determine the type of radar sensing information generated by the JSC DNN structure 103.For example, the JSC DNN structure 103 provides the range estimates for each associated object 109a to 109n, the Doppler estimates for each associated object 109a to 109n, the location or position estimates for each associated object 109a to 109n, the delay spreads for each associated object 109a to 109n, the Doppler spreads for each associated object 109a to 109n, the mean delays for each associated object 109a to 109n, and the values ​​of each associated object 109a to 109n. The process involves generating radar sensing information 105 which includes generating radar sensing information (or measured / estimated values) for each of the objects 109a to 109n, which includes generating radar sensing information (or measured values / estimates) for each of the objects 109a to 109n, which includes generating radar sensing information (or measured values / estimates) for each of the objects 109a to 109n, which includes generating radar sensing information 105 which includes generating radar sensing information (or measured values / estimates) for each of the objects 109a to 109n, which includes generating radar sensing information (or measured values / estimates) for each of the objects 109a to 109n, which includes generating radar sensing information (or measured values / estimates) for each of the objects 109a to 109n, which includes generating radar sensing information (or measured values / estimates) for each of the objects 109a to 109n, which includes generating radar sensing information (or measured values / estimated values

[0018] The first device 101 and the second device 112 have JSC DNN structures 103 and 114 for use in JSC DNN operation during a communication session between them. Once JSC DNN operation is established during the communication session, the first device 101 and the second device 112 configure their JSC DNN structures 103 and 114 with corresponding trained DNN models, at least in part according to the JSC system type that the first device 101 selects to generate unique radar sensing information 105 of one or more objects 109a to 109n in the environment. As illustrated with reference to Figure 11, a machine learning (ML) algorithm is used to co-train (e.g., using supervised or unsupervised learning) a combination of DNN models for each specific JSC system type, or a DNN and a combination of communication and / or radar sensing channel characteristics (e.g., various channel conditions and radar sensing requirements), to form trained pairs of JSC DNN structures 103 and 114, respectively, when JSC is executed by a first device 101 and a second device 112. Each trained pair of JSC DNN structures 103 and 114 includes one or more trained transmitter and / or receiver DNN model configurations based on the JSC system type (e.g., bistatic, multistatic, or monostatic JSC), channel conditions / characteristics, communication performance requirements, and radar sensing requirements, which are used when co-training these DNN models. Each trained pair of JSC DNN structures can provide end-to-end JSC without the complexity of a conventional transmitter and receiver JSC processing chain. The JSC DNN structures 103 and 114 extend and / or replace the conventional transmitter and receiver JSC processing chains.

[0019] For example, the configuration of a trained pair of JSC DNN structures is used to configure the JSC DNN structure 103 for the first device 101 and the JSC DNN structure 114 for the second device 112. After configuration, the JSC DNN structure 103 generates transmitted JSC signal waveforms representing input communication data 102 and radar signals that are suitable for efficiently overcoming various channel conditions / environments used during training (e.g., beyond line of sight (NLOS) communication, within line of sight (LOS) communication, multipath interference, multiple access interference, and narrowband interference, changes in weather or atmospheric conditions, different target sizes and types, and other radar and / or communication signal interference), and at the same time processes the input communication data 102 to meet the communication performance requirements of the wireless communication system 100 (e.g., corresponding 5G or 6G or later communication standards), depending on the type of JSC system used during training. The corresponding JSC DNN structure 114 of the second device 112 overcomes the various channel conditions / environments described above when processing the received JSC signal waveform to generate reconstructed communication data 115 that meets the communication performance requirements of the wireless communication system 100 (e.g., the corresponding 5G or 6G or later communication standard), and / or, depending on the JSC system type, to generate an RFB signal waveform 117 for transmission to the first device 101 for input to the JSC DNN structure 103 of the first device 101 for accurate target object detection and / or range / velocity estimation.

[0020] The JSC DNN operation performed during a communication session between the first device 101 and the second device 112 is established based on the JSC system type when the first device 101 identifies the channel conditions for the wireless communication / radar channel 108 between the first device 101 and the second device 112, and according to the JSC performance requirements to be met during the communication session. The JSC performance requirements include the JSC system type, radar sensing requirements, and further communication performance requirements. The radar sensing requirements include, but are not limited to, one or more of the following, which may affect the frequency range selection of the transmitted JSC signal: range resolution and Doppler resolution. The communication performance requirements include, but are not limited to, one or more of the following, which are defined by the communication standard (e.g., 5G or 6G or later) implemented by the wireless communication system 100: throughput, latency, block error rate, and / or other performance metrics / parameters. For example, channel conditions for a wireless communication (and / or radar) channel 108 that affect communication and / or radar performance include, but are not limited to, one or more of the following: beyond-line-of-sight communication channel conditions, within-line-of-sight communication channel conditions, meteorological or atmospheric conditions, radar channel conditions, target characteristics, channel throughput, channel frequency or frequency band, channel bandwidth, channel delay spread, channel Doppler spread, channel angle spread, or any other type of conditions affecting the communication channel between the first and second devices of the JSC.

[0021] For example, the first device 101 identifies a JSC system type based on radar sensing requirements (e.g., range resolution, Doppler resolution, frequency band, etc.) for generating radar sensing information 105 (e.g., range, Doppler, velocity, position, or other radar measurements) for one or more potential objects 109a-109n, and communication performance requirements (e.g., block error rate, latency, throughput, etc.) to be met during a communication session with the second device 112. The first device 101 also identifies the channel conditions of the wireless communication channel 108 and, using the selected JSC system type, searches for, selects, and retrieves an appropriate trained pair of JSC DNN structures from multiple trained pairs of JSC DNN structures, with each trained pair of JSC DNN structures associated with a JSC DNN identifier. The selected pair of trained JSC DNN structures satisfies the JSC performance requirements. After selection, the first device 101 sends a control message to the second device 112, which includes a JSC DNN identifier associated with the selected trained pair of JSC DNN structures. The JSC DNN identifier specifies which JSC DNN structures the first device 101 and the second device 112 will acquire and use for one or more time slots when performing JSC DNN operations.

[0022] Depending on the type of wireless communication system 100, the control message also specifies one or more communication resource parameters (e.g., downlink (DL) and / or uplink (UL) resource block (RB) configuration, DL and / or UL frequencies, etc.) for use by the first device 101 and the second device 112 when receiving the JSC signal waveform 107 transmitted from the first device 101. Depending on the JSC system type, the control message also specifies communication resource parameters for use by the second device 112 when transmitting the RFB signal waveform 117 corresponding to one or more time slots to the first device 101.

[0023] In bistatic JSC, the first device 101 transmits input communication data 102 and a JSC signal waveform 107 representing the radar signal within a time slot, and the second device 112 provides an RFB signal waveform 117 corresponding to the time slot to assist the first device 101 during radar sensing of the time slot with respect to the JSC signal waveform 107 transmitted within the time slot. In this case, the JSC system type is a bistatic JSC system type, the trained JSC DNN structures 103 and 114 of the first device 101 and the second device 112 respectively have bistatic JSC processing capabilities, and the second device 112 assists the first device 101 during radar sensing of the time slot by providing the RFB signal waveform 117 corresponding to the time slot to the first device 101. The JSC DNN structure 103 of the first device 101 processes the input communication data 102 of the time slot and generates an output JSC signal 106a as a JSC signal waveform 107 for transmission within the time slot by the RF front-end Tx / Rx subsystem 104. The JSC signal waveform 107 represents the input communication data 102 and the radar signal of the time slot.

[0024] The JSC DNN structure 103 of the first device 101 also processes one or more radar sensing signals 106b derived from the RFB signal 117 transmitted from the second device 112a to the first device 101, after the JSC DNN structure 114 of the second device 112 generates an RFB signal 116b from at least the received component 107a of the transmitted JSC signal waveform 107 and the reflected JSC signal waveforms 111a, 111b, and 111c-111n resulting from the reflection of the transmitted JSC signal waveform 107 from objects 109a-109n. The JSC DNN structure 114 of the second device 112 processes the received component 107a of the transmitted JSC signal waveform 107 and the reflected JSC signal waveforms 111a, 111b, and 111c-111n to generate both reconstructed communication data 115 corresponding to one or more objects 109a-109n and RFB signals 116b (e.g., Doppler, velocity, and range). The second device 112 transmits the RFB signal 116b as RFB signal waveform 117 to the first device 101. The generated reconstructed communication data 115 corresponds to the input communication data 102 transmitted in the JSC signal waveform 107 during the corresponding time slot.

[0025] The RF front-end Tx / Rx subsystem 104 receives the RFB signal waveform 117 for the corresponding time slot from the second device 112, processes it (e.g., ADC and frequency down-converting to baseband), and converts it into a radar sensing signal 106b. The JSC DNN structure 103 of the first device 101 processes the radar sensing signal 106b derived from the radar sensing feedback signal waveform 117 for the corresponding time slot in order to generate radar sensing information 105 for one or more objects 109a to 109n for the corresponding time slot. The bistatic JSC system type has the advantage that the first device 101 does not need to be in full-duplex communication mode to receive and process the reflections 110a to 110n of the JSC signal waveform 107 transmitted from one or more objects 109a to 109n when generating the radar sensing information 105.

[0026] In multistatic JSC, the first device 101 and the second device 112 perform bistatic JSC, but the first device 101 also receives reflections 110a-110n of the JSC signal waveform 107 transmitted within a time slot from one or more objects 109a-109n (e.g., in full-duplex communication). In this case, the JSC system type is a multistatic JSC system type in which both JSC DNN structures 103 and 114 are trained to have bistatic JSC processing capability, and furthermore, JSC DNN structure 103 is also trained to have monostatic JSC processing capability. The RF front-end Tx / Rx subsystem 104 of the first device 101 receives the reflections 110a-110n and processes them (e.g., ADC and frequency down-conversion to baseband) to produce one or more radar sensing signals 106b. The RF front-end Tx / Rx subsystem 104 receives and processes an RFB signal waveform 117 associated with a time slot from the second device 112, converting it into one or more radar sensing signals 106b. The radar sensing signals 106b are then processed by the JSC DNN structure 103 of the first device 101 to generate radar sensing information 105 for one or more objects 109a to 109n with respect to the time slot. The multistatic scenario offers the advantage of additional radar sensing accuracy or tracking accuracy by having the first device 101 operate in full-duplex communication mode.

[0027] Other bistatic / multistatic scenarios are also possible, for example, a third device (see, for example, the third device 341 in Figure 3d) providing other RFB signals to the first device 101. For example, the first device 101 may establish JSC DNN operation for one or more time slots in a communication session with the second device 112, while the first device 101 employs the third device to provide radar feedback support in response to JSC signal waveforms 107 transmitted between the first device 101 and the second device 112. The third device includes a radar sensing DNN structure solely for processing the received JSC signal waveforms 107 transmitted from the first device 101 and the corresponding reflections of the transmitted JSC signal waveforms 107 for each time slot, and for transmitting the generated RFB signals for the time slots corresponding to the first device 101. In this scenario, the second device 112 performs a JSC DNN operation in which the JSC DNN structure 114 of the second device 112 processes the received JSC signal waveform 107 and its reflections 111a-111n from one or more objects 109a-109n in the transmitted JSC signal waveform 107 and / or RFB signal waveform 117 to generate reconstructed communication data 115 corresponding to the input communication data 102 transmitted to the second device 112. Optionally, the second device 112 generates only the reconstructed communication data 115 for the time slot, and the third device provides the RFB signal for the time slot to the first device 101. This configuration offers the advantage of further improving the resolution of radar sensing within the wireless communication system 100, and / or providing flexibility by enabling JSC operation between the first device 101 and the second device 112, by allowing the second device 112 to process the transmitted JSC signal waveform 107 to reconstruct the input communication data 102.

[0028] In monostatic JSC, the first device 101 transmits input communication data 102 and a JSC signal waveform 107 representing the radar signal of the time slot within the time slot, and at the same time receives reflections 110a to 110n of the transmitted JSC signal waveform 107 associated with the time slot from one or more objects 109a to 109n (for example, in full-duplex communication mode). The second device 112 generates only reconstructed communication data 115 from the received JSC signal waveform 107 and / or its reflections 111a to 111n. In this case, the JSC system type is a monostatic JSC system type in which the first device 101 configures a JSC DNN structure 103 for monostatic JSC processing capability and the second device 112 configures a JSC DNN structure 114 to reconstruct only the communication data 115. The RF front-end Tx / Rx subsystem 104 of the first device 101 receives reflections 110a to 110n of the transmitted JSC signal waveform 107, processes them (e.g., ADC and frequency down-conversion to baseband), and converts them into one or more radar sensing signals 106b. The JSC DNN structure 103 processes these signals to generate radar sensing information 105 associated with one or more objects 109a to 109n. By operating in full-duplex communication mode, the first device 101 can simultaneously transmit the JSC signal waveform 107 to generate radar sensing information 105 for one or more objects 109a to 109n and receive reflections 110a to 110n of the JSC signal waveform from one or more objects 109a to 109n. This minimizes latency in the JSC DNN structure 103 that receives one or more radar sensing signals 106b, resulting in the advantage of faster generation of radar sensing information 105 for one or more objects 109a to 109n for each time slot.

[0029] Optionally, the first device 101 also has a protocol stack comprising multiple protocol layers. After generating radar sensing information 105 for a specific time slot, or for one or more time slots, the first device 101 transmits the radar sensing information 105 for one or more time slots to one or more higher-layer protocols of the first device 101's protocol stack. In the first device 101's protocol stack, lower layers serve the role of providing services to higher layers, and higher layers use those services to provide their own functions. For example, radar sensing information 105 is generated in the physical layer of the protocol stack, passed up to the application layer of the protocol stack by each of the higher layers, processed, and the corresponding radar sensing information 105 is used, but is not limited to, display to a user, further processing, and / or transmission to one or more applications of the first device 101 or the second device 112 for further processing and / or consumption of the radar sensing information 105.

[0030] Similarly, the second device 112 also has a protocol stack comprising multiple protocol layers. After the JSC DNN structure 114 generates reconstructed communication data 115 for a specific time slot or one or more time slots, the second device 112 transmits the reconstructed communication data 115 for one or more time slots to one or more higher-layer protocols of the second device 112's protocol stack. In the second device 112's protocol stack, lower layers serve the role of providing services to higher layers, and higher layers use those services to provide their own functionality. For example, the reconstructed communication data 115 is generated in the physical layer of the protocol stack, passed by each of the higher layers up to the application layer of the protocol stack, processed, and the corresponding reconstructed communication data 115 is used, but is not limited to, display to a user, further processing, and / or transmission to one or more applications of the second device 112 for further processing and / or consumption of the reconstructed communication data 115.

[0031] The first device 101 and the second device 112 can be any type of communication device for use in the wireless communication system 100, including, but not limited to, any combination of a terrestrial network (TN) base station (BS), a non-terrestrial network (NTN) BS (or a network device including a satellite, drone, or high-altitude platform station (HAPS)), user equipment (UE), or other RAN elements in the wireless communication system 100. For example, the first device 101 and the second device 112 may be two BSs, or two UEs, or a BS and a UE, or a UE and a BS, or any other combination of communication devices, depending on the requirements of the application. Figure 2 shows a wireless communication system 200 in which the first device 201 is a TN BS and the second device 212 is a UE.

[0032] Figure 2 shows another example of a wireless communication system 200 in which a first device 201 performs JSC DNN communication with a second device 212. Reference numerals in Figure 1 are reused in Figure 2 for similar or identical components or features. In this example, the first device 201 is a TN BS called BS201, and the second device 212 is a UE called UE212. BS201 connects to the core network of the wireless communication system 200 via one or more interfaces. For example, the wireless communication system 200 may be a 5G or 6G wireless communication system.

[0033] In this example, BS201 and UE212 communicate via downlink (DL) and uplink (UL) transmissions over wireless communication channel 108. Wireless communication channel 108 may include a DL communication channel (e.g., a physical downlink shared channel (PDSCH)) for transmitting DL transmission signals from BS201 to UE212, and a UL communication channel (e.g., a physical uplink shared channel (PUSCH)) for transmitting UL transmission signals from UE212 to BS201. The DL communication channel may also include a DL control channel (e.g., a physical downlink control channel (PDCCH)), and the UL communication channel may also include a UL control channel (e.g., a physical uplink control channel (PUCCH)). In this example, UE212 transmits UL transmission signals via the UL control channel.

[0034] BS201 is implemented as a computing system / device for performing any of the corresponding methods, JSC DNN operations or processes described herein, and / or for implementing any of the corresponding systems, units and / or devices described herein. BS201 includes interconnected RF front-end Tx / Rx subsystems 204, one or more transceivers 220, one or more processors 221, and memory units 223. Those skilled in the art will understand that other types of computing devices / systems / platforms, such as distributed computing systems, may be used instead to implement BS201 and the methods described herein, depending on the requirements of the application. BS201 includes one or more processors 221 (e.g., a central processing unit (CPU)). One or more processors 221 control the operation of other components of BS201, such as the RF front-end Tx / Rx subsystems 204, one or more transceivers 220, and memory units 223. One or more processors 221 may be a single-core device or a multi-core device. One or more processors 221 may include a CPU or a graphics processing unit (GPU). Alternatively, one or more processors 221 may include specialized processing hardware, such as a reduced instruction set computer (RISC) processor, or programmable hardware with built-in firmware. BS201 may include multiple processor cores. In some embodiments, one or more processors 221 may be part of a distributed computing system, such as a cloud computing system and / or a cloud computing platform.

[0035] One or more processors 221 of BS201 may be connected to a network interface such as a transceiver 220 including a transmitter (Tx) and receiver (Rx) for communicating via an RF front-end Tx / Rx subsystem 204 over a wireless communication channel 108 of the network with other devices and systems, such as UE212, other communication devices, network equipment, RAN entities or devices, operators, and / or any other devices, services, systems, and / or devices requested by the application. Optionally, one or more processors 221 may be connected to a user interface (UI) for user or operator input to instruct or use BS201 and / or the underlying computing system, and / or output data from there. Optionally, one or more processors 221 may be connected to a display for displaying output to the user or operator.

[0036] BS201 includes a memory system or memory unit 223 that includes working or volatile memory. One or more processors 221 may access the volatile memory to process data and may control the storage of data into the memory. The volatile memory may include any type of random access memory (RAM), such as static RAM (SRAM), dynamic RAM (DRAM), etc., or may include flash memory, such as a secure digital card. In some embodiments, the memory unit 223 and / or one or more volatile memories may include a number of multiple memories that form part of a distributed computing system, such as a cloud computing system and / or a cloud computing platform. BS201 also includes non-volatile memory. The non-volatile memory may store a set of operation or operating system instructions and / or a set of software instructions in the form of computer-readable instructions for controlling the operation of the processor 221, which, when executed on one or more processors 221, cause the processor 221 to perform the methods, processes, operations and / or functions of the operation and / or methods of the JSC DNN described herein. Non-volatile memory may be any type of memory, depending on the application requirements, such as read-only memory (ROM), flash memory, secure digital drive, magnetic drive memory, or magnetic disk drive memory. In some embodiments, non-volatile memory may include a number of multiple non-volatile memories that form part of a distributed computing system, such as a cloud computing system and / or cloud computing platform.

[0037] The non-volatile memory of the BS201's memory unit 223 contains computer program code and / or instructions for executing the BS DNN controller (BS DNNC) 224 and / or the BS JSC DNN structure 203. When executed on one or more processors 221, the BS DNNC 224 uses the BS JSC DNN structure 203 and the BS DNN configuration store 225 stored in the memory unit 223 to control the JSC DNN operation on the BS201. Although the BS DNNC 224 is part of the memory unit 223, this is merely an example, and the BS DNNC is not limited in this way. Those skilled in the art will understand that the BS DNNC 224 may be executed in hardware and / or software on the BS201 as required by the application. In this example, BS DNNC224 configures a BS JSC DNN structure 203 to run a DL JSC application, including a BS DL transmitting communication and radar DNN model (BS DL Tx CR-DNN) 203a coupled to a BS DL receiving radar DNN model (BS DL Rx R-DNN) 203b. For example, BS DL Tx CR-DNN203a processes the input communication data to generate an output JSC signal, which is transmitted by BS201 as a JSC signal waveform representing the input communication data converted to be operable as a radar signal, as described with respect to Figure 1.Depending on the selected JSC system type in which the BS DL Rx R-DNN203b is configured (e.g., bistatic, multistatic, or monostatic), the BS DL Rx R-DNN203b receives as input one or more radar sensing signals generated from either a) or b) (e.g., multistatic JSC), where a) BS 201 receives reflections 110a-110n of transmitted JSC signal waveforms from one or more objects 109a-109n (e.g., monostatic or multistatic JSC), where b) BS 201 receives RFB signal waveforms 117 transmitted from UE 212 corresponding to reflections 111a-111n of transmitted JSC signal waveforms received and processed by UE 212 (e.g., bistatic or multistatic), or both a) and b) (e.g., multistatic JSC). The BS DL Rx R-DNN203b processes one or more radar sensing signals to generate radar sensing information (e.g., Doppler, tracking, position, etc.) about one or more objects 109a~109n. The BS DNNC224 uses a selected JSC system type (e.g., bistatic, multistatic, or monostatic JSC) to obtain appropriate model parameters or weights for the BS DL Tx CR-DNN203a and BS DL Rx R-DNN203b from the BS DNN configuration store 225. In this example, the BS DNNC224 configures the BS JSC DNN structure 203 to perform either bistatic, multistatic, or monostatic JSC when running on one or more processors 221.

[0038] Similarly, UE212 is implemented as a computing system / device for performing any of the corresponding methods, JSC DNN operations or processes described herein, and / or for implementing any of the corresponding systems, units and / or devices described herein. UE212 includes interconnected RF front-end Tx / Rx subsystems 213, one or more transceivers 230, one or more processors 231, and a memory unit 233. Those skilled in the art will understand that other types of computing devices / systems / platforms may be used instead to perform UE212 and the methods described herein. UE212 includes one or more processors 221 (e.g., CPUs). One or more processors 221 control the operation of other components of UE212, such as the RF front-end Tx / Rx subsystems 213, one or more transceivers 230, and the memory unit 233. One or more processors 231 may be single-core devices or multi-core devices. One or more processors 231 may include a CPU and / or a GPU. Alternatively, one or more processors 231 may include special processing hardware, such as a RISC processor, or programmable hardware with built-in firmware. The UE212 may include multiple processors.

[0039] One or more processors 231 of the UE212 may be connected to a network interface such as a transceiver 230 including Tx and Rx for communicating via the RF front-end Tx / Rx subsystem 213 over a wireless communication channel 108 of the network with other devices and systems, such as BS201, other communication devices, network equipment, RAN entities or devices, users or operators, and / or any other devices, services, systems and / or devices requested by the application. Optionally, one or more processors 231 may be connected to a UI for user input to instruct or use the UE212 and / or the underlying computing system, and / or output data from there. Optionally, one or more processors 231 may be connected to a display for displaying output to the user.

[0040] The UE212 includes a memory system or memory unit 233, which includes working or volatile memory. One or more processors 231 may access the volatile memory to process data and control the storage of data into the memory. The volatile memory may include any type of RAM, such as SRAM, DRAM, etc., or flash memory, such as a secure digital card. The UE212 also includes non-volatile memory. The non-volatile memory may store a set of operation or operating system instructions and / or a set of software instructions in the form of computer-readable instructions for controlling the operation of the processor 231, which, when executed on one or more processors 231, cause the processor 231 to perform the corresponding methods, processes, operations and / or functions of the operation and / or methods of the JSC DNN in the UE212, as described herein. The non-volatile memory may be any type of memory, such as ROM, flash memory, secure digital drive, magnetic drive memory, or magnetic disk drive memory, depending on the requirements of the application.

[0041] The non-volatile memory of the UE212's memory unit 233 contains computer program code and / or instructions for executing the UE DNN controller (UE DNNC) 234 and / or the UE JSC DNN structure 214. When executed on one or more processors 231, the UE DNNC 234 uses the UE JSC DNN structure 214 and the UE DNN configuration store 235 stored in the memory unit 233 to control the JSC DNN operation in the UE212. While the UE DNNC 234 is part of the memory unit 233, this is merely an example, and the UE DNNC 234 is not limited in this way. Those skilled in the art will understand that the UE DNNC 234 may be executed on any combination of the UE212's hardware and / or software, and / or as required by the application. In this example, if the selected JSC system type is either bistatic or multistatic, the UE DNNC 234 configures the DL JSC UE JSC DNN structure 214 to include a UE DL receiver-side communication and radar DNN model (UE DL Rx CR-DNN) 214a, which is coupled to a UE DL transmitter-side radar feedback DNN model (UE DL Tx RFB-DNN) 214b. The UE DL Rx CR-DNN 214a processes the received JSC signal waveform corresponding to the JSC signal waveform transmission from BS201 to generate reconstructed communication data corresponding to the input communication data incorporated into the received JSC signal waveform. The UE DL Rx CR-DNN214a processes the radar sensing signal generated by the UE212, which receives one or more reflections 111a-111n from one or more objects 109a-109n of the JSC signal waveform transmission from BS201, and generates radar sensing feedback (RFB) information (e.g., Doppler, tracking, and / or position) for one or more objects 109a-109n.The UE DL Tx RFB-DNN214b processes RFB information about one or more objects 109a-109n and generates an RFB signal for transmission to BS201 as an RFB signal waveform 117 for use by BS201 (bistatic or multistatic JSC) when generating radar sensing information about one or more objects 109a-109n. The UE DNNC234 uses the selected JSC system type (e.g., bistatic or multistatic JSC) to configure the UE JSC DNN structure 214 for bistatic or multistatic DL JSCs by retrieving the model parameters and / or weights of the UE DL Rx CR-DNN214a and UE DL Tx RFB-DNN214b from the UE DNN configuration store 235. In this example, the UE JSC DNN structure 214 runs a bistatic or multistatic radar sensing application when executed on one or more processors 231.

[0042] Additionally or alternatively, when a monostatic JSC system type is selected, UE DNNC234 configures the UE JSC DNN structure 214 to run a monostatic JSC application by replacing UE DL Rx CR-DNN214a and UE DL Tx RFB-DNN214b with the UE DL receiver-side communication DNN model (UE DL Rx C-DNN) 214c when running on one or more processors 231. UE DNNC234 uses the selected JSC system type (e.g., monostatic JSC) to obtain the model parameters and / or weights of UE DL Rx C-DNN214c from the UE DNN configuration store 235 in order to configure the UE JSC DNN structure 214 for monostatic DL JSC. In a monostatic JSC application, BS201 does not use the RFB signal waveform 117 from UE212. Rather, the UE DL Rx C-DNN214c receives and processes the received JSC signal waveform, and / or one or more reflections 111a~111n of the JSC signal waveform, corresponding to the JSC signal waveform transmission from BS201, in order to generate reconstructed communication data corresponding to the input communication data incorporated into the received JSC signal waveform.

[0043] Optionally, UE212 can perform UL JSC, similar to BS201 which performs DL JSC, and UE DNNC234, when running on one or more processors 231, configures UE JSC DNN structure 214 so that UE212 performs JSC transmission and radar sensing in a UL JSC application. For example, in the case of UL JSC, UE JSC DNN structure 214 includes UE UL Transmitter-Side Communication and Radar DNN Model (UE UL Tx CR-DNN) 214d and UE UL Receiver-Side Radar DNN Model (UE UL Rx R-DNN) 214e. These UE UL DNN models 214d and 214e operate in a similar manner to the corresponding BS DL DNN models 203a and 203b when BS201 performs DL JSC with UE212, depending on the selected JSC system type (e.g., bistatic, multistatic, or monostatic). For UL JSC on BS201, BS DNNC224 configures BS JSC DNN structure 203 to execute UL JSC applications when running on one or more processors 221. For example, for UL JSC on BS201, BS JSC DNN structure 203 includes one or more of the following, depending on the selected JSC system type (e.g., bistatic, multistatic, or monostatic): BS UL receiver (Rx) communication and radar (CR) DNN model (BS UL Rx CR-DNN) 203c, BS UL transmitter (Tx) radar feedback (RFB) DNN model (BS UL Tx RFB-DNN) 203d, and / or BS UL receiver (Rx) communication (C) DNN model (BS UL Rx C-DNN) 203e. These BS UL DNN models 203c, 203d, and 203e operate in a similar manner to the corresponding UE DL DNN models 214a, 214b, and 214c when the UE212 runs BS201 and DL JSC, depending on the selected JSC system type (e.g., bistatic, multistatic, or monostatic).

[0044] Multiple trained pairs of JSC DNN structures, including different combinations of the above DNN models for use by BS201 and UE212 in JSC applications, are stored in a first neural network table (e.g., the first neural network table 1010 in Figure 10b) in the BS DNN configuration store 225, along with the corresponding channel characteristics / conditions and JSC system type (e.g., bistatic, multistatic, or monostatic JSC) used when training each trained pair of JSC DNN structures. Each trained pair of JSC DNN structures in the first neural network table is assigned a JSC DNN identifier / index. Similarly, UE212 stores a second neural network table (e.g., the second neural network table 1020 in Figure 10c) in the UE DNN configuration store 235, containing the trained JSC DNN structures used by UE212 from each trained pair of JSC DNN structures and their associated corresponding JSC DNN identifiers. In another example, the first neural network table of BS201 stores only the corresponding JSC DNN structure used by BS201 for each trained pair of JSC DNN structures, for each JSC DNN identifier. Similarly, the second neural network table of UE212 stores, for each JSC DNN identifier, the corresponding JSC DNN structure used by UE212 to match the other trained pair of JSC DNN structures.

[0045] In a further example, in addition to BS201 performing JSC DNN operation with UE212 using bistatic or multistatic JSC system types, BS201 also employs a third device, as described with reference to Figure 1, or a third device 341 in Figure 3d, to support bistatic or multistatic JSC.

[0046] In the example, BS201 transmits radar sensing information for one or more objects 109a-109n per time slot to one or more higher-layer protocols in BS201's protocol stack (for example, to the application protocol layer of the protocol stack for use by one or more applications running on BS201). Similarly, UE212 transmits reconstructed communication data to UE212's data sink, or transmits reconstructed communication data to one or more higher-layer protocols in UE212's protocol stack (for example, to the application protocol layer of the protocol stack for use by one or more applications running on UE212).

[0047] During operation, DL JSC DNN operation can be established when BS201 identifies the channel conditions of the DL communication channel (PDSCH) between BS201 and UE212 and selects a JSC system type according to JSC performance requirements. A suitable trained pair of JSC DNN structures is selected from a first neural network table using the identified channel conditions and the selected JSC system type, and BS201 sends a control message to UE212 containing fields associated with the corresponding JSC DNN identifier of the selected trained pair of JSC DNN structures. BS201 uses the JSC DNN identifier to configure BS JSC DNN structure 203 for DL ​​JSC DNN operation, depending on whether it obtains the corresponding BS JSC DNN structure for the selected trained pair of JSC DNN structures. Similarly, UE212 uses the JSC DNN identifier to configure UE JSC DNN structure 214 for DL ​​JSC DNN operation, depending on whether it obtains the corresponding UE JSC DNN structure for the selected trained pair of JSC DNN structures. In this example, the JSC system type is either bistatic or multistatic JSC, and the BS JSC DNN structure 203 includes a BS DL Tx CR-DNN203a coupled with a BS DL Rx R-DNN203b. The UE JSC DNN structure 214 includes a UE DL Rx CR-DNN214a coupled with a UE DL Tx RFB-DNN214b.

[0048] In this example, the control message uses DL control plane signaling (e.g., PDCCH) to specify the UE JSC DNN structure 214, one or more time slots, and associated communication resources. For example, the control message is either a Radio Resource Control (RRC) message or a Downlink Control Indicator (DCI) signal. The control message indicates a specific resource block (RB) / frequency to be used to transmit the JSC signal waveform 107 over the DL communication channel (e.g., PDSCH) for each time slot when the JSC DNN operation is performed. The UE 212 transmits the RFB signal waveform 117 corresponding to each time slot over the wireless communication channel 108. The UE 212 may, but is not limited to, using conventional UL communication control channels (e.g., PDCCH) to transmit the RFB signal waveform 117. Alternatively or additionally, the control message further includes a set of UL RBs / frequencies and subsequent UL time slots that the UE 212 can use when transmitting the RFB signal waveform 117 to BS 201 over the UL within a subsequent UL time slot. Each RFB signal waveform 117 corresponds to a specific DL time slot used by BS201 to transmit the JSC signal waveform 107.

[0049] Alternatively, if the pair of JSC DNN structures 203 and 214 of BS201 or UE212 outputs a JSC signal waveform 107 or RFB signal waveform 117 that does not use RB or a specific carrier frequency, the control message only needs to specify the JSC DNN identifier and one or more DL / UL time slots and / or subsequent DL / UL time slots when BS201 transmits each JSC signal waveform 107 via DL and when UE212 transmits the corresponding RFB signal waveform 117 via UL. After configuring BS201 and UE212 to perform DL JSC DNN operation during a communication session, BS201 and UE212 begin performing DL JSC DNN operation during the communication session in a manner similar to the JSC DNN operation described with reference to Figures 1 and 3a to 11.

[0050] In another example, for UL JSC DNN operation, BS201 instructs UE212 on the UL RB / frequency configuration and UL time slots that can be used when transmitting one or more JSC signal waveforms over the UL. During operation, the establishment of UL JSC DNN operation occurs when UE212 or BS201 identifies the channel conditions of the UL communication channel (e.g., PUSCH) between BS201 and UE212, and then identifies the JSC system type for UL JSC DNN operation according to the JSC performance requirements. In a similar manner to DL JSC DNN operation, a suitable trained pair of JSC DNN structures is selected using the identified channel conditions, the selected JSC system type, and the JSC performance requirements (see, for example, the first and second neural network tables 1010 and 1020 in Figure 10b or Figure 10c). BS DNNC224 obtains a BS JSC DNN structure from a trained pair of JSC DNN structures and configures a BS JSC DNN structure 203 using the obtained trained JSC DNN structures to perform UL JSC DNN operations within one or more time slots. UE DNNC234 obtains a UE JSC DNN structure from a trained pair of JSC DNN structures and configures a UE JSC DNN structure 214 using the obtained trained JSC DNN structures to perform UL JSC DNN operations within one or more time slots.

[0051] In another example, if UE212 selects a trained pair of JSC DNN structures intended to use a given channel condition, JSC system type, and JSC performance requirements, UE212 sends a control message to BS201 specifying the JSC DNN identifier of the selected trained pair of JSC DNN structures. BS201 uses the received JSC DNN identifier to retrieve the corresponding JSC DNN structures of the selected trained pair of JSC DNN structures from a first neural network table for use by BS201. UE DNNC234 retrieves the UE JSC DNN structure of the trained pair of JSC DNN structures and configures UE JSC DNN structure 214 with the retrieved trained JSC DNN structures to perform UL JSC DNN operation within one or more time slots. BS DNNC224 configures BS JSC DNN structure 203 with the retrieved trained JSC DNN structures to perform UL JSC DNN operation within one or more time slots. In response to the control message, BS201 sends a control message to UE212 specifying one or more communication resource parameters (e.g., DL / UL time slot and / or RB configuration, DL / UL frequency, etc.) for use by UE212 when transmitting JSC signal waveforms via UL, and for use by UE212 when receiving RFB signal waveforms 117 from BS201 via DL for use by UE212 when radar sensing associated with one or more objects 109~109n.

[0052] If the JSC system type is either bistatic or multistatic JSC for UL JSC DNN operation, the BS JSC DNN structure 203 includes a BS UL Rx CR-DNN203c coupled with a BS UL Tx RFB-DNN203d. The UE JSC DNN structure 214 includes a UE UL Tx CR-DNN214d coupled with a UE UL Rx R-DNN214e. After configuring BS201 and UE212 to perform UL JSC DNN operation during a communication session, BS201 and UE212 begin performing UL JSC DNN operation during the communication session in a similar manner to the JSC DNN operation described with reference to Figures 1 and 3a to 11.

[0053] In the case of BS201, at least one processor 221 comprises at least one memory unit 223 and computer program code or instructions stored therein, and is configured to cause at least a corresponding operation, method, and / or process disclosed with respect to operation, for example, a schematic diagram, flowchart, or operation describing any of Figures 1 to 12 and its associated features, to be executed by the computing system of BS201. In the case of UE212, at least one processor 231 comprises at least one memory unit 233 and computer program code or instructions stored therein, and is configured to cause at least a corresponding operation, method, and / or process disclosed with respect to operation, for example, a schematic diagram, flowchart, or operation describing any of Figures 1 to 12 and its associated features, to be executed by the computing system of UE212.

[0054] The JSC communication system 200, including BS201 and UE212, offers the advantage of performing accurate sensing (e.g., target object detection and / or range / velocity estimation) while simultaneously performing efficient JSC during communication sessions between BS201 and UE212 in 5G / 6G or later communication systems for various communication channel conditions, and meeting the performance requirements of the corresponding communication standards. Further advantages include efficiently designing JSC waveforms for use in 5G / 6G or later communication systems and for various sensing and communication channel conditions, while simultaneously performing accurate sensing and meeting the requirements of the corresponding communication standards. Additional advantages of the JSC communication system 200 and / or additional advantages described herein result in efficient, rapid, and dynamic reconfiguration of JSC DNN operation (e.g., bistatic, multistatic, monostatic JSC radar sensing) within communication sessions between BS201 and UE212 while maintaining accurate sensing and communication performance.

[0055] Wireless communication systems 100 or 200 are described with reference to Figure 1 or Figure 2 and / or as described herein, but this is merely an example and not limited in this way. Those skilled in the art will understand that any type of communication system or network, e.g., any telecommunication system or network, any wired communication network, any wireless communication network, satellite network, peer-to-peer communication network, communication system or network using technologies of third-generation (3G), fourth-generation (4G), 5G, and / or 6G or later standards, Wi-Fi communication network, optical communication network, fiber optic communication network, and / or any other network for communication between a first device and a second device, combinations thereof, modifications thereof, and / or as required by the application, are applicable. Although the first device 101 is described as BS201 with reference to Figure 2 and / or as described herein, this is merely an example and not limited in this way. Those skilled in the art will understand that the first device 101 may be any type of communication device capable of communicating with the second device 112, including, but not limited to, UE212, BS, satellite, mobile phone or smartphone, laptop, computing device, device using technology of 3G, 4G, 5G and / or 6G or later standards, and / or any other device used for communication with the second device 112, combinations thereof, modifications thereof, and / or as required by the application.Although the second device 112 is described as UE212 with reference to Figure 2 and / or as described herein, this is merely an example and not limited in this way. Those skilled in the art will understand that the second device 112 may be any type of communication device capable of communicating with the first device 101, including, but not limited to, UE212, BS, satellite, mobile phone or smartphone, laptop, computing device, device using 3G, 4G, 5G, and / or 6G or later standard technologies, and / or any other device used for communication with the first device 101, combinations thereof, modifications thereof, and / or as required by the application.

[0056] Figure 3a shows an example of a wireless communication system 300a comprising a first device 301 and a second device 312 for performing bistatic JSC. Figure 3a modifies the wireless communication systems 100 and 200 of Figures 1 and 2 by further defining the JSC DNN structures of the first device 101 and the second device 112, or BS201 and UE212, respectively. The first device 301 and the second device 312 have already established a communication session over the wireless communication / radar channel 308, as described herein with reference to Figures 1 or 2 and / or Figures 4 to 12. The first device 301 and the second device 312 establish and perform bistatic JSC DNN operation within one or more specific time slots of the communication session over the wireless communication / radar channel 308.

[0057] Bistatic JSC DNN operation is established when the first device 301 selects a trained pair of JSC DNN structures corresponding to the bistatic JSC system type. The selected trained pair of JSC DNN structures includes a first trained JSC DNN structure with a transmitter / receiver DNN model configured for bistatic JSC, and a second trained JSC DNN structure. The first trained JSC DNN structure includes a trained transmitter communication and radar DNN model (Tx CR-DNN) 303a coupled to a trained receiver radar DNN model (Rx R-DNN) 303b for configuring the JSC DNN structure 303 of the first device 301 for bistatic JSC DNN operation. The second trained JSC DNN structure includes a trained receiver-side communication and radar DNN model (Rx CR-DNN) 314a coupled to a trained transmitter-side radar feedback DNN model (Tx RFB-DNN) 314b for configuring the JSC DNN structure 314 of the second device 312 in bistatic JSC DNN operation. The Tx CR-DNN 303a and Rx R-DNN 303b of the first JSC DNN structure and the Rx CR-DNN 314a and Tx RFB-DNN 314b of the second JSC DNN structure are jointly trained and stored for selection by the first device 301, as described with reference to, for example, Figures 10a to 11.

[0058] In this example, the first device 301 constitutes a JSC DNN structure 303 with a trained Tx CR-DNN 303a coupled to a trained Rx R-DNN 303b. When the Tx CR-DNN 303a performs radar sensing for a specific time slot out of one or more time slots, it generates an optional feedforward communication and radar (FFCR) signal 303c (or feedforward signal) which is passed to the Rx R-DNN 303b. The FFCR signal 303c includes, but is not limited to, one or more of the following: input communication data 302 for a specific time slot, output JSC signal 306a for a specific time slot, inputs to one or more neural network layers of the Tx CR-DNN 303a, outputs to one or more neural network layers of the Tx CR-DNN 303a, and / or combinations thereof. The trained Tx CR-DNN303a receives input communication data 102 as input for transmission within a specific time slot and generates output JSC signal 306a representing the input communication data 102 and the radar signal for that specific time slot.

[0059] The first device 301 transmits an output JSC signal 306a within a specific time slot as a JSC signal waveform 307 via the DAC / RF Tx antenna component 304a of the RF front-end Tx / Rx subsystem 304. The first device 301 transmits the JSC signal waveform 307 within a specific time slot to the second device 312 via the wireless communication / radar channel 308. The transmission of the JSC signal waveform 307 within a specific time slot represents the input communication data 302 and radar signal for transmission to the second device 312. Within the environment of the wireless communication / radar channel 308, there are one or more objects 309a~309n (e.g., people, animals, obstacles, trees, buildings, vehicles, and / or any other objects that can reflect the transmitted JSC signal waveform 307).

[0060] The second device 312 includes an RF front-end Tx / Rx subsystem 313 having an RF Rx antenna / ADC (RF Rx / ADC) component 313b coupled to a JSC DNN structure 314. The RF Rx / ADC component 313b receives, within a specific time slot, the transmitted JSC signal waveform 307, or at least one component 307a of the JSC signal waveform 307, and one or more further reflected JSC signal waveforms 311a, 311b, and 311c-311n based on reflections of the JSC signal waveform 307 from one or more corresponding objects 309a, 309b, and 309c-309n in the environment of the radio communication / radar channel 308. The RF Rx / ADC component 313b (e.g., a frequency downconverter / ADC to baseband) processes the received JSC signal waveform 307 and / or its component 307a and one or more further reflected JSC signal waveforms 311a, 311b, and 311c-311n to form a JSC signal 316a for a specific time slot.

[0061] The JSC DNN structure 314 is coupled to the RF front-end Tx / Rx subsystem 313. The second device 312 configures the JSC DNN structure 314 to include a trained Rx CR-DNN 314a coupled to a trained Tx RFB-DNN 314b. The JSC signal 316a for a specific time slot is input to the Rx CR-DNN 314a of the JSC DNN structure 314. The Rx CR-DNN 314a processes the JSC signal 316a and generates reconstructed communication data 315 for the specific time slot, as well as radar sensing feedback (RFB) information 314d for the specific time slot. The reconstructed communication data 315 for the specific time slot corresponds to the input communication data 102 transmitted within the JSC signal waveform 307 during the specific time slot. The RFB information 314d represents, but is not limited to, one or more of the following radar measurements / parameters for one or more objects 309a, 309b, and 309c-309n within the environment of the wireless communication / radar channel 308: range, Doppler, velocity, position, azimuth, elevation, or other radar measurements. The generated RFB information 314d is input to Tx RFB-DNN 314b. Tx RFB-DNN 314b processes the RFB information 314d (e.g., range, Doppler, velocity, position, or other radar measurements) and generates an RFB signal 316b corresponding to a specific time slot for transmission to the first device 301.

[0062] The RF front-end Tx / Rx subsystem 313 of the second device 312 also includes a DAC / RF Tx antenna (DAC / RF Tx) component 313a coupled to the Tx RFB-DNN 314b of the JSC DNN structure 314. The DAC / RF Tx component 313a receives and processes the RFB signal 316b for transmission to the first device 301 as an RFB signal waveform 317 corresponding to a specific time slot. This assists the trained JSC DNN structure 303 of the first device 301 in generating radar sensing information (e.g., range, Doppler, velocity, position, or other radar measurements) corresponding to one or more objects 309a-309n in the environment of the wireless communication / radar channel 308 for a specific time slot.

[0063] The first device 301 receives and processes an RFB signal waveform 317 corresponding to a specific time slot via the RF Rx / ADC component 304b of the RF front-end Tx / Rx subsystem 304, as a radar sensing signal 306b corresponding to a specific time slot. The radar sensing signal 306b and FFCR signal 303c corresponding to a specific time slot are input to the Rx R-DNN 303b, which generates radar sensing information 305 (e.g., range, Doppler, velocity, position, or other radar measurements / parameters) for one or more objects 309a, 309b, and 309c-309n of the specific time slot.

[0064] Figure 3b shows an example of a wireless communication system 300b with the first device 301 and the second device 312 of Figure 3a configured for multistatic JSC. Figure 3b modifies the wireless communication systems 100, 200, and 300a of Figures 1, 2, and 3a into multistatic JSC system types by further configuring the JSC DNN structure of the first device 101 and the second device 112, BS201 and UE212, or the first device 301 and the second device 312 of Figure 3a, respectively. In this example, the first device 301 and the second device 312 perform JSC using a co-trained DNN when the JSC system type is multistatic JSC.

[0065] In a manner similar to that shown in Figure 3a, a selected trained pair of JSC DNN structures for a multi-static JSC system type includes a first trained JSC DNN structure and a second trained JSC DNN structure, each featuring a transmitter / receiver DNN model configured for multi-static JSC. The first trained JSC DNN structure includes a trained Tx CR-DNN303a coupled to an Rx R-DNN303b trained to constitute the JSC DNN structure 303 of the first device 301, and is trained for multi-static JSC DNN operation. The second trained JSC DNN structure includes a trained Rx CR-DNN314a coupled to an Rx RFB-DNN314b trained to constitute the JSC DNN structure 314 of the second device 312, and is trained for multi-static JSC DNN operation. The Tx CR-DNN303a and Rx R-DNN303b of the first JSC DNN structure and the Rx CR-DNN314a and Tx RFB-DNN314b of the second JSC DNN structure are jointly trained for multistatic JSC, as described with reference to, for example, Figures 10a to 11, and stored for selection by the first device 301.

[0066] The multistatic JSC system type is similar to the bistatic JSC system type described with respect to Figure 3a, with the additional modification that the first device 301 operates in full-duplex communication mode to respond to the transmission of a JSC signal waveform 307 within a specific time slot and to receive multiple reflected JSC signal waveforms 310a, 310b, and 310c-310n reflected from one or more corresponding objects 309a, 309b, and 309c-309n for each specific time slot. The first device 301 receives and processes the reflected JSC signal waveforms 310a, 310b, and 310c-310n, as well as an RFB signal waveform 317 corresponding to a specific time slot, via the RF Rx / ADC component 304b of the RF front-end Tx / Rx subsystem 304, in order to generate one or more radar sensing signals 306b corresponding to a specific time slot. The trained Rx R-DNN303b receives multiple radar sensing signals 306b as input. The radar sensing signals 306b and FFCR signals 303c corresponding to a particular time slot are input to the trained Rx R-DNN303b, which generates radar sensing information 305 (e.g., range, Doppler, velocity, position, or other radar measurements / parameters) for one or more objects 309a, 309b, and 309c-309n of that particular time slot.

[0067] Since the radar sensing signal 306b and the RFB signal waveform 317 are multiple inputs that are reflected from the JSC signal waveforms 310a, 310b, and 310c-310n, it should be noted that the trained Rx R-DNN 303b will have at least different layers, nodes, weights, and biases compared to the trained Rx R-DNN 303b in Figure 3a. The second device 312 is configured identically to the bistatic radar sensing configuration in Figure 3a and operates in the same manner as described with reference to Figure 3a.

[0068] Figure 3c shows an exemplary wireless communication system 300c with the first device 301 and the second device 312 of Figure 3a arranged for monostatic JSC. Figure 3c modifies the wireless communication systems 100, 200, 300a, and 300b of Figures 1, 2, 3a, and 3b into monostatic JSC system types by further configuring the JSC DNN structure of the first device 101 and the second device 112, BS201 and UE212, or the first device 301 and the second device 312, respectively. In this example, the first device 301 and the second device 312 perform JSC using a co-trained DNN when the JSC system type is monostatic JSC.

[0069] In a manner similar to that shown in Figure 3b, a selected trained pair of JSC DNN structures for a monostatic JSC system type includes a first trained JSC DNN structure with a transmitter / receiver DNN model configured for monostatic JSC, and a second trained JSC DNN structure. The first trained JSC DNN structure includes a trained Tx CR-DNN 303a coupled to an Rx R-DNN 303b trained to constitute the JSC DNN structure 303 of the first device 301, and is trained for monostatic JSC DNN operation. However, for monostatic JSC, the second trained JSC DNN structure includes only a receiver-side communication DNN (Rx C-DNN) 314c trained to constitute the JSC DNN structure 314 of the second device 312, and is trained for monostatic JSC DNN operation. The Tx CR-DNN303a and Rx R-DNN303b of the first JSC DNN structure and the Rx C-DNN314c of the second JSC DNN structure are jointly trained for monostatic JSC, as described with reference to, for example, Figures 10a to 11, and stored for selection by the first device 301.

[0070] The monostatic JSC system type is a subset of the multistatic JSC system type described with reference to Figure 3b, but with additional modifications, the first device 301 operates in full-duplex communication mode to respond to the transmission of a JSC signal waveform 307 within a specific time slot and to receive only multiple reflected JSC signal waveforms 310a, 310b, and 310c-310n reflected from one or more corresponding objects 309a, 309b, and 309c-309n for each specific time slot. The second device 312 does not provide an RFB signal waveform 317 as described with reference to Figures 3a and 3b. Instead, the first device 301 receives and processes reflected JSC signal waveforms 310a, 310b, and 310c-310n corresponding to a particular time slot via the RF Rx / ADC component 304b of the RF front-end Tx / Rx subsystem 304 to generate one or more radar sensing signals 306b corresponding to a particular time slot. The trained Rx R-DNN 303b is configured to receive multiple radar sensing signals 306b as input. The radar sensing signals 306b and FFCR signals 303c corresponding to a particular time slot are input to the trained Rx R-DNN 303b, which generates radar sensing information 305 (e.g., range, Doppler, velocity, position, or other radar measurements / parameters) for one or more objects 309a, 309b, and 309c-309n of the particular time slot.

[0071] Similarly, the RF Rx / ADC component 313b of the second device 312 receives the transmitted JSC signal waveform 307 within a specific time slot. The RF Rx / ADC component 313b (e.g., a frequency downconverter / ADC to baseband) processes the received transmitted JSC signal waveform 307 into a JSC signal 316a for the specific time slot. The received JSC signal 316a for the specific time slot is input to a trained Rx C-DNN 314c that generates only the reconstructed communication data 315 for the specific time slot. Optionally, the second device 312 can also, via the RF Rx / ADC component 313b, respond to the transmission of a JSC signal waveform 307 within a specific time slot and further receive and process multiple reflected JSC signal waveforms 311a, 311b, and 311c-311n, which are reflected by the Rx C-DNN 314c (or coupled to the JSC signal 316a) to generate reconstructed communication data 315 from one or more corresponding objects 309a, 309b, and 309c-309n for input as multiple JSC signals 316a.

[0072] Figure 3d shows an example of a wireless communication system 300d comprising a first device 301 and a second device 312 from Figure 3a or 3b, arranged for bistatic or multistatic JSC, and a third device 341. Figure 3d modifies the wireless communication systems 100, 200, 300a, and 300b from Figures 1, 2, 3a, and 3b, where the third device 341 uses a modified JSC DNN structure to support the first device 301 in bistatic or multistatic JSC. In this example, the first device 301 and the second device 312 perform JSC using a DNN, with the JSC system type being bistatic or multistatic JSC, as described with reference to Figures 3a and 3b. However, the third device 341 operates only in a bistatic radar sensing configuration and does not reconstruct communication data targeting the second device 312. Instead, the RF front-end Tx / Rx342 of the third device 341 receives, within a specific time slot, the transmitted JSC signal waveform 307 or its component 307a and one or more further reflected JSC signal waveforms 340a, 340b, and 340c-340n based on the reflection of the JSC signal waveform 307 transmitted from one or more corresponding objects 309a, 309b, and 309c-309n. The RF front-end Tx / Rx342 processes the received transmitted JSC signal waveform 307 and any one or more further received reflected JSC signal waveforms 340a-340n to form the received JSC signal 346a for the specific time slot. The JSC DNN structure of the third device 341 is a radar sensing JSC DNN structure 343 (radar DNN structure) which includes only a trained receiving radar DNN model (Rx R-DNN) coupled to a trained Tx RFB-DNN model. The Rx R-DNN of the radar DNN structure 343 processes the received JSC signal 346a and generates RFB information for a specific time slot. The generated RFB information is input to the Tx RFB-DNN of the radar DNN structure 343.The Tx RFB-DNN of the radar DNN structure 343 processes RFB information (e.g., range, Doppler, velocity, position, or other radar measurements) and generates an RFB signal 346b for a specific time slot. The RF front-end Tx / Rx 342 processes the RFB signal 346b for transmission to the first device 301 as an RFB signal waveform 345 for a specific time slot. This additional RFB signal waveform 345 is received by the first device 301 in a similar manner to how the RFB signal waveform 317 is received by the second device 312 for a specific time slot, as described with reference to Figures 3a and 3b. The additional RFB signal waveform 345 assists the trained JSC DNN structure 303 of the first device 301 in generating radar sensing information (e.g., range, Doppler, velocity, position, or other radar measurements) similar to the RFB signal waveform 317 from the second device 312 in Figures 3a and 3b (e.g., range, Doppler, velocity, position, or other radar measurements).

[0073] Figures 3a to 3c illustrate and describe three JSC system types, namely bistatic JSC, multistatic JSC, and monostatic JSC, while Figure 3d shows a combination of two JSC system types (e.g., bistatic JSC or multistatic JSC) using a first device 301, a second device 312, and a third device 341. In the following disclosure, Figures 4 to 6 show and describe signaling diagrams for three JSC types, bistatic JSC, multistatic JSC, and monostatic JSC, as merely examples, and those skilled in the art will understand that the processes and / or systems described by the signaling diagrams can be further modified and / or combined based on a combination of two or more JSC system types, as described and / or depending on the requirements of the application.

[0074] Figure 4 shows a signal flow diagram of an exemplary bistatic JSC DNN operation 400 for one or more predetermined time slots during a communication session between the first device 101 and the second device 112. The first device 101 and the second device 112 in Figure 1 perform the bistatic JSC DNN operation 400 using any of the bistatic JSC embodiments described with reference to Figures 1, 2, 3a, 3b, and / or 3d. The signal flow of the bistatic JSC DNN operation 400 for a communication session between the first device 101 and the second device 112 includes the following signal flow operation.

[0075] In operation 402 / 422, the first device 101 and the second device 112 establish bistatic JSC DNN communication with each other. That is, the first device 101 establishes JSC DNN operation for one or more time slots of a communication session with the second device 112. While establishing the bistatic DNN connection, the first device 101 and the second device 112 communicate with each other to define, agree, and / or configure the type of JSC DNN structure that each of the first device 101 and the second device 112 will use, including the transmit / receive DNN models used to perform end-to-end JSC between them. As illustrated with reference to Figures 2, 3a, 3b, and 3d, the JSC DNN structure of the first device 101 includes a Tx CR-DNN model and an Rx R-DNN model. As illustrated with reference to Figures 2, 3a, 3b, and 3d, the JSC DNN structure of the second device 112 includes an Rx CR-DNN model and a Tx RFB-DNN model. In this example, the first device 101 identifies the channel conditions / characteristics of the communication / radar channel between the first device 101 and the second device 112, which the first device 101 uses to select a pair of bistatic JSC system type JSC DNN structures from the first neural network table or the second neural network table (provided to the first device 101 or the second device 112, respectively), as illustrated with reference to Figures 10a-10c. For example, after the first device 101 initiates a conventional or standard communication session with the second device 112, the first device 101 selects a pair of JSC DNN structures for use in a bistatic JSC DNN connection with the second device 112, depending on the communication / radar channel conditions / environment, the radar and / or communication performance requirements of the JSC DNN connection (e.g., JSC performance requirements), and the data communication characteristics (e.g., voice communication, data communication, multimedia streaming, QoS parameters, network slicing parameters, etc.).In the example, the first device 101 requests the machine learning processing capabilities of the second device 112 to assist the first device 101 in selecting a pair of JSC DNN structures for the first device 101 and the second device 112 for bistatic JSC DNN communication / operation.

[0076] After selecting a pair of JSC DNN structures for a bistatic DNN connection, the first device 101 initiates the bistatic DNN connection by sending a bistatic JSC DNN connection request message to the second device 112, in which fields specify the bistatic JSC DNN connection and JSC DNN indicator corresponding to the selected pair of JSC DNN structures. The JSC DNN indicator allows the second device 112 to select the corresponding bistatic JSC DNN structure to assist the first device 101 in performing radar sensing of one or more objects 109a-109n. For example, the first device 101 sends one or more DL control messages to the second device 112 indicating the selected type of JSC DNN structure (e.g., a selected unique JSC DNN identifier) ​​for use by the second device 112 when processing the transmitted JSC signal waveform.

[0077] In operation 403, the first device 101 configures its JSC DNN structure based on a selected pair of JSC DNN structures or a selected unique JSC DNN identifier. The selected JSC DNN structure of the first device 101 performs JSC transmission and radar sensing operations. In this example, as illustrated with reference to Figures 1, 2, 3a, 3b, and 3d, the first device 101 configures its JSC DNN structure to include a Tx CR-DNN model and an Rx-R-DNN model. In operation 423, the second device configures its JSC DNN structure based on a selected pair of JSC DNN structures or a selected unique JSC DNN identifier. The selected JSC DNN structure of the second device performs communication data reception and radar feedback operations. For example, in operation 423, when the first device 101 receives one or more control messages indicating the type of JSC DNN structure selected for use by the second device 112 when processing the received JSC signal waveform, the second device 112 configures its JSC DNN structure based on the type of JSC DNN structure selected for use within one or more time slots. In this example, as illustrated with reference to Figures 1, 2, 3a, 3b, and 3d, the second device 112 configures its JSC DNN structure to include an Rx CR-DNN model and a Tx RFB-DNN model.

[0078] For each time slot of one or more predetermined time slots, the first device 101 and the second device 112 perform the following operations: In operation 404, the first device 101 acquires input communication data for transmission to the second device 112 within the time slot. In operation 406, the first device 101 processes the input communication data using the Tx CR-DNN of the first device's JSC DNN structure to generate an output JSC signal representing the input communication data and radar signal. In operation 408, the first device 101 transmits the output JSC signal to the second device 112 as a Tx JSC signal waveform within the time slot via the communication channel.

[0079] In operation 424, the second device 112 receives the Tx JSC signal waveform transmitted from the first device 101 within a time slot via the communication channel. The received Tx JSC signal waveform represents the input communication data and radar signals transmitted within the time slot. In operation 426, the second device (SD) 112 also receives one or more objects 109a~109n (e.g., O1, O2, ..., O N Based on the reflection of the Tx JSC signal waveform transmission from ) one or more further JSC signal waveforms in the time slot (e.g., O1SD JSC signal waveform, O2SD JSC signal waveform, ..., O NThe SD JSC signal waveform is received. In operation 428, the second device 112 processes the received Tx JSC signal waveform of the time slot and one or more further received JSC signal waveforms using an Rx CR-DNN model of a JSC DNN structure in the second device 112 to generate reconstructed communication data corresponding to the input communication data transmitted within the time slot and radar sensing feedback (RFB) information (e.g., range, Doppler, velocity, etc.) associated with one or more objects 109a-109n. In operation 430, the Tx RFB-DNN processes the RFB information of the time slot to generate an RFB signal for transmission to the first device 101. In operation 432, the second device 112 transmits the RFB signal as a Tx RFB signal waveform (e.g., RFB signal waveform) to the first device 101 for use by the first device 101 in generating radar sensing information about one or more objects 109a-109n. In operation 434, the second device 112 sends the reconstructed time slot communication data to the data sink. For example, the second device 112 sends the reconstructed time slot communication data to one or more higher protocol layers of the second device 112's protocol stack. For example, the second device 112 sends the reconstructed communication data to the application protocol layer of the second device 112's protocol stack for use by one or more applications running on the second device 112. In operation 436, the second device 112 proceeds to operations 424 / 426 to receive other Tx JSC signal waveforms transmitted from the first device 101 within the next time slot (or subsequent time slots) via the communication channel. For example, the next time slot is a subsequent time slot designated by the first device 101 for performing a bistatic JSC DNN operation.

[0080] In operation 410, the first device 101 receives the Tx RFB signal waveform transmitted by the second device 112 in the time slot as a received radar sensing signal, which includes information representing RFB information associated with one or more objects 109a-109n, generated by the second device 112 in operations 428-430. The Rx R-DNN of the JSC DNN structure of the first device 101 processes the received radar sensing signal to generate radar sensing information for one or more objects 109a-109n in the communication channel. In operation 414, the first device 101 transmits the radar sensing information for one or more objects 109a-109n in the time slot to one or more higher-layer application protocols in the protocol stack of the first device 101. For example, the first device 101 transmits radar sensing information for one or more objects 109a to 109n for each time slot to one or more higher-layer protocols of the first device 101's protocol stack (for example, to the application protocol layer of the protocol stack for use by one or more applications running on the first device 101). One or more applications may run on the first device and / or one or more other devices and may access the radar sensing information for one or more objects 109a to 109n accordingly. For example, the second device 112 transmits the reconstructed communication data for the time slot to one or more higher-layer protocols of the second device 112's protocol stack. For example, the second device 112 transmits the reconstructed communication data to the application protocol layer of the second device 112's protocol stack for use by one or more applications running on the second device 112.

[0081] In operation 416, the first device 101 uses the bistatic JSC operation to determine if there are any further predetermined time slots available for transmitting input communication data. If so, the signal flow of the bistatic JSC DNN operation 400 proceeds to operation 404 for transmitting further input communication data in the next time slot. If there are no further predetermined time slots available for using the bistatic JSC DNN operation 400, the signal flow of the bistatic JSC DNN operation 400 proceeds to operation 418. In operation 418, the first device 101 sends a control message to the second device to disable bistatic JSC communication. The first device 101 disables bistatic JSC communication. In operation 438, upon receiving the control message to disable bistatic JSC communication, the second device 112 also disables bistatic JSC communication.

[0082] In operation 439, the first device 101 and the second device 112, in response to the application's request, continue the communication session using conventional communication and / or terminate the communication session.

[0083] Figure 5 shows a signal flow diagram of an exemplary multistatic JSC DNN operation 500 for one or more predetermined time slots during a communication session between the first device 101 and the second device 112. The bistatic JSC DNN operation 400 in Figure 4 is further modified so that the first device 101 and the second device 112 are configured to perform the multistatic JSC configuration described with reference to Figures 1, 2, 3a, 3b, and 3d. The signal flow of the multistatic JSC DNN operation 500 for a communication session between the first device 101 and the second device 112 includes the following signal flow operation.

[0084] In operation 502 / 522, the first device 101 and the second device 112 establish JSC DNN communication with each other, similar to operation 402 / 422 in Figure 4. In this case, the first device 101 establishes a multistatic JSC DNN operation in one or more predetermined time slots of the communication session with the second device 112. During establishment, the first device 101 selects a pair of JSC DNN structures from a first or second neural network table (described with reference to, for example, Figures 10a-10c) associated with the multistatic JSC system type and one or more communication channel conditions / characteristics identified by the first device 101 with respect to the communication channel. As described with reference to Figure 4, the selected pair of JSC DNN structures includes a similar JSC DNN structure of the first device 101 and a similar JSC DNN structure of the second device 112, and the corresponding models are jointly trained for the multistatic JSC system type.

[0085] Operations 503 and 523 substantially correspond to operations 403 and 523 in Figure 4, except that the trained Tx CR-DNN model and the trained Rx R-DNN model perform JSC transmission and multistatic radar sensing (instead of bistatic JSC).

[0086] For each time slot of one or more predetermined time slots, the first device 101 and the second device 112 perform multistatic JSC operations 504 to 508, which substantially correspond to operations 404 to 408 in Figure 4. Operations 524 to 536 substantially correspond to operations 424 to 436 in Figure 4. In operation 532, the second device 112 transmits a radar sensing feedback (RFB) signal as a Tx RFB signal waveform to the first device 101 for use by the first device 101 when generating multistatic radar sensing information for one or more objects 109a to 109n.

[0087] In operation 510, the first device 101 receives the Tx RFB waveform transmitted by the second device 112 in the time slot. In contrast to operation 410 in Figure 4, the first device (FD) 101 receives the Tx JSC signal waveform from one or more objects 109a~109n (e.g., O1, O2, ..., O N Because it is reflected from ), multiple reflected Tx JSC signal waveforms (e.g., O1FD JSC signal waveform, O2FD JSC signal waveform, ..., O N The RF front-end Tx / Rx of the first device 101 receives the received Tx RFB waveform transmitted from the second device 112 in the time slot and multiple reflected Tx JSC signal waveforms received in the time slot (e.g., O1FD JSC signal waveform, O2FD JSC signal waveform, ...O N In operation 512, the Rx R-DNN of the JSC DNN structure of the first device 101 processes the received radar sensing signals (e.g., reflected FD JSC signal waveform and Tx RFB signal waveform) to generate radar sensing information for one or more objects 109a-109n in the communication channel. In operation 514, the first device 101 transmits the radar sensing information for one or more objects 109a-109n for a time slot to one or more upper layers of the protocol stack of the first device 101. For example, the first device 101 transmits the radar sensing information for one or more objects 109a-109n for this time slot to one or more upper layer protocols of the protocol stack of the first device 101 (e.g., to the application protocol layer of the protocol stack for use by one or more applications running on the first device 101). One or more applications may run on a first device and / or one or more other devices, and may, accordingly, access radar sensing information of one or more objects 109a to 109n.

[0088] In operation 516, the first device 101 uses the multistatic JSC operation to determine if there are any further time slots in a given time slot for transmitting input communication data. If so, the signal flow of the multistatic JSC DNN operation 500 proceeds to operation 504 for transmitting further input communication data in the next time slot. If there are no further predetermined time slots, the signal flow of the multistatic JSC DNN operation 500 proceeds to operations 518, 538, and 539, corresponding to operations 418, 438, and 439 in Figure 4.

[0089] Figure 6 shows a signal flow diagram of an exemplary monostatic JSC DNN operation 600 for one or more predetermined time slots during a communication session between a first device 101 and a second device 112. Monostatic JSC is a subset of multistatic JSC as described with reference to Figure 5. In monostatic JSC, the first device 101 receives reflections of the JSC signal waveform transmitted from the first device 101 to generate radar sensing information, and the second device 112 generates only reconstructed data from the received JSC signal waveform transmitted by the first device 101 and / or its reflections from one or more objects 109a-109n. The multistatic JSC DNN operation 500 in Figure 5 is further modified so that the first device 101 and the second device 112 are configured to perform the monostatic JSC configuration described with reference to Figures 1, 2, and 3c. The signal flow of the monostatic JSC DNN operation 600 for the communication session between the first device 101 and the second device 112 includes the following signal flow operation:

[0090] In operation 602 / 622, the first device 101 and the second device 112 establish JSC DNN communication with each other, similar to operation 502 / 522 in Figure 5. In this case, the first device 101 establishes monostatic JSC DNN operation in one or more predetermined time slots of the communication session with the second device 112. As illustrated with reference to Figures 1, 2, and 3c, the selected pair of JSC DNN structures includes a similar JSC DNN structure for the first device 101 and a similar JSC DNN structure for the second device 112, and the corresponding DNN models are jointly trained for the monostatic JSC system type.

[0091] Operations 603 and 623 substantially correspond to operations 503 and 523 in Figure 5, except that the trained Tx CR-DNN model and trained Rx R-DNN model of the selected pair of JSC DNN structures in the first device 101 perform JSC transmission and monostatic JSC (rather than multistatic JSC). The trained Rx C-DNN model of the selected pair of JSC DNN structures in the second device 112 performs reconstruction of any input communication data transmitted from the first device 101 in Tx JSC signal waveforms for each time slot, and / or its reflected versions from one or more objects 109a-109n.

[0092] For each time slot of one or more predetermined time slots, the first device 101 and the second device 112 perform monostatic JSC operations 604 to 608, which substantially correspond to operations 504 to 508 in Figure 5, in which Tx JSC signal waveforms representing input communication data and radar signals are transmitted over the communication channel within the time slot.

[0093] In operation 624, the second device 112 receives the Tx JSC signal waveform of the time slot. In operation 632, the second device 112 processes the received Tx JSC signal waveform of the time slot using the Rx C-DNN model of the JSC DNN structure of the second device 112 to generate reconstructed communication data corresponding to the input communication data transmitted within the Tx JSC signal waveform of the time slot. Optionally, the Rx-C-DNN model can also process any received reflections of the Tx JSC signal waveform transmitted from one or more objects 109a-109n once it has generated the reconstructed communication data for the time slot. In operation 634, the second device 112 transmits the reconstructed communication data for the time slot to the data sink of the second device 112 or to one or more higher protocol layers of the protocol stack of the second device 112. Operation 636 substantially corresponds to operations 436 or 536 in Figure 4 or 5, respectively.

[0094] Operations 610-616 in Figure 6 substantially correspond to operations 510-516 in Figure 5, but the Rx R-DNN of the JSC DNN structure of the first device 101 generates monostatic radar sensing information for one or more objects 109a-109n in the communication channel of the time slot, using the transmitted Tx JSC signal waveform, e.g., O1FD JSC signal waveform, O2FD JSC signal waveform, ...O NAs a result of the FD JSC signal waveform, only the multiple received reflected Tx JSC signal waveforms are processed to make the received radar sensing signal for input to the Rx R-DNN. In operation 614, the first device 101 transmits monostatic radar sensing information for one or more objects 109a~109n in a time slot to one or more higher-layer protocols in the protocol stack of the first device 101. In operation 616, the first device 101 uses the monostatic JSC DNN operation to determine if there are further time slots in a given time slot for transmitting input communication data. If so, the signal flow of monostatic JSC DNN operation 600 proceeds to operation 604 to transmit further input communication data in the next time slot. If there are no further given time slots in monostatic JSC DNN operation 600, the signal flow of monostatic JSC DNN operation 600 proceeds to operations 618, 638 and 639, corresponding to operations 418, 438 and 439 in Figure 4.

[0095] Figure 7a shows a signal flow diagram of an exemplary bistatic JSC DNN operation 700a for one or more predetermined time slots (TS) (e.g., TS{a, b, c, d}) of a communication session between a first device 201 and a second device 212 in a wireless communication system 200. In this example, the first device 201 is a BS called BS201, and the second device 212 is a UE called UE212. Although this example describes the first device 201 as BS201 and the second device 212 as UE212, this is merely an example and is not limited thereto. Those skilled in the art will understand that the first device 201 and the second device 212 can be any type of communication device for use in the wireless communication system 200, including, for example, a BS, a network device, a UE, or any combination of radio access network elements including other RAN elements in the wireless communication system 200. For example, the first device 201 and the second device 212 may be two BSs, or two UEs, or one BS and one UE, or one UE and one BS, or any other combination of communication devices, depending on the requirements of the application. Optionally, the wireless communication system 200 may include one or more other communication devices (e.g., a third device as described in Figure 3d) to support bistatic radar sensing of BS201 per time slot during the bistatic JSC DNN operation 700a in Figure 7a.

[0096] In this example, the data communication channel includes a DL data channel, e.g., PDSCH for transmitting JSC signal waveforms from BS201 to UE212, and a DL control channel, e.g., PDCCH for establishing bistatic JSC DNN operation during a communication session between BS201 and UE212. The data communication channel also includes a UL data channel, e.g., PUSCH for transmitting radar sensing feedback (RFB) signal waveforms 117 from UE212 to BS201. Alternatively, UE212 may use a UL control channel, e.g., PUCCH for transmitting RFB signal waveforms 117 from UE212 to BS201, instead of using the UL data channel.

[0097] BS201 includes a BS DNNC224 connected to a BS JSC DNN structure 203 that performs bistatic JSC DNN operation as described with reference to Figures 1, 2, 3a, 3b, 4, and 5. Similarly, UE212 includes a UE DNNC234 connected to a UE JSC DNN structure 214 that also performs bistatic JSC DNN operation as described with reference to Figures 1, 2, 3a, 3b, 4, and 5.

[0098] Operations 702 and 722 are similar to operations 402 and 422 in Figure 4, in which BS DNNC224 establishes a DL bistatic JSC DNN communication session between BS201 and UE212. In operation 702, BS DNNC224 determines the PDSCH channel characteristics / conditions and obtains a unique JSC DNN identifier (e.g., DNN ID1) that identifies from the neural network table a pair of JSC DNN structures to be provided to BS201 for a bistatic radar system type that addresses known channel conditions, communication performance requirements, and / or radar performance requirements. The selected pair of JSC DNN structures includes the BS JSC DNN structure and the UE JSC DNN structure for the bistatic JSC system type. Optionally, BS201 receives UE capability information (e.g., processing capabilities such as central processing unit or graphics processing unit capability, memory, DNN capability, operating system and software version) and / or UE support information (e.g., power saving (low battery), battery level, processor utilization, thermal status, and / or temperature threshold / overheating) from UE212. BS DNNC224 uses the UE capability and / or support information to select and acquire pairs of JSC DNN structures for bistatic radar system types in which the UE can configure a pair of JSC DNN structures. The BS JSC DNN structures and US JSC DNN structures are based on those described, for example, with reference to Figure 4.BS201 sends an RRC establishment request message that includes the DNN type or unique JSC DNN identifier (e.g., DNN ID1) associated with the selected BS JSC DNN / UE JSC DNN pair, a predetermined set of DL time slots (TS) (e.g., DL TSs{a,b,c,d}), a set of DL frequencies and / or resource blocks (RBs) for use in each of the predetermined set of DL time slots, a predetermined set of UL TS (e.g., UL TS{e,f,g,h}), and a set of UL frequencies and / or resource blocks (RBs) for use in each of the predetermined set of UL time slots (e.g., RRC JSC DNN Establishment Req(DNN ID1,DL TS{a,b,c,d},DL freq. / RBs,UL TS{e,f,g,h},UL freq. / RBs)).

[0099] Operation 703 is similar to Operation 403 in Figure 4, where BS DNNC 224 issues a configuration instruction (e.g., Cfg(DNN ID1)) associated with the selected BS JSC DNN structure, and BS 201 uses the neural network table provided to BS 201 (e.g., the master neural network table 1000 or the first neural network table 1010 as described with reference to Figures 10a-10b) to configure the BS JSC DNN structure 203 according to its unique JSC DNN identifier (e.g., DNN ID1). For example, BS 201 uses the JSC DNN identifier (e.g., DNN ID1) to retrieve the DNN parameters / weights and configuration parameters of the trained BS Tx CR-DNN and trained BS Rx R-DNN corresponding to the selected BS JSC DNN structure from the neural network table. BS201 configures the BS JSC DNN structure 203 with the DNN parameters / weights and / or configuration parameters of the selected BS JSC DNN structure, a trained BS Tx CR-DNN, and a trained BS Rx R-DNN. After the configuration of the BS JSC DNN structure 203, the BS JSC DNN structure 203 includes the trained BS Tx CR-DNN and the trained BS Rx R-DNN of the selected BS JSC DNN structure for bistatic JSC operation.

[0100] Operation 723 is similar to operation 423 in Figure 4, where UE DNNC234 receives an RRC establishment request message (e.g., RRC JSC DNN Establishment Req(DNN ID1,DL TS{a,b,c,d},DL freq. / RBs,UL TS{e,f,g,h},UL freq. / RBs)), and UE DNNC234 issues a configuration instruction (e.g., Cfg(DNN ID1)) associated with the selected UE JSC DNN structure. UE212 uses the neural network table provided to UE212 (e.g., a second neural network table 1020 described with reference to Figures 10a-10c) to configure the UE JSC DNN structure 214 according to its unique JSC DNN identifier (e.g., DNN ID1). For example, UE212 uses the JSC DNN identifier (e.g., DNN ID1) to configure the trained UE Rx corresponding to the UE JSC DNN structure selected from the neural network table. The DNN parameters / weights and configuration parameters of the CR-DNN and the trained UE Tx RFB-DNN are obtained. UE212 configures the UE JSC DNN structure 214 with the DNN parameters / weights and / or configuration parameters of the selected UE JSC DNN structure, the trained UE Rx CR-DNN and the trained UE Tx RFB-DNN. After the configuration of the UE JSC DNN structure 214, the UE JSC DNN structure 214 includes the trained UE Rx CR-DNN and the trained UE Tx RFB-DNN of the selected UE JSC DNN structure for bistatic JSC operation.

[0101] In operation 722, after configuring the UE JSC DNN structure 214, the UE DNNC 234 of UE212 sends an RRC response (e.g., RRC JSC DNN Establishment Resp (ACK)) to BS201 indicating that the UE JSC DNN structure 214 has been configured. Operations 702-703 and 722-723 complete the establishment of the DL bistatic JSC DNN communication session between BS201 and UE212.

[0102] Upon receiving an acknowledgment, operation 704a applies to the Tx CR-DNN model of the BS JSC DNN structure 203, which generates an output JSC signal (e.g., DNN_IDXa_Sens) representing the combination of IDXa and a radar signal (e.g., Sens) within a specific time slot, TS a (e.g., input communication data X(IDXa) of TS a). That is, the output JSC signal (e.g., DNN_IDXa_Sens) represents the digital integrated communication and radar signal with the input communication data (e.g., IDXa) of TS a incorporated therein and / or as described with reference to Figure 1. For example, the BS JSC DNN structure 203 generates an output JSC signal (e.g., DNN_IDXa_Sens) by processing the input communication data of TS a, and the output JSC signal of TS a (e.g., DNN_IDXa_Sens) represents the input communication data that has been digitally converted (or adapted) by the BS JSC DNN structure 203, for transmission within TS a to UE212, so that it also functions as a radar signal for radar sensing. In operation 706a, the BS JSC DNN structure 203 provides the output JSC signal (e.g., DNN_IDXa_Sens) to the BS DNNC224 for transmission as a JSC signal waveform via the PDSCH within TS a to UE210. Upon receiving DNN_IDXa_Sens, the BS DNNC224 buffers the DNN_IDXa_Sens data until transmission within TS a. In a specific TS a, BS201 processes the output JSC signal (e.g., DNN_IDXa_Sens) and transmits it via the PDSCH as a JSC signal waveform signal. In operation 708a, BS DNNC224 transmits the JSC signal waveform of TS a into TS a via the PDSCH (e.g., PDSCH RF Tx JSC WAVEFORM(DNN_IDXa_Sens_TS_a))). The JSC signal waveform of TS a incorporates the input communication data of TS a to operate as a radar signal for radar sensing and / or as described with reference to Figure 1, but represents a JSC waveform optimized for channel conditions / communication and / or radar performance requirements / JSC system type.

[0103] Operations 724a to 734a substantially correspond to operations 424 to 434 in FIG. 4. In operation 724a, UE 212 receives, via PDSCH, a transmitted JSC signal waveform within TS a, processes it (e.g., down-converts it), and makes it the first received JSC signal of TS a (e.g., RxIDXa_Sens_TS_a). The transmitted JSC signal waveform is also reflected from objects 109a to 109n (e.g., O1, O2, ..., O N ) to generate a plurality of reflected transmitted JSC signal waveforms for TS a (e.g., O1UE JSC-signal waveform_TS_a, O2UE JSC-signal waveform_TS_a, ..., O N UE JSC-signal waveform_TS_a). In each of operations 726a-1, 726a-2 to 726a-N, UE 212 receives the corresponding reflected transmitted JSC signal waveform for TS a, and UE DNNC 234 processes, for example, the reflected waveforms O1UE JSC-signal waveform_TS_a, O2UE JSC-signal waveform_TS_a, ..., O N UE JSC-signal waveform_TS_a received from one or more of the objects 109a to 109n for TS a, and makes them the corresponding plurality of received reflected JSC signals for TS a (e.g., O1UEJSC_TS_a, O2UEJSC_TS_a, ..., O N UEJSC_TS_a).

[0104] As described in operation 428 of FIG. 4, the first received JSC signal of TS a (e.g., RxIDXa_Sens_TS_a) and the plurality of received reflected JSC signals of TS a (e.g., O1UEJSC_TS_a, O2UEJSC_TS_a, ..., O NThe UEJSC_TS_a) is input to the UE Rx CR-DNN of the UE JSC DNN structure 214 to generate reconstructed communication data (e.g., RCIDXa) for TS a, and is processed by this UE Rx CR-DNN to generate radar sensing feedback (RFB) information (e.g., range, Doppler, velocity) for one or more objects. The RFB information is input to the UE Tx RFB-DNN of the UE JSC DNN structure 214 to generate the RFB signal (e.g., Tx RFB_TS_a) for TS a. The RFB signal (e.g., Tx RFB_TS_a) for TS a represents the range, Doppler, velocity, position, or other radar measurements of one or more objects 109a-109n, as illustrated with reference to Figure 4.

[0105] In operation 730a, the RFB signal (e.g., Tx RFB_TS_a) is transmitted to UE DNNC234 as the RFB signal waveform for TS a in one of the available UL time slots (e.g., TS e), and in operation 732a, the RFB signal for TS a (e.g., Tx RFB_TS_a) is transmitted to BS201 via PUSCH as the transmitted RFB signal waveform for TS a in TS e (e.g., PUSCH Tx JSC DATA(Tx RFB_TS_a, TS e)). Alternatively, the RFB signal for TS a can be transmitted via PUSCH. In operation 734a, UE212 transmits the reconstructed communication data for TS a (e.g., RCIDXa) to a data sink, for example, as one or more upper-layer protocols in the protocol stack of UE212 and / or as described herein.

[0106] In operation 710a, BS201 receives and processes the RFB signal waveform transmitted to TS a in TS e via PUSCH (e.g., down-converting to baseband) to obtain the received radar sensing signal for TS a (e.g., Rx RS_RFB_TS_a). BS DNNC224 buffers the received radar sensing signal for TS a received in TS e (e.g., Rx RS_RFB_TS_a) until the Rx R-DNN model of the BS JSC DNN structure 203 is ready to process the received radar sensing signal for TS a. Operations 712a-714a substantially correspond to operations 412-414 in Figure 4. In operation 712a, the BS DNNC224 inputs the received radar sensing signal Rx RS_RFB_TS_a to the Rx R-DNN model to generate radar sensing information (e.g., RSTS_a) for TS a about one or more objects 109a-109n. RSTS_a represents one or more of a group of values ​​such as range, Doppler, velocity, position, or other radar measurements for one or more objects 109a-109n in TS a. For example, the RSTS_a data represents one or more of the following: range estimates associated with each object 109a-109n, Doppler estimates associated with each object 109a-109n, velocity estimates associated with each object 109a-109n, location or position estimates associated with each object 109a-109n, delay spread associated with each object 109a-109n, mean delay associated with each object 109a-109n, angle estimates associated with each object 109a-109n, azimuth and / or elevation estimates associated with each object 109a-109n, and / or any other appropriate radar measurements associated with each object 109a-109n, radar sensing information or radar sensing parameter estimates for one or more objects 109a-109n in TS a.

[0107] For BS201 and UE212, operations 704a-714a and 724a-734a are repeated for each set of input communication data (e.g., IDXb for TS b, IDYc for TS c, IDZd for TS d, ..., etc.) for transmission in subsequent time slots in UE212, in order to generate reconstructed communication data (e.g., RCIDXb for TS b, RCIDYc for TS c, RCIDZd for TS d, ..., etc.) for each subsequent time slot (e.g., TS{b, c, d}) of objects 109a-109n, in order to generate radar sensing information (e.g., RSTS_b for TS b, RSTS_c for TS c, RSTS_d for TS d, ..., etc.). For example, operations 704d-714d and 724d-734d are performed in a similar manner to operations 704a-714a and 724a-734a, in which BS201 and UE212 generate radar sensing information (e.g., RSTS_d) for TS d relating to one or more objects 109a-109n, and reconstructed communication data (e.g., RCIDZd) for TS d corresponding to input communication data (e.g., IDZd) transmitted into TS d.

[0108] Regarding operations 418 and 438 in Figure 4, the bistatic JSC DNN operation is repeated until BS DNNC224 determines that the bistatic JSC DNN operation needs to be disabled. In operation 718, BS DNNC224 sends an RRC control message (e.g., RRC JSC DNN Disable Req(.)) to UE212 to disable bistatic JSC communication. BS DNNC224 disables bistatic JSC communication. In operation 738, upon receiving the RRC control message to disable bistatic JSC communication, UE212 also disables bistatic JSC communication and sends an acknowledgment (e.g., RRC JSC DNN Disable Resp(ACK)) accordingly.

[0109] Figures 7b and 7c show signal flow diagrams for exemplary multistatic JSC DNN operation 700b and exemplary monostatic JSC DNN operation 700c, respectively. The multistatic DNN operation 700b in Figure 7b is essentially the same as the bistatic DNN operation 700a in Figure 7a, except that BS201 also processes reflected signals further received from objects 109a-109n in order to further improve the accuracy of radar sensing information. The monostatic DNN operation 700c in Figure 7c differs from the respective bistatic DNN operation 700a and multistatic DNN operation 700c in Figures 7a and 7c, in that BS201 processes only reflected signals further received from objects 109a-109n in BS201, without UE212 transmitting RFB signal waveforms (e.g., radar sensing feedback signal waveforms) to BS201, as described in Figures 7a and 7b. The monostatic DNN operation 700c further simplifies the processing of radar sensing information in BS201. BS201 and UE212 use the monostatic DNN operation 700c when UE212 lacks the capability for more complex JSC processing or JSC DNN operation. Furthermore, the monostatic DNN operation 700c reduces the latency in generating radar sensing information in BS201 compared to the bistatic DNN operation 700a and multistatic DNN operation 700b, where BS202 waits for the RFB signal waveform corresponding to the TS from UE212 before generating radar sensing information for that TS.

[0110] Figure 7b shows a signal flow diagram of an exemplary multistatic JSC DNN operation 700b for one or more predetermined time slots (e.g., TS{a, b, c, d}) of a communication session between BS201 and UE212. The multistatic JSC DNN operation 700b modifies the bistatic JSC DNN operation 700a in Figure 7a by including processing of reflected signals further received from objects 109a-109n at BS201 to enhance radar sensing information.

[0111] Operations 702, 722, 703, and 723 in Figure 7b substantially correspond to operations 702, 712, 703, and 723 in Figure 7a, except that BS DNNC224 selects a pair of JSC DNN structures of the multistatic JSC system type, rather than the bistatic JSC system type as in Figure 7a. In this example, the BS JSC DNN structure includes a trained BS Tx CR-DNN model and a trained BS Rx R-DNN model of the multistatic JSC system type. The UE JSC DNN structure includes a trained UE Rx CR-DNN model and a trained UE Tx RFB-DNN model of the multistatic JSC system type. After configuring the BS JSC DNN structure 203 in a manner similar to that described with reference to Figure 7a, in operation 703, the BS JSC DNN structure 203 includes a trained BS Tx CR-DNN and a trained BS Rx R-DNN of the BS JSC DNN structure selected for multi-static JSC operation. After configuring the UE JSC DNN structure 214 in a manner similar to that described with reference to Figure 7a, in operation 723, the UE JSC DNN structure 214 includes a trained UE Rx CR-DNN and a trained UE Tx RFB-DNN of the UE JSC DNN structure selected for multi-static JSC operation. In operation 723, when configuring the UE JSC DNN structure 214, the UE DNNC 234 of UE 212 sends an RRC response (e.g., RRC JSC DNN Establishment Resp(ACK)) to BS201, instructing it to acknowledge that the UE JSC DNN structure 214 was configured in operation 722.

[0112] In operation 722, BS201 and UE212 execute multistatic DNN operation 700b after BS201 receives an acknowledgment (ACK) from UE212 (e.g., RRC JSC DNN Establishment Resp(ACK)) indicating that the UE JSC DNN structure 214 has been successfully configured. Except that BS JSC DNN and UE JSC DNN are of the multistatic JSC type, operations 700b are performed, with operations 704a, 706a, 708a, 724a, 726a-1, 726a-2, 726a-N, 730a, 732a / 710a, 712a, and 734a substantially corresponding to the bistatic DNN operations 704a, 706a, 708a, 724a, 726a-1, 726a-2, 726a-N, 730a, 732a / 710a, 712a, and 734a in Figure 7a. In addition, when transmitting the JSC signal waveform of TS a via PDSCH in operation 708a, the transmitted JSC signal waveform is also objects 109a~109n (e.g., O1, O2, ..., O N ) is also reflected from multiple BS received reflected transmitted JSC signal waveforms for TS a (e.g., O1BS JSC-signal waveform_TS_a, O2BS JSC-signal waveform_TS_a, ..., O N BS generates a JSC-signal waveform (TS_a). For each operation 710a-1, 710a-2, ~710a-N, BS201 generates the corresponding BS reflected transmit JSC signal waveform for TS a (e.g., O1BS JSC-signal waveform_TS_a, O2BS JSC-signal waveform_TS_a, ..., O N BS JSC-signal waveform_TS_a) is received. BS DNNC224 receives, for example, O1BS JSC-signal waveform_TS_a, O2BS JSC-signal waveform_TS_a, ..., O NProcess the BS received reflected waveform from objects 109a~109n for TS a, including BS JSC-signal waveform_TS_a, and the corresponding multiple radar sensing signals for TS a (e.g., O1BSJSC_RS_TS_a, O2BSJSC_RS_TS_a, ..., O N Set to BSJSC_RS_TS_a). The BS DNNC224 receives the received radar sensing signals of TS a (e.g., O1BSJSC_RS_TS_a, O2BSJSC_RS_TS_a, ..., O) until the BS JSC DNN has a multistatic input for generating radar sensing information for TS a. N BSJSC_RS_TS_a) is buffered. As described in operation 528 of Figure 5, and also in operations 724a, 726a-1, 726a-2, 726a-N, 730a, 732a / 710a of Figure 7a, UE212 generates the RFB signal for TS a (e.g., Tx RFB_TS_a) which is sent to BS201 via PUSCH in TS e as an RFB signal waveform (e.g., PUSCH Tx JSC DATA(Tx RFB_TS_a, TS e)) in operation 732a. Alternatively, the RFB signal for TS a can also be sent via PUSCH. Similarly, in operation 710a, BS201 receives the transmitted RFB signal waveform of TS a in TS e via PUSCH from UE212, processes the received RFB signal waveform of TS a (e.g., downconverts to baseband), and converts it into a further received radar sensing signal of TS a (e.g., Rx RS_RFB_TS_a).

[0113] In operation 712a of Figure 7b, multiple received radar sensing signals of TS a (e.g., O1BSJSC_RS_TS_a, O2BSJSC_RS_TS_a, ..., O NExcept that UE212 inputs BSJSC_RS_TS_a) and further received radar sensing signals of TS a (e.g., Rx RS_RFB_TS_a) to the Rx R-DNN model of the BS JSC DNN structure 203 to generate radar sensing information of TS a (e.g., RSTS_a) for one or more objects 109a-109n, operations 712a-714a substantially correspond to operations 512-514 in Figure 5 and operations 712a and 714a described with respect to Figure 7a.

[0114] The RSTS_a data for TS a represents one or more of a group of radar measurements, such as range, Doppler, velocity, position, or other radar measurements, for one or more objects 109a-109n of TS a. For example, RSTS_a includes data representing one or more from a group such as range estimates associated with each object 109a-109n, Doppler estimates associated with each object 109a-109n, velocity estimates associated with each object 109a-109n, location or position estimates associated with each object 109a-109n, delay spread associated with each object 109a-109n, mean delay associated with each object 109a-109n, angle estimates associated with each object 109a-109n, azimuth and / or elevation estimates associated with each object 109a-109n, and / or any other appropriate radar measurements associated with each object 109a-109n, radar sensing information or radar sensing parameter estimates for one or more objects 109a-109n of TS a.

[0115] For BS201 and UE212, operations 704a-714a and 724a-734a each repeatedly generate reconstructed communication data (e.g., RCIDXb for TS b, RCIDYc for TS c, IDZd for TS d, etc.) for each subsequent time slot in UE212 for transmission within the subsequent time slots TS b, TS c, and TS d, and generate radar sensing information (e.g., RSTS_b for TS b, RSTS_c for TS c, RSTS_d for TS d, etc.) for each subsequent time slot in objects 109a-109n. Regarding operations 518 and 538 in Figure 5, the multistatic JSC DNN operation is repeated until BS DNNC224 determines that the multistatic JSC DNN operation needs to be disabled, and in Figure 7b, operations 718 and 738 corresponding to operations 718 and 738 in Figure 7a are executed.

[0116] Figure 7c shows a signal flow diagram of an exemplary monostatic JSC DNN operation 700c for one or more predetermined time slots (e.g., TS{a, b, c, d}) of a communication session between BS201 and UE212. The monostatic JSC DNN operation 700c modifies the multistatic JSC DNN operation 700b in Figure 7b only by removing UE processing of further received reflected signals from objects 109a-109n in UE212 so that RFB signal waveforms are not generated, while BS201 retains processing of BS received reflected signals from objects 109a-109n. This has the advantage of simplifying the processing of radar sensing information when UE212 does not have the capability for more complex JSC processing. Other advantages include throughput in generating radar sensing information per time slot compared to bistatic or multistatic JSC system types, because BS201 does not need to wait for the reception of the corresponding RFB signal waveform from UE212 for each time slot.

[0117] Operations 702, 722, 703, and 723 in Figure 7b substantially correspond to the operations described above with respect to Figure 7a or Figure 7b, except that BS DNNC224 selects a pair of JSC DNN structures for the monostatic JSC system type. In this example, the BS JSC DNN structure includes a trained BS Tx CR-DNN model and a trained BS Rx R-DNN model for the monostatic JSC system type. The UE JSC DNN structure includes only a trained UE Rx C-DNN model. After configuring the BS JSC DNN structure 203, in operation 703, the BS JSC DNN structure 203 includes the trained BS Tx CR-DNN and trained BS Rx R-DNN of the BS JSC DNN structure selected for monostatic JSC DNN operation 700c. After the UE JSC DNN structure 214 is configured, in operation 723, the UE JSC DNN structure 214 includes a trained UE Rx C-DNN for monostatic JSC DNN operation. In operation 723, after the UE JSC DNN structure 214 is configured, the UE DNNC 234 of UE212 sends an RRC response (e.g., RRC JSC DNN Establishment Resp(ACK)) to BS201 in operation 722, indicating that the UE JSC DNN structure 214 has been configured.

[0118] In operation 722, BS201 and UE212 execute monostatic JSC DNN operation 700c after BS201 receives an acknowledgment from UE212 (e.g., RRC JSC DNN Establishment Resp(ACK)) indicating that the UE JSC DNN structure 214 has been successfully configured. Monostatic JSC DNN operation 700c is executed, except that operations 704a, 706a, 708a, and 710a-1, 710a-2, 710a-N substantially correspond to the multistatic DNN operations 704a, 706a, 708a, and 710a-1, 710a-2, 710a-N in Figure 7b, except that the BS JSC DNN and UE JSC DNN are configured for the monostatic JSC system type.

[0119] In operation 708a, BS DNNC224 transmits the DNN_IDXa_Sens data to UE212 as the transmitted JSC signal waveform via PDSCH in TS a (e.g., PDSCH RF Tx JSC WAVEFORM(DNN_IDXa_Sens, TS a))). UE212 receives the transmitted JSC signal waveform via PDSCH in TS a, processes it (e.g., downconverts it), and makes it the first received JSC signal of TS a (e.g., RxIDXa_Sens_TS_a). In operation 724a, UE DNNC234 inputs the received JSC signal of TS a (e.g., RxIDXa_Sens_TS_a) to the Rx C-DNN model of UE JSC DNN structure 214 to generate the reconstructed communication data of TS a (e.g., RCIDXa). The UE DNNC234 of the UE212 transmits the reconstructed communication data (e.g., RCIDXa) of TS a, corresponding to the input communication data IDXa transmitted into TS a, to the data sink or to a higher-layer protocol of the UE212's protocol stack, and / or as described herein.

[0120] In operation 708a, when transmitting the JSC signal waveform of TS a via PDSCH, the transmitted JSC signal waveform is also an object 109a~109n (e.g., O1, O2, ..., O N ) is also reflected from multiple BS received reflected transmitted JSC signal waveforms for TS a (e.g., O1BS JSC-signal waveform_TS_a, O2BS JSC-signal waveform_TS_a, ..., O N BS201 generates a BS JSC-signal waveform (TS_a). For operations 710a-1, 710a-2, ~710a-N in Figure 7b, BS201 processes the corresponding BS received reflected transmitted JSC signal waveform of TS a and generates one or more received radar sensing signals of TS a (e.g., O1BSJSC_RS_TS_a, O2BSJSC_RS_TS_a, ..., O NSet to BSJSC_RS_TS_a). In response to operations 710a-1, 710a-2, ~710a-N, the BS DNNC224 inputs the received radar sensing signals of TS a into the Rx R-DNN model of the BS JSC DNN structure 203 so as to generate radar sensing information (e.g., RSTS_a) of TS a for one or more objects 109a~109n.

[0121] The radar sensing information of TS a for one or more objects 109a to 109n (e.g., RSTS_a) represents one or more of the group of values ​​for one or more objects 109a to 109n of TS a, such as range estimates for each related object, Doppler estimates for each related object, velocity estimates for each related object, position or location estimates for each related object, delay spread, Doppler spread, mean delay, angle estimates for each related object, azimuth and / or elevation estimates for each related object, or other radar measurements. For example, RSTS_a contains data representing one or more of the group of values ​​for one or more objects 109a to 109n of TS a, such as range estimates, Doppler estimates, velocity estimates, position or location estimates, delay spread, mean delay, angle estimates, azimuth and / or elevation estimates, and / or any other appropriate radar measurements, radar sensing information, or radar sensing parameter estimates.

[0122] For BS201 and UE212, operations 704a-714a and 724a-744a are repeated for each set of input communication data (e.g., IDXb for TS b, IDYc for TS c, IDZd for TS d, ..., etc.) for transmission in subsequent time slots, respectively, in UE212 to generate reconstructed communication data (e.g., RCIDXb for TS b, RCIDYc for TS c, RCIDZd for TS d, ..., etc.) for each subsequent time slot, and for radar sensing information (e.g., RSTS_b for TS b, RSTS_c for TS c, RSTS_d for TS d, ..., etc.) for each subsequent time slot of objects 109a-109n. The monostatic JSC DNN operation is repeated until BS DNNC224 determines that the monostatic JSC DNN operation needs to be disabled, and in Figure 7c, operations 718 and 748 are executed, corresponding to operations 718 and 738 in Figure 7a.

[0123] Figures 7a and 7c illustrate the downlink JSC DNN operation between BS201 and UE212, but this is merely illustrative and not limiting. Those skilled in the art will understand that the first device 201 and the second device 212 can be the UE and BS, respectively, and that the first device 201 (i.e., the UE) and the second device 212 (i.e., the BS) perform uplink JSC DNN operation via PUSCH.

[0124] Figure 8a shows a flowchart of an exemplary JSC DNN process 800 of a first device that performs JSC DNN operations within one or more time slots of a communication session with a second device. The JSC DNN process 800 includes the following steps:

[0125] Step 802 establishes JSC DNN operation for one or more time slots of a communication session with the second device. In an example, step 802 includes establishing JSC DNN communication and configuring the JSC DNN structure of the first device with respect to a bistatic JSC system type, as illustrated with reference to operations 402 and 403 in Figure 4 and operations 702, 703, 722, and 723 in Figure 7a. In another example, step 802 includes establishing JSC DNN communication and configuring the JSC DNN structure of the first device with respect to a multistatic JSC system type, as illustrated with reference to operations 502 and 503 in Figure 5 and operations 702, 703, 722, and 723 in Figure 7b. In a further example, step 802 includes establishing JSC DNN communication with respect to a monostatic JSC system type and configuring the JSC DNN structure of the first device, as described with reference to operations 602 and 603 in Figure 6 and operations 702, 703, 722, and 723 in Figure 7c.

[0126] For each time slot of one or more time slots, the JSC DNN process 800 executes steps 804-816 based on the following:

[0127] Step 804 involves acquiring input communication data for transmission to the second device for each time slot. For example, step 804 includes acquiring input communication data as described with reference to operations 404, 504, 604, 704a-704d, with reference to Figures 4-7c.

[0128] In step 806, the input communication data is processed by a JSC DNN structure of the first device to generate an output JSC signal, the output JSC signal representing the input communication data, the radar signal for transmission per time slot, and / or those described with reference to Figure 1. For example, the JSC DNN structure includes a transmitting-side communication and radar DNN (Tx CR-DNN) model to output a JSC signal for transmission per time slot, given input communication data as input. In an example, step 806 includes processing the input communication data using a Tx CR-DNN to generate an output JSC signal for a bistatic JSC system type, as described with reference to operation 406 in Figure 4 and operations 706a-706d in Figure 7a. In another example, step 806 includes processing the input communication data using a Tx CR-DNN to generate an output JSC signal for a multistatic JSC system type, as described with reference to operation 506 in Figure 5 and operations 706a-706d in Figure 7b. In a further example, step 806 includes processing input communication data using a Tx CR-DNN to generate an output JSC signal, with respect to a monostatic JSC system type, as illustrated with reference to operation 606 in Figure 4 and operations 706a-706d in Figure 7c.

[0129] Step 808 transmits the output JSC signal as a JSC signal waveform for each time slot to a second device via a communication channel. For example, step 808 includes transmitting the output JSC signal as a JSC signal waveform to a second device, as described with reference to operations 408, 508, 608, and 708a-708d with reference to Figures 4-7c.

[0130] In step 810, one or more radar sensing signals are received for each time slot based on reflections of JSC signal waveforms transmitted from one or more objects. For example, in the case of a bistatic JSC system type, step 810 includes receiving RFB signal waveforms transmitted from a second device, as described with reference to operations 410 and 732a / 710a-732d / 710d, respectively, with reference to either Figure 4 or Figure 7a. For example, one or more radar sensing signals are generated from the first device receiving RFB signal waveforms generated by the second device using reflections of JSC signal waveforms transmitted from one or more objects in the time slot. The first device performs RF-to-baseband processing of the received RFB signal waveforms for the time slot to generate the radar sensing signals for the time slot. In other examples, for a multistatic JSC system type, step 810 includes receiving an RFB signal waveform transmitted from a second device, as described with reference to operation 510 and operation 732a / 710a~732d / 710d and operation 710a-1~710a-N~710d-1~710d-N, respectively, with reference to Figures 5 and 7b, and also receiving a reflection of a transmitted JSC signal waveform from one or more objects. In other examples, for a monostatic JSC system type, step 810 includes receiving a reflection of a transmitted JSC signal waveform from one or more objects, as described with reference to operation 610 and operation 732a / 710a~732d / 710d and operation 710a-1~710a-N~710d-1~710d-N, respectively, with reference to Figures 6 and 7c. For example, one or more radar sensing signals are generated when a first device receives one or more reflected JSC signal waveforms in time slots, resulting from the reflection of JSC signal waveforms transmitted from one or more objects.The first device performs RF-to-baseband processing of one or more received reflected JSC signal waveforms associated with one or more objects in a time slot, resulting in one or more radar sensing signals for the time slots corresponding to one or more objects.

[0131] In step 812, the received radar sensing signal is processed by a JSC DNN structure of the first device (e.g., BS JSC DNN structure 203 in Figure 2) to generate radar sensing information for one or more objects. For example, the JSC DNN structure includes a receiving radar DNN (Rx R-DNN) model for generating radar sensing information when given a received radar sensing signal as input. For example, in the case of a bistatic JSC system type, in step 812, the radar sensing signal representing the received RFB signal waveform is input to the Rx R-DNN model to generate radar sensing information, referring to operations 412 and 712a-712d, respectively, referring to either Figure 4 or Figure 7a. In other examples, for a multistatic JSC system type, step 812 inputs radar sensing signals representing the received RFB signal waveform and the reflection of the transmitted JSC signal waveform from one or more objects into the Rx R-DNN model to generate radar sensing information, as described with reference to operations 512 and operations 712a-712d, respectively, with reference to Figures 5 and 7b. In other examples, for a monostatic JSC system type, step 812 inputs radar sensing signals representing the reflection of the transmitted JSC signal waveform from one or more objects into the Rx R-DNN model to generate radar sensing information, as described with reference to operations 612 and operations 712a-712d, respectively, with reference to Figures 6 and 7c.

[0132] In step 814, radar sensing information for one or more objects per time slot is transmitted to one or more higher protocol layers of the first device's protocol stack. In the example, the first device transmits radar sensing information for one or more objects per time slot to one or more higher layer protocols of the first device's protocol stack. For example, the radar sensing information is transmitted up to the application protocol layer of the protocol stack for use by one or more applications running on the first device.

[0133] Step 816 determines whether to continue the JSC DNN operation. If there are further time slots within one or more time slots (e.g., "Y"), proceed to step 804 to obtain the input communication data for the next time slot among the one or more time slots. If there are no further time slots for performing the JSC DNN operation (e.g., "N"), proceed to step 818.

[0134] Step 818 disables the JSC DNN operation for the communication session with the second device. See, for example, operations 418 and 438 in Figure 4, operations 518 and 538 in Figure 5, operations 618 and 638 in Figure 6, and operations 718 and 738 in any of Figures 7a to 7c.

[0135] Figure 8b shows a flowchart illustrating the JSC DNN establishment process of step 802 in Figure 8a, which is performed by the first device after establishing JSC DNN operation for one or more time slots of a communication session with the second device. The JSC DNN establishment process of step 802 includes the following steps:

[0136] In step 802a, one or more control messages are sent to the second device that indicate the JSC DNN structure selected for use per time slot. For example, one or more control messages may be RRC establishment request messages or equivalents thereof, as described in operation 702 with reference to any of Figures 7a to 7c.

[0137] In step 802b, the JSC DNN structure of the first device is configured with a selected JSC DNN structure for use in processing input communication data and receiving one or more radar sensing signals. For example, the first device is configured with the JSC DNN structure of the first device as described in operation 703 with reference to any of Figures 7a to 7c.

[0138] Step 802 of the JSC DNN establishment process proceeds to step 804 of the JSC DNN process 800 in Figure 8.

[0139] Figure 9a is a flowchart showing an exemplary JSC DNN process 920 of a second device performing bistatic or multistatic JSC operation within one or more time slots of a communication session with a first device. The bistatic or multistatic JSC DNN operation is based on the bistatic or multistatic JSC DNN operation as described with reference to Figures 3a, 3b, 4, 5, 7a, and 7b. The JSC DNN process 920 of the second device includes the following steps:

[0140] Step 922 establishes JSC DNN operation for one or more time slots of a communication session with the first device. In an example, step 922 includes establishing JSC DNN communication with the first device and configuring the JSC DNN structure of the second device with respect to a bistatic JSC system type, as described with reference to operations 422 and 423 in Figure 4 and operations 722 and 723 in Figure 7a. In another example, step 922 includes establishing JSC DNN communication with the first device and configuring the JSC DNN structure of the second device with respect to a multistatic JSC system type, as described with reference to operations 522 and 523 in Figure 5 and operations 722 and 723 in Figure 7b.

[0141] For each time slot of one or more time slots, the JSC DNN process 920 executes steps 924-936 based on the following:

[0142] In step 924, the JSC signal waveform is received from the first device via the communication channel for each time slot, and the JSC signal waveform represents the input communication data and radar signal. For example, step 924 includes receiving the JSC signal waveform in the second device, as described with reference to operations 424, 524, 724a-724d with reference to Figures 4, 5, 7a and 7b.

[0143] In step 926, one or more additional JSC signal waveforms are received per time slot based on reflections of transmissions of JSC signal waveforms from one or more objects in the communication channel. For example, in the case of a bistatic or multistatic JSC system type, in step 926, one or more additional JSC signal waveforms include reflections of JSC signal waveforms from one or more objects, as described with reference to operations 426, 526, and 726a-1 to 726a-N and 726d-1 to 726d-N, respectively, with reference to either Figure 4, Figure 5, and Figures 7a and 7b.

[0144] In step 928, the received JSC signal waveform and, if present, further JSC signal waveforms (e.g., reflections) are processed by a JSC DNN structure in a second device to generate reconstructed communication data for each time slot corresponding to the input communication data transmitted for each time slot, and to generate radar sensing feedback (RFB) information associated with one or more objects for each time slot. For example, the JSC DNN structure in the second device includes a receiver-side communication and radar DNN (Rx CR-DNN) model to generate reconstructed communication data and RFB information (e.g., range, Doppler, velocity, etc.) for each time slot. For example, in the case of a bistatic or multistatic JSC system type, in step 928, the received JSC signal waveform and, if present, further JSC signal waveforms are input to the Rx CR-DNN model to generate reconstructed communication data and RFB information, as described in either operation 428 or 528, respectively, with reference to Figure 4 or Figure 5.

[0145] In step 930, the JSC DNN structure further includes a Transmitter (Tx) Radar Feedback (RFB) DNN, which processes RFB information for each time slot using a Tx RFB DNN, and the RFB DNN generates an RFB signal for transmission to a first device as an RFB signal waveform. For example, in the case of a bistatic or multistatic JSC system type, in step 930, the RFB information is input to the Tx RFB DNN model to generate an RFB signal, as described in operations 430, 530, or 730a-730d, respectively, with reference to Figures 4, 5, or 7a and 7b.

[0146] In step 932, the RFB signal is transmitted to the first device as an RFB signal waveform for use by the first device in generating radar sensing information for one or more objects. For example, in the case of a bistatic or multistatic JSC system type, in step 932, the second device transmits the RFB signal as described in operations 432, 532, or 732a-732d, respectively, with reference to Figures 4, 5, or 7a and 7b. For example, the RFB signal waveform for each time slot is received as a received radar sensing signal in step 810 of the JSC DNN process 800 and processed in step 812 to assist in generating radar sensing information corresponding to one or more objects. In some examples, the generated RFB signal data is transmitted to the first device in one or more further time slots via a communication data channel or a communication control channel, as is conventional.

[0147] In step 934, the JSC DNN structure of the second device (for example, the JSC DNN structure 214 in Figure 2) sends the reconstructed communication data for each time slot to the data sink or to one or more higher protocol layers of the second device's protocol stack. The reconstructed communication data for each time slot corresponds to the input communication data sent for each corresponding time slot. In the example, the second device sends the reconstructed communication data to one or more higher protocol layers of the second device's protocol stack. For example, it sends it to the application protocol layer of the protocol stack for use by one or more applications running on the second device.

[0148] Step 936 determines whether to continue the JSC DNN operation. If the second device needs to receive further JSC signal waveforms in corresponding further time slots of one or more time slots (e.g., "Y"), then proceed to step 924 to receive further JSC signal waveforms transmitted by the first device in further time slots of one or more time slots. If there are no further time slots for performing the JSC DNN operation (e.g., "N"), proceed to step 938. Additionally or alternatively, determining whether to continue the JSC DNN operation includes, for example, receiving a control message from the first device instructing the JSC DNN operation to be disabled, and if such a control message is received (e.g., "N"), proceed to step 938; otherwise, proceed to step 924.

[0149] Step 938 disables the JSC DNN operation for the communication session with the first device. See, for example, operations 418 and 438 in Figure 4, operations 518 and 538 in Figure 5, and operations 718 and 738 in either Figure 7a or Figure 7b.

[0150] Figure 9b is a flowchart showing another exemplary JSC DNN process 940 of a second device performing monostatic JSC DNN operation within one or more time slots of a communication session with the first device. The monostatic JSC DNN operation is based on the monostatic JSC DNN operation as described with reference to Figures 3c, 6, and 7c. The exemplary JSC DNN process 940 includes the following steps:

[0151] Step 922 establishes JSC DNN operation for one or more time slots of a communication session with the first device. In the example, step 922 includes establishing JSC DNN communication with the first device and configuring the JSC DNN structure of the second device with respect to a monostatic JSC system type, as described with reference to operations 622 and 623 in Figure 6 and operations 722 and 723 in Figure 7c.

[0152] For each time slot of one or more time slots, the JSC DNN process 940 executes steps 924 / 926, 942-948 based on the following:

[0153] In steps 924 / 926, the JSC signal waveform is received from the first device via the communication channel for each time slot, the JSC signal waveform representing input communication data and radar signals, and is transmitted for each time slot. In some examples, steps 924 / 926 include receiving one or more additional JSC signal waveforms for each time slot based on reflections of transmissions of JSC signal waveforms from one or more objects. In other examples, steps 924 / 926 include receiving the JSC signal waveform in the second device as described with reference to operations 624 and any of operations 724a-724d with reference to Figures 6 and 7c.

[0154] In step 942, the second device processes the received JSC signal waveform and / or any further JSC signal waveforms associated with reflections from one or more objects using a communication DNN structure. The communication DNN structure is configured to generate reconstructed communication data for each time slot corresponding to the input communication data transmitted from the first device for each corresponding time slot. For example, in the case of a monostatic JSC system type, in step 942, the received JSC signal waveform and / or any further JSC signal waveforms associated with reflections from one or more objects are input to an Rx C-DNN model for processing and generating reconstructed communication data, as described in operation 642 with reference to Figure 6.

[0155] In step 944, the reconstructed communication data for each time slot is sent to the data sink. In the example, the second device sends the reconstructed communication data to one or more higher protocol layers of the second device's protocol stack. For example, it sends it to the application protocol layer of the protocol stack for use by one or more applications running on the second device.

[0156] In step 946, the second device decides whether to continue the JSC DNN operation. If there are further time slots allocated to perform the JSC DNN operation (e.g., "Y"), proceed to step 924 / 926 to receive further JSC signal waveforms within those time slots. If there are no further time slots available to perform the JSC DNN operation (e.g., "N"), proceed to step 948. In other examples, if the second device receives a control message from the first device instructing it to disable the JSC DNN operation (e.g., "N"), proceed to step 948; otherwise, proceed to step 924 / 926.

[0157] Step 948 disables the JSC DNN operation for the communication session with the first device. See, for example, operations 618 and 638 in Figure 6 and operations 718 and 738 in Figure 7c.

[0158] Figure 9c is a flowchart illustrating an exemplary JSC DNN establishment process of step 922 in Figure 9a or Figure 9b, which is performed by the second device after establishing JSC DNN operation for one or more time slots of a communication session with the first device. For example, step 922 of JSC DNN process 920 or 940 performs the JSC DNN establishment process of step 922. The JSC DNN establishment process of step 922 includes the following steps:

[0159] In step 922a, the first device receives one or more control messages from the first device indicating the type of JSC DNN structure selected for use by the second device when processing the received JSC signal waveform transmitted from the first device. For example, one or more control messages may be RRC establishment request messages or equivalents thereof, as described in operation 702 with reference to any of Figures 7a to 7c.

[0160] In step 923, a JSC DNN structure of the second device is configured based on the selected type of JSC DNN structure for use within one or more time slots. The JSC DNN structure is configured to generate communication data and / or radar feedback data from the received JSC signal waveform transmitted by the first device. For example, the second device configures the JSC DNN structure as described in operations 403, 503, or 703 with reference to any of Figures 4-6 and 7a-7c.

[0161] The JSC DNN establishment process in Step 922 proceeds to Step 924 of the JSC DNN process 920, or to Steps 924 / 926 of the JSC DNN process 940.

[0162] Figure 10a shows an exemplary master neural network table 1000, and Figures 10b, 10c, and 10d show exemplary first neural network tables 1010, second neural network table 1020, and third neural network table 1030 provided to and used by a first device, a second device, and an optional third device, respectively. The first, second, and third neural network tables 1010, 1020, and 1030 are subtables of the master neural network table 1000. For example, as illustrated with reference to Figures 1 to 9c, the first, second, and third devices use the neural network tables 1010, 1020, and 1030, respectively, to select the corresponding JSC DNN structure for JSC DNN operation within one or more time slots of a communication session between the first, second, and third devices.

[0163] With respect to Figure 10a, the master neural network table 1000 includes a JSC DNN identifier column, a JSC DNN structure column for the first device, JSC DNN structure columns for the second / third devices, a communication channel condition / characteristic column, and a JSC system type column (e.g., bistatic JSC, multistatic JSC, monostatic JSC, etc.). The JSC DNN identifier column holds a unique JSC DNN identifier for each row (e.g., JSC DNN ID=1, 2, ..., 10, ..., 15a, 15b, ..., etc.). The JSC DNN structure column for the first device holds the configuration data or weights of the trained DNN model for the JSC DNN structure of the first device for each row. The JSC DNN structure column for the second / third devices holds the configuration data or weights of the trained DNN model for the JSC DNN structure of the second / third devices for each row. The DNN models held row by row in the JSC DNN structure column of the first device and the JSC DNN structure column of the second / third device form pairs of JSC DNN structures that are co-trained based on communication channel conditions / characteristics and radar configuration, as indicated in the same row of the columns for communication channel conditions / characteristics and JSC system type. The communication channel conditions / characteristics column holds row by row an indicator of the communication channel conditions / characteristics (e.g., LOS, NLOS, WEATHER CONDITION1 / NLOS, etc.) that the trained DNN models in the columns for the JSC DNN structure of the first device and the JSC DNN structure of the second / third device have co-trained. The JSC system type column holds row by row the type of JSC system that the corresponding trained DNN models in the columns for the JSC DNN structure of the first device and the JSC DNN structure of the second / third device have co-trained for, for example, bistatic JSC, multistatic JSC, monostatic JSC, and / or any combination thereof, modified forms, and / or variant forms.

[0164] Each row in the master neural network table 1000 is populated with a JSC DNN identifier unique to a pair of JSC DNN structures for a first device and a second / third device, and the pair of JSC DNN structures for the first device and the second / third device is trained for specific communication channel conditions / characteristics and specific radar or JSC system types (e.g., bistatic JSC, multistatic JSC, or monostatic JSC). For example, the first row of the master neural network table 1000 contains a JSC DNN ID value of 1, in which case the JSC DNN structure of the first device contains a JSC DNN structure of the first device that includes a Tx CR-DNN / Rx R-DNN model, the JSC DNN structure of the second / third device contains an Rx CR-DNN / Tx RFB-DNN model, the Tx CR-DNN / Rx R-DNN model and the Rx CR-DNN / Tx RFB-DNN model are jointly trained for line-of-sight (LOS) communication conditions, and the Tx CR-DNN / Rx R-DNN model and the Rx CR-DNN / Tx RFB-DNN model have inputs and outputs corresponding to bistatic JSC system types.

[0165] With respect to Figure 10b, the first neural network table 1010 includes a JSC DNN identifier column, a JSC DNN structure column for the first device, a communication channel condition / characteristics column, and a JSC system type column (e.g., bistatic JSC, multistatic JSC, monostatic JSC, etc.). The JSC DNN identifier column holds a unique JSC DNN identifier for each row (e.g., JSC DNN ID=1, 2, ..., 10, ..., 15a, 15b, ..., etc.). The JSC DNN structure column of the first device holds, row by row (e.g., row by The communication channel conditions / characteristics are stored row by row, with trained DNN models co-trained in the JSC DNN structure of the first device in the first neural network table 1010 and in the columns of the JSC DNN structure of the second / third device in the second / third neural network tables 1020 / 1030 for the same row, each holding an indicator of the communication channel conditions / characteristics (e.g., LOS, NLOS, WEATHER CONDITION1 / NLOS, etc.).The JSC system type column holds, row by row, the type of JSC system co-trained by the corresponding trained DNN models in the columns for the JSC DNN structure of the first device in the first neural network table 1010 and the JSC DNN structure of the second / third device in the second / third neural network tables 1020 / 1030 in the same row, for example, bistatic JSC, multistatic JSC, monostatic JSC, and / or any combination thereof, modified forms, and / or variant forms.

[0166] Optionally, the first neural network table 1010 may include JSC DNN structure columns for the second / third device of the master neural network table 1000, which allows the first device to determine a suitable pair of JSC DNN structures based on the capabilities of both the first and second devices when it accesses the first neural network table 1010. For example, the first device can determine the complexity of the second device's JSC DNN structures and whether the second device's capability means that it can configure and operate the second device's JSC DNN structures of the pair. If not, the first device can identify other pairs of JSC DNN structures that the second device can configure and operate. The first device then uses a control message to inform the second device of the JSC DNN identifier of the selected pair of JSC DNN structures.

[0167] Optionally, in some examples, the master neural network table 1000 is provided to and / or accessible by the first or second device for use in selecting a pair of JSC DNN structures based on communication channel conditions and JSC system type. This provides the advantage that the first device selects an appropriate pair of JSC DNN structures, and the second device can execute and / or operate the JSC DNN structures of the second device of the selected pair of JSC DNN structures. For example, the first device receives capability information or processing capability information associated with the second device and uses this information to select a pair of JSC DNN structures that is more likely to enable the second device to generate reconstructed data and / or RFB signal waveforms using the JSC DNN structures of the second device of the selected pair of JSC DNN structures. Alternatively, the first device is provided with a first neural network table 1010, which is a subset of the master neural network table 1000, as shown in Figure 10b. Alternatively, the second device may be provided with a neural network table 1020, which is a subset of the master neural network table 1000, as shown in Figure 10c.

[0168] With respect to Figure 10b, the first neural network table 1010, provided to and used by the first device, is a subset of the master neural network table 1000 and includes only the JSC DNN ID column, the corresponding JSC DNN structure column for the first device, and the communication channel conditions / characteristics and JSC system type columns.

[0169] With respect to Figure 10c, the second neural network table 1020, provided to and used by the second device, is a subset of the master neural network table 1000 and includes only the JSC DNN ID column from the JSC DNN structure column of the second / third device and the corresponding JSC DNN structure of the second device.

[0170] With respect to Figure 10d, the third neural network table 1030, which is provided to and used by the third device, is also a subset of the master neural network table 1000 and includes only the JSC DNN ID column from the JSC DNN structure column of the second / third device and the corresponding JSC DNN structure of the third device.

[0171] With respect to Figures 10c and 10d, in this example, the second neural network table 1020 of the second device includes JSC DNN IDs 1, 2, 10, and 15a, excluding those JSC DNN IDs assigned to the third device, such as JSC DNN ID 15b, which is assigned to the JSC DNN structure of the third device as seen in the third neural network table 1030. Similarly, the third neural network table 1030 of the third device includes JSC DNN IDs, excluding those assigned to the second device.

[0172] In some examples, the second neural network table 1020 includes all JSC DNN IDs assigned to the second device, as well as the corresponding columns of the master neural network table 1000. This allows the second device to identify the JSC DNN ID and select the first JSC DNN structure with specific channel conditions / characteristics and radar configuration in order to establish JSC DNN operation during a communication session with the first device. Since the second JSC DNN structure is assigned to the first device, the first device supports the radar sensing that the second device is performing. The first device, the second device, and / or the third device use the first, second, and third neural network tables 1010, 1020, and / or 1030 to select and configure their corresponding JSC DNN structures with respect to the JSC system type and / or radar / channel conditions or performance requirements (e.g., JSC performance requirements) as described with reference to any of Figures 1 to 9b.

[0173] During operation, the first device selects a pair of JSC DNN structures from the first neural network table 1010, which has multiple first JSC DNN structures in the JSC DNN structure column of the first device, for selection and use by the first device. As seen in the master neural network table 1000, each of the multiple first JSC DNN structures corresponds to a pair of JSC DNN structures, which includes a first JSC DNN structure for use by the first device and a second JSC DNN structure for use by the second device. Each pair of JSC DNN structures is mapped to a JSC DNN identifier. For each JSC DNN structure of the first device, the first neural network table 1010 includes the communication channel conditions on which the JSC DNN structure is trained and the JSC system type to which the JSC DNN structure is associated. After selecting a JSC DNN structure based on channel conditions and JSC system type, the first device, while establishing JSC DNN operation within the communication session, sends one or more control messages to the second device specifying the JSC DNN identifier of the selected pair of JSC DNN structures. The second device uses the received JSC DNN identifier from the second neural network table 1020 in Figure 10c to identify the JSC DNN structure for use in the second device to receive JSC DNN communication from the first device. In this case, the second device supports the first device when performing radar sensing.

[0174] Generally, the first, second, and third neural network tables 1010, 1020, and 1030 are provided to and used by the first, second, and third devices, respectively, but such neural network tables may be stored in any configuration or combination as needed. For example, the master neural network table 1000 may store together information from each of the described first, second, and third neural network tables 1010, 1020, and 1030, along with the corresponding elements accessed by the first, second, and / or third devices accordingly. Conversely, each of the neural network tables 1000, 1010, 1020, and / or 1030 may be stored in a distributed manner across several physical devices and may be accessible by the first, second, and / or third devices as required by the application. In other examples, the master neural network table 1000 or the first neural network table 1010 is stored in the first device, the second neural network table 1020 is stored in the second device, and the third neural network table 1030 is stored in the third device. In other examples, the master neural network table 1000 is stored in the second device. In other examples, the master neural network table 1000 is stored in a network device or entity of a communication system, where the first and / or second devices restrict access to various columns / rows of the master neural network table 1000 according to the granted authorizations and / or privileges.

[0175] Figure 10e shows a flowchart of an exemplary JSC DNN structure selection process 1050 of the first device during the establishment of JSC DNN operation with the second device. In this case, the first device accesses either the master neural network table 1000 in Figure 10a or the first neural network table 1010 in Figure 10b, and the second device accesses the second neural network table 1020 in Figure 10c. For example, the first neural network table 1010 includes, for each pair of JSC DNN structures with a JSC DNN identifier, associated communication channel conditions and associated JSC system types, as shown in the JSC DNN identifier column, the JSC DNN structure column of the first device, the communication channel conditions / characteristics column, and the JSC system type column in Figure 10b. In other examples, the JSC DNN structure selection process 1050 may be performed by a second device having a second neural network table 1020, which for each pair of JSC DNN structures with a JSC DNN identifier, includes associated communication channel conditions and associated JSC system types, as shown in the JSC DNN identifier column, the JSC DNN structure column of the second device, the communication channel conditions / characteristics column, and the JSC system type column in Figure 10c. The JSC DNN structure selection process 1050 of the first device includes the following steps:

[0176] Step 1051 identifies the communication channel conditions for the communication channel between the first device and the second device.

[0177] Step 1052 involves selecting a JSC system type for use in generating radar sensing information. The JSC system type includes at least one of the bistatic, multistatic, or monostatic JSC system types (or radar configurations) described with reference to Figures 1, 2, and / or Figures 3a to 3d, and / or described herein.

[0178] In step 1053, a JSC DNN identifier is selected from the master neural network table 1000 or the first neural network table 1010 based on the identified communication channel conditions and the selected JSC system type. In the example, the first device also selects a JSC DNN identifier based on the capabilities or processing capabilities of the second device, as illustrated with reference to Figure 10b. Optionally, the JSC DNN structure selection process 1050 may be performed by the second device (for example, when performing an uplink JSC), and in step 1053, a JSC DNN identifier may be selected from the second neural network table 1020.

[0179] After step 1053, the JSC DNN structure selection process 1050 proceeds to execute the JSC DNN process 800 or process 922 / 942 to establish JSC DNN operation during a communication session between the first device and the second device.

[0180] Optionally, once the second device establishes JSC DNN operation with the first device, the second device performs the JSC DNN structure selection process 1050, and in steps 1051-1053, the second neural network table 1020 is used instead of the first neural network table 1010.

[0181] Figure 11 shows a flowchart of an exemplary co-training process 1100 for training first and second JSC DNN structures (pairs of JSC DNN structures) for post-training use by a first device and a second device, respectively. The co-training process 1100 stores the corresponding trained first and second JSC DNN structures of the trained pair of JSC DNN structures in, for example, a master neural network table 1000, a first neural network table 1010, and / or a second neural network table 1020, as described with reference to Figures 10a to 10e. The first and second JSC DNN structures of each pair of JSC DNN structures include one or more transmitter and / or receiver DNN models that are co-trained for JSC performance requirements having a specific JSC system type (e.g., bistatic JSC, multistatic JSC, and monostatic JSC), one or more communication and / or radar sensing channel characteristics. Each of the transmitting DNN models and / or receiving DNN models includes any type of ML model algorithm / architecture. The co-training process 1100 co-trains the transmitting DNN models and / or receiving DNN models of the first and second JSC DNN structures using supervised or unsupervised learning. The ML model algorithm / architecture may be, but may not be limited to, one or more of the following: neural networks, fully connected neural networks, convolutional neural networks, long short-term memory (LSTM) neural networks, and transformer neural networks, and / or any other suitable DNN architecture, combinations thereof, or modifications thereof, as described herein and / or as required by the application.For example, supervised co-training of a pair of JSC DNN structures for a first device and a second device uses gradient backpropagation-based techniques to update the weights / parameters of the corresponding pair of JSC DNN structures using, for example, an integrated radar and communication loss function. It is assumed that the DNN model architecture and DNN model configuration have been determined for the pair of JSC DNN structures for the first device and the second device, as described with reference to Figures 3a to 3d. The co-training process 1100 includes the following steps:

[0182] Step 1160 acquires a training dataset, such as a batch of training data containing multiple training data instances, where each training data instance contains known sensing data (e.g., known range, Doppler, velocity, acceleration, position, and / or any other known radar measurement data) associated with one or more objects in an environment having a communication / radar channel. The communication / radar channel is associated with selected JSC performance requirements, such as the conditions / characteristics of a particular communication and / or radar channel (e.g., NLOS communication, LOS communication, multipath interference, multiple access interference, and narrowband interference, meteorological or atmospheric conditions affecting the communication / radar channel, different target sizes and types, and one or more other radar and / or communication signal interferences, as described with reference to Figures 1 to 9b). The selected JSC performance requirements may further include communication performance requirements for the communication channel / link between the first and second devices (e.g., communication performance requirements for a specific communication standard (e.g., 5G, 6G, etc.)), radar performance requirements for the first device when detecting objects, and so on.

[0183] In step 1162, for each training data instance in a batch, a first JSC DNN structure processes a set of input communication data, which may be randomly generated and / or included in the batch from which the training data was acquired. Each training data instance may represent a time slot. The first JSC DNN structure connects to a second JSC DNN structure via a communication channel model (or communication channel). The first JSC DNN structure processes the set of input communication data to generate an output JSC signal for the time slot, which is RF processed, passes through a communication channel model (e.g., simulating RF processing such as ADC and frequency upconversion, simulating transmission over a communication channel with one or more objects based on known sensing data of the training data instance and selected JSC performance requirements and / or channel conditions), and is passed to the second JSC DNN structure as a JSC signal waveform, along with reflections of the JSC signal waveform from one or more objects based on known sensing data of the training data instance. The output JSC signal is a digital representation of the set of input communication data transformed by the first JSC DNN structure to include the radar signal and / or radar signal characteristics of the JSC. The JSC signal waveform is the RF or analog representation of the output JSC signal. The communication channel model includes one or more objects arranged to be represented by acquired training data instances. The joint training process 1100 generates an output JSC signal given a set of input communication data and configures the first JSC DNN structure to transmit the output JSC signal, representing the set of input communication data and the radar signal / characteristics, as a JSC signal waveform via a communication channel model (or communication channel) for reception by the second JSC DNN structure. Simultaneously, the training process 110 configures the first JSC DNN structure to generate radar sensing information or data (e.g., range, Doppler, velocity, etc.) of one or more objects.A communication channel model models a communication channel based on communication channel conditions / characteristics and one or more corresponding objects.

[0184] The second device receives a JSC signal waveform as a JSC signal, which is input to a second JSC DNN structure to generate a set of reconstructed communication data corresponding to a set of input communication data, and / or an RFB signal waveform based on radar sensing feedback (RFB) information associated with one or more objects, depending on the JSC system type (e.g., bistatic or multistatic) from which the pair of JSC DNN structures are trained, as well as a reflection of the transmitted JSC signal waveform received in the second device via the communication channel model. Bistatic and multistatic JSC systems use the RFB signal waveform as described with reference to Figures 3a, 3b, and 3d, and / or as described herein. In the case of bistatic and multistatic JSC system types, a second JSC DNN structure provides RFB information (e.g., Doppler, velocity, position / location, etc.), which the second device transmits to the first device as an RFB signal waveform by a communication channel model (e.g., simulation of RF processing such as ADC and frequency upconversion, simulation of transmission over the communication channel, etc.). The first device processes the RFB signal waveform (e.g., by simulating RF processing such as DAC and frequency downconversion to baseband, etc.) to generate a first radar sensing signal (or more) representing the RFB signal waveform and the RFB information within it. These first radar sensing signals are input to the first JSC DNN structure to generate radar sensing information (e.g., Doppler, velocity, position / location, etc.) for one or more detected objects. In the case of a multistatic JSC system type, the communication channel model also processes the JSC signal waveforms of time slots and provides the first JSC DNN structure with further radar sensing signals derived from the reflections of the JSC signal waveforms for one or more objects modeled in the communication channel model and received by the first device.These additional radar sensing signals are input to a first JSC DNN structure along with the first radar sensing signals to generate radar sensing information about one or more objects to be detected. In the case of a monostatic JSC system type, the communication channel model processes the JSC signal waveforms for time slots and provides the radar sensing signals to the first JSC DNN structure using only the reflections of the JSC signal waveforms for one or more objects modeled by the communication channel model and received by the first device. These radar sensing signals are processed by the first JSC DNN structure to generate radar sensing information about one or more objects to be detected.

[0185] The radar sensing information for a time slot (i.e., per training data instance) relating to one or more detected objects (e.g., range, Doppler, velocity of the radar object) represents one or more from a group such as per-object range estimate, per-object Doppler estimate, per-object velocity estimate, per-object position or location estimate, per-object delay spread, per-object Doppler spread, per-object mean delay, per-object angle estimate, per-object azimuth and / or elevation estimate, or other radar measurements, relating to one or more objects in a time slot (i.e., per training data instance). Step 1162 is performed for each training data instance in a batch of training datasets.

[0186] The generation of radar sensing information (or radar measurements) is performed by a first JSC DNN structure, and the DNN model of the first JSC DNN structure learns which type of radar measurement (e.g., correlation) is best given the JSC signal waveform output and reflected from one or more objects. That is, the DNN architecture of the first JSC DNN structure converges during training to the optimal scheme and / or JSC signal waveform for performing JSC for a specific selected JSC system type and set of JSC requirements. The DNN model of the second JSC DNN structure learns how to separate received radar / reflections from the data signals of the transmitted JSC signal waveform and how to generate reconstructed communication data corresponding to the input communication data embedded in the JSC signal waveform. The communication data signals and radar signals in the JSC signal waveform may be subject to constraints that they are separated because they use different resources (e.g., time / frequency / space / code, etc.). For example, training DNN models for first and second JSC DNN structures can converge to a solution where radar sensing may use different resources (time / frequency / space / code). For instance, during training, or due to constraints, the JSC signal waveform may result in the input communication data and radar signal occupying different communication resources (e.g., time / frequency / space / code). In such a case, the first device can filter out the transmitted input communication data from the received JSC signal waveform so that only the radar reflection of the JSC signal waveform is included in the radar sensing signal processed by the DNN model of the first JSC DNN structure. For multiple second devices (e.g., multiple UEs) communicating with the first device (e.g., BS), the first device can support the multiple second devices using different communication resources (e.g., different time slots or frequency / RB / bandwidth portions (BWP)).Alternatively or additionally, joint training of a first JSC DNN structure and a second JSC DNN structure may converge to produce an output JSC signal, which, when transmitted, forms an optimal JSC signal waveform with characteristics incorporating the input communication data and the radar signal / radar signal characteristics of the JSC, but is constrained within JSC performance requirements (e.g., communication and / or radar performance requirements, and / or performance requirements of one or more communication standards (e.g., 4G, 5G, and / or 6G, etc.)).

[0187] In step 1164, after all training data instances in a batch have been processed, a joint loss function is calculated using performance metrics associated with the radar sensing information generated by the first JSC DNN structure and the reconstructed set of transmitted communication data for the batch generated by the second JSC DNN structure. For example, the joint loss function is a combination of the performance metrics associated with the radar sensing information generated by the first JSC DNN structure of the first device and the performance metrics associated with the reconstructed communication data generated by the second JSC DNN structure of the second device in the batch. For example, the performance metric associated with the generated radar sensing information is the radar sensing error rate (or sensing error) derived from the radar sensing information generated by the first JSC DNN structure and the corresponding known radar sensor data of the corresponding training data instances in the batch. For example, the radar sensing error rate for a batch is calculated from the mean squared error of the radar sensing information and known radar sensing data for the corresponding training data instances in the batch. Other error metrics include, for example, squared error, least squares error, and / or any other error metric appropriate depending on the application requirements. The performance metrics associated with the reconstructed communication data are the block or bit error rate (or block error, bit error, signal-to-interference-to-noise ratio (SINR), or other communication performance metrics) derived from the reconstructed communication data generated by the second JSC DNN structure and the corresponding input communication dataset associated with the batch's training data instances. The integrated radar and communication loss function uses the combined performance metrics (e.g., radar sensing error rate and block or bit error rate) when updating the weights of the DNN model placements in the first and second JSC DNN structures.For example, the integrated radar and communication loss function is a weighted sum of the radar sensing error rate (or sensing error) and the blocking error rate (or blocking error). For example, a gradient backpropagation-based technique using the integrated radar and communication loss function can be used to update the weights of the DNN model for the first and second DNN structures.

[0188] Step 1166 determines whether the JSC DNN structures of the first and second devices are adequately trained. For example, using the joint radar and communication loss functions, it is determined whether the joint loss falls below a given joint loss threshold. If so, the JSC DNN structure may be considered adequately trained; otherwise, further training of the JSC DNN structure may be performed. If the JSC DNN structures of the first and second devices are adequately trained (e.g., "Y"), proceed to step 1170; otherwise (e.g., "N"), proceed to step 1168 to update the model parameters and / or weights.

[0189] In step 1168, the model parameters and / or weights of the DNN models in the first and second JSC DNN structures of the pair of JSC DNN structures for the first and second devices are updated based on the joint radar and communication loss functions. For example, in supervised joint training of the first and second JSC DNN structures of the pair of JSC DNN structures for the first and second devices, the weights of the corresponding DNN models in the first and second JSC DNN structures are updated using a gradient backpropagation-based technique that uses the joint radar and communication loss functions. After the update, the process proceeds to step 1160 to obtain other batches of training data instances. The first JSC DNN structure and the second JSC DNN structure are jointly trained to incorporate the input communication data so that the JSC signal waveform is the optimal and / or appropriate waveform that can accommodate channel conditions (e.g., based on selected JSC requirements) for use in both a) radar sensing using the reflection of the JSC signal waveform from one or more objects and / or the RFB signal waveform from the second JSC DNN structure, and b) communication with a second device, where the second JSC DNN structure is trained and optimized to generate reconstructed communication data corresponding to the input communication data incorporated into the transmitted JSC signal waveform.

[0190] In step 1170, after training, the collaborative training process 1100 stores the selection and use of pairs of trained JSC DNN structures (e.g., sent to the master neural network table 1000 in Figure 10a), JSC DNN operation, processes, systems and / or devices, as described with reference to Figures 1 to 10e. For example, the trained first and second JSC DNN structures form a pair of trained JSC DNN structures for the first and second devices, and the JSC system type on which the pair of trained JSC DNN structures were trained, along with the channel / radar conditions and / or characteristics, are assigned to a JSC identifier (or index), as described with reference to Figure 10a, and stored in the master neural network table 1000. For example, the pair of trained JSC DNN structures for the first and second devices are assigned to a JSC identifier (or index) and stored in a row in the master neural network table 1000. The network device or the first device populates a trained first JSC DNN structure in a first neural network table 1010, along with channel conditions, JSC system type, and JSC identifier, as described with reference to Figures 10a and 10b. The network device or the second device populates a trained second JSC DNN structure in a second neural network table 1020, along with channel conditions, JSC system type, and JSC identifier, as described with reference to Figures 10a and 10c.

[0191] In the example above, as illustrated with reference to Figures 1 to 3d and Figures 10a to 10e, the first JSC DNN structure includes a transmitting JSC DNN model and a receiving radar sensing DNN model. After training and storage in the master neural network table 1000 in Figure 10a, the transmitting JSC DNN model can be configured to process input communication data in a time slot and to generate an output JSC signal for transmission as a JSC signal waveform within the time slot. The output JSC signal and JSC signal waveform represent the input communication data in the time slot and the radar signal in the time slot. The receiving radar sensing DNN model can be configured to process one or more received radar sensing signals and to generate predicted radar sensing information (e.g., range, Doppler, velocity, position, acceleration, and / or any other parameters) for one or more objects.

[0192] As illustrated with reference to Figures 3a to 3d, for example, the transmitting JSC DNN model of the first JSC DNN structure outputs a feedforward signal to the receiving radar sensing DNN (Rx R-DNN) model of the first JSC DNN structure for each time slot. The first device configures the receiving radar sensing DNN model to process the feedforward signal of the time slot and one or more radar sensing signals received in the time slot, and generates radar sensing information for one or more objects. In the example, the feedforward signal includes one or more of the time slot's input communication data, one or more hidden layer outputs of the time slot's transmitting JSC DNN model, one or more hidden layer inputs of the time slot's transmitting JSC DNN model, and the time slot's output JSC signals.

[0193] In a further example, as illustrated with reference to Figures 2, 3a, 3b, and 3d, and Figures 7a-7b, the second JSC DNN structure of the second device includes a receiving-side communication and radar (CR) DNN (Rx CR-DNN) model and a transmitting-side radar feedback (RFB) DNN (Tx RFB-DNN) model. The second device configures the Rx CR-DNN model to process data representing one or more received JSC signal waveforms of a time slot, to generate reconstructed communication data associated with the input communication data of the time slot, and to generate radar sensing feedback (RFB) information of the time slot. The second device configures the Tx RFB DNN model to process the generated RFB information, as illustrated with reference to, for example, Figures 2, 3a, 3b, and 3d, and Figures 7a-7b, and to generate an RFB signal for transmission to the first device as an RFB signal waveform over a communication channel.

[0194] The collaborative training process 1100 may be applied to any of the JSC DNN structures 103 and 114, 203 and 214, and 303 and 314, combinations thereof, or modifications thereof, as described herein and / or as required by the application, with reference to Figures 1 to 10e. In one example, the collaborative training process 1100 uses a simulated communication channel model that randomly generates input communication data for each training data instance within each batch of training data instances. In another example, the collaborative training process 1100 uses an online real communication channel, in which the input communication data is real communication data for transmission over the real communication channel from a first device to a second device, and the radar sensing information associated with the object is known, or the sensed object is a known object having known range, Doppler, velocity, acceleration, position and / or other known radar measurements. Techniques are known for estimating the range, Doppler, velocity, and other known radar measurements of an object to train a first JSC DNN structure on a first device and a second JSC DNN structure on a second device. In other examples, a joint training process 1100 uses federated learning techniques and / or any other real-time machine learning techniques to train the first and second JSC DNN structures in real time on the first and second devices, etc.

[0195] As illustrated with reference to Figures 1 to 11, the JSC signal waveform includes both the radar sensing signal and the input communication data, adapted for transmission. The JSC signal waveform, incorporating the input communication data and radar signal characteristics / or radar signal, is learned during joint training of the DNN model in the first and second JSC DNN structures. The DNN model learns or converges to the optimal JSC signal waveform given a specific JSC system type and set and / or other constraints of the JSC performance requirements, using feedback of the joint cost function or joint loss function optimized by the first and second JSC DNN structures. The joint cost or loss function can be the sum of radar sensing errors (e.g., Doppler error, range error, velocity error, etc.) and input communication data errors (e.g., block error rate (BLER)). The radar sensing errors are based on error metrics such as the mean squared error of the radar sensing errors (e.g., range error, Doppler error, etc.). Gradient backpropagation techniques or other DNN updating techniques are applied to update the weights / parameters of the DNN models of the first and second JSC DNN structures with the aim of minimizing the joint cost function / joint loss function. Upon convergence, the result is a trained first JSC DNN structure that generates an optimal output JSC signal and subsequent transmitted JSC signal waveform for the JSC system type and set of JSC performance requirements used when training the first and second JSC DNN structures, the first JSC DNN structure then generates a trained second JSC DNN structure that extracts and generates radar sensing information based on reflections of the JSC signal waveform from one or more objects, and reconstructed communication data from the JSC signal waveform, and / or generates an RFB feedback waveform to assist the first JSC DNN structure in generating the radar sensing information.

[0196] Figure 12 shows non-temporary media 1200 (or computer program product) in several embodiments. Non-temporary media 1200 may include a computer-readable storage medium 1202 (or computer-readable medium) and / or an input / output mechanism 1204 for enabling a computing system to access the computer-readable storage medium 1202. In this example, the non-temporary media is a Universal Serial Bus (USB) stick, but this is merely an example and is not limited thereto. Those skilled in the art will understand that non-temporary media 1200 may be any other type of computer-readable medium, one or more computer-readable media, such as a compact disc, digital video disc, USB stick, Blu-ray disc, flash drive, etc., and / or any other computer-readable medium as required by the application. The non-temporary medium 1200 stores computer program code or instructions, which, when executed by the device's processor, cause the device to perform one or more of the aforementioned process methods, operations, or processes, such as those disclosed with respect to the flowcharts and schematics in Figures 1 to 11 and their associated features.

[0197] Embodiments of the methods or processes described herein can be implemented as digital electronic circuits, integrated circuits, specially designed ASICs (application-specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These may include computer program products (e.g., software stored on magnetic disks, optical disks, memory, programmable logic devices, etc.), which may include stored computer-readable instructions, which, when executed by a processor, cause the processor to perform one or more of the methods described herein.

[0198] Any system features described herein may also be provided as method or process features, and vice versa. Where used herein, means-plus-function features may be expressed alternatively with respect to their corresponding structures. In particular, method embodiments may be applied to system embodiments, and vice versa. Furthermore, any, some, and / or all features in one embodiment may be applied to any, some, and / or all features in any other embodiment, in any suitable combination. It should also be understood that specific combinations of various features described and defined in any embodiment of the present invention may be independently implemented and / or supplied and / or used. While several embodiments have been shown and described, those skilled in the art will understand that modifications to these embodiments may be made without departing from the principles of this disclosure, and the scope of such modifications is defined in the claims and their equivalents.

Claims

1. A method performed by a first device (101), This includes establishing the operation of an integrated sensing and communication (JSC) deep neural network (DNN) for one or more time slots of a communication session with a second device (112) (802), For each of the one or more time slots, the method further: For each of the aforementioned time slots, input communication data (102) for transmission to the second device is acquired (804), To generate the input communication data (102) and the output JSC signal representing the radar signal, the input communication data is processed (806) by the JSC DNN structure (103) of the first device (101), The output JSC signal is transmitted (808) to the second device (112) via the communication channel (108) as a JSC signal waveform (107) for each time slot. Based on the reflection (110a-110n, 111a-111n) of the transmitted JSC signal waveform (107) from one or more objects (109a-109n), one or more radar sensing signals (117, 110a-110n) are received for each time slot (810), To generate radar sensing information (105) of one or more objects (109a to 109n) in the communication channel, the JSC DNN structure (103) of the first device (101) processes the one or more received radar sensing signals (812), A method comprising transmitting (814) the radar sensing information (105) of one or more objects (109a to 109n) for each time slot to one or more upper protocol layers of the protocol stack of the first device.

2. Establishing the JSC DNN operation (802) is, Sending one or more control messages to the second device (112) (802a) that indicate the JSC DNN structure selected for use per time slot, The method according to claim 1, comprising configuring the JSC DNN structure (103) of the first device (101) with the selected JSC DNN structure (802b) for use in processing the input communication data (102) and receiving one or more radar sensing signals (117, 110a to 110n).

3. The method further comprises the first device selecting a pair of JSC DNN structures from a neural network table (1010) containing a plurality of pairs of JSC DNN structures, each pair of JSC DNN structures comprising a first JSC DNN structure for use by the first device and a second JSC DNN structure for use by the second device, and each pair of JSC DNN structures being identified by a JSC DNN identifier. The one or more control messages include the JSC DNN identifier of the selected pair of JSC DNN structures, The method according to claim 2, wherein the JSC DNN structure of the first device comprises the first JSC DNN structure of the selected pair of JSC DNN structures.

4. The neural network table (1010) further includes, for each pair of JSC DNN structures, associated communication channel conditions and associated JSC system types, and the method further, Identifying (1051) the communication channel conditions of the communication channel (108) between the first device (101) and the second device (112), Select the JSC system type for use when generating radar sensing information (1052), The method according to claim 3, further comprising selecting the JSC DNN identifier from the neural network table (1010) (1053) based on the identified communication channel conditions and the selected JSC system type.

5. The method according to claim 4, wherein the communication channel conditions include one or more of the group of conditions of any other type affecting the communication channel between the first device (101) and the second device (112) of the JSC, such as: beyond line of sight communication channel conditions, within line of sight communication channel conditions, meteorological or atmospheric conditions, radar channel conditions, target characteristics, channel throughput, channel frequency or frequency band, channel bandwidth, channel delay spread, channel Doppler spread, channel angle spread, or any other type of condition affecting the communication channel between the first device (101) and the second device (112) of the JSC.

6. The aforementioned JSC system type is Monostatic JSC system type, Bistatic JSC system type, or Multistatic JSC system type, The method according to any one of claims 4 or 5, comprising at least one of the following.

7. Receiving one or more radar sensing signals (810) and processing one or more received radar sensing signals (812) further means Receiving one or more reflected JSC signal waveforms (110a to 110n) of the time slot by the reflection of the transmitted JSC signal waveform (107) from one or more objects (109a to 109n) (810), This includes processing one or more received reflected JSC signal waveforms (110a to 110n) of the time slot into one or more radar sensing signals of the time slot, from radio frequency RF to baseband. The method according to any of the prior claims, wherein the processing (812) further comprises processing the one or more radar sensing signals of the time slots using the JSC DNN structure (103) of the first device (101) to generate radar sensing information (105) of the time slots corresponding to the one or more objects (109a to 109n).

8. Receiving (810) one or more radar sensing signals and processing (812) one or more received radar sensing signals further, The radar sensing feedback signal waveform (117) generated using the reflections (111a to 111n) of the JSC signal waveform (107) transmitted from one or more objects (109a to 109n) is received (810) from the second device (112), This includes processing the received radar sensing feedback signal waveform (117) of the time slot from radio frequency RF to baseband to obtain one or more further radar sensing signals of the time slot. The method according to any of the prior claims, wherein the processing (812) further comprises processing the one or more further radar sensing signals of the time slot using the JSC DNN structure (103) of the first device (101) to generate radar sensing information (105) of the time slot corresponding to the one or more objects (109a to 109n).

9. The method according to any of the prior claims, further comprising co-training (1100) the JSC DNN structure (103) of the first device (101) and the JSC DNN structure (114) of the second device (112) via a communication channel (108) with the one or more objects (109a to 109n) using a training dataset comprising known radar sensor data associated with one or more objects (109a to 109n) and a corresponding set of input communication data, wherein the weights of the JSC DNN structures (103, 114) of the first and second devices (101, 112) are updated based on a joint loss function which includes a combination of performance metrics associated with the radar sensing information generated by the JSC DNN structure (103) of the first device (101) and the reconstructed communication data generated by the JSC DNN structure (114) of the second device (112).

10. The method according to claim 9, wherein the performance metric associated with the generated radar sensing information is a radar sensing error rate derived from the generated radar sensing information of the JSC DNN structure (103) of the first device (101) and the corresponding known radar sensor data of the training dataset, and the performance metric associated with the reconstructed communication data is a block or bit error rate derived from the reconstructed communication data generated from the JSC DNN structure (114) of the second device (112) and the corresponding set of input communication data associated with the training dataset.

11. The method according to claim 9 or 10, wherein the JSC DNN structure (303) of the first device (101) includes a transmitting JSC DNN (303a) and a receiving radar sensing DNN (303b), the transmitting JSC DNN (303a) is configured to process the input communication data (302) of the time slot and to generate an output JSC signal for transmission within the time slot, and the receiving radar sensing DNN (303b) is configured to process one or more radar sensing signals and to generate predictive radar sensing information (305) of one or more objects (309a to 309n).

12. The method according to claim 11, wherein the transmitting JSC DNN (303a) is configured to output a feedforward signal (303c) to the receiving radar sensing DNN (303b) of the time slot, and the receiving radar sensing DNN (303b) is configured to process the feedforward signal (303c) of the time slot and the one or more radar sensing signals of the time slot, and generates radar sensing information (305) of the one or more objects (309a to 309n).

13. The aforementioned feedforward signal (303c) is The communication data entered into the aforementioned time slot, One or more hidden layer outputs of the transmitting JSC DNN (303a), or The output JSC signal of the aforementioned time slot, The method according to claim 12, comprising one or more of the above.

14. Using the resource block and / or frequency set of the time slot, the output JSC signal of the JSC DNN structure (103) of the first device (101) is processed for transmission, and the method further, The transmission of the output JSC signal includes sending a control message (802a) to the second device (112) that instructs the set of resource blocks and / or frequencies of the time slot, The method according to any one of claims 1 to 13, wherein transmitting the output JSC signal to the second device (112) as a JSC signal waveform (107) in the time slot (808) further comprises transmitting the JSC signal waveform (107) in the resource block and / or set of frequencies.

15. The output JSC signal is further transmitted (808) to the second device (112) via the communication channel (108) as a JSC signal waveform (107) for each time slot. The process includes processing the output JSC signal generated by the JSC DNN structure (103) of the first device (101) into a radio frequency RF for transmission as a JSC signal waveform for each time slot, wherein the JSC signal waveform represents the input communication data and the radar signal, and the transmission further includes, The method according to any one of claims 1 to 14, comprising transmitting the JSC signal waveform (107) for each time slot to the second device (112) via the communication channel (108) by one or more objects (109a to 109n) (808).

16. The method according to any one of claims 1 to 15, wherein the received radar sensing signal for each time slot includes one or more component radar signals or radar signal characteristics associated with the reflections (110a to 110n, 111a to 111n) of the transmitted JSC signal waveform (107) for the time slot from one or more objects (109a to 109n).

17. The method according to any one of claims 1 to 16, wherein establishing the JSC DNN operation for one or more time slots (802) includes sending a control message that includes a JSC DNN identifier and the one or more time slots as a downlink control indicator (DCI) message.

18. The method according to claim 17, wherein transmitting the control messages includes transmitting each control message as a radio resource control (RRC) message (802a).

19. Sending the aforementioned control message means JSC DNN identifier, A set of one or more primary time slots used for JSC DNN operation. For use when transmitting the JSC signal waveform (107) for each time slot of the first set of time slots, a first resource block and / or set of frequencies, When transmitting the radar sensing feedback signal waveform (117), one or more sets of second time slots for use by the second device, When transmitting a radar sensing feedback signal waveform (117) for each time slot of the second set of time slots, a second set of resource blocks and / or frequencies for use by the second device, The method according to any one of claim 17 or 18, comprising transmitting the control message (802a) which includes one or more of the following:

20. The generation of the radar sensing information is Range estimates for each related object, Doppler estimates for each related object, Velocity estimates for each related object, Location or estimated position of each related object, Delayed spread, Doppler spread, Average delay, Angle estimates for each related object, Estimated azimuth and / or elevation angles for each related object, or Any other appropriate radar sensing information or parameters estimated for each related object, A method according to any of the prior claims, comprising generating radar sensing information which includes one or more of the above.

21. A method performed by a second device (112), This includes establishing the operation of an integrated sensing and communication (JSC) deep neural network (DNN) for one or more time slots of a communication session with a first device (101) (922), For each of the one or more time slots, the method further: The method further includes receiving (924) a JSC signal waveform (107) transmitted from the first device (101) for each time slot via a communication channel, wherein the JSC signal waveform represents input communication data and radar signals, and the method further includes Based on the reflections (111a to 111n) of the transmitted JSC signal waveform (107) from one or more objects in the communication channel, one or more further JSC signal waveforms are received (926) for each time slot, To generate reconstructed communication data corresponding to the input communication data and radar sensing feedback signals associated with one or more objects, the received JSC signal waveform and further JSC signal waveforms (107, 111a to 111n) are processed in the second device (112) by a JSC DNN structure (114) (928 to 930), For use in generating radar sensing information relating to one or more of the aforementioned objects, the radar sensing feedback signal is transmitted to the first device (101) as a radar sensing feedback signal waveform (117) (932), A method comprising transmitting the reconstructed communication data to the data sink of the second device (934).

22. Establishing the JSC DNN operation for one or more time slots of the communication session with the first device (101) (422) is: For use by the second device (112) when processing one or more of the aforementioned received JSC signal waveforms (107, 111a to 111n), one or more control messages indicating the type of selected JSC DNN structure are received from the first device (101) (922a), The method according to claim 21, comprising configuring the JSC DNN structure (114) of the second device (112) based on the type of the selected JSC DNN structure for use within one or more time slots (423, 923).

23. The neural network table (1020) further includes storing data representing multiple JSC DNN structures, each JSC DNN structure including a unique JSC DNN identifier. Receiving one or more control messages from the first device (101) (922a) means that the one or more control messages from the first device (101) include a JSC DNN identifier of the selected JSC DNN structure type, The method according to claim 22, wherein configuring (923) the JSC DNN structure (114) of the second device (112) uses the neural network table (1020) and the received JSC DNN identifier for use within one or more time slots.

24. The neural network table (1020) further includes, for each JSC DNN structure, the communication channel conditions on which the JSC DNN structure is trained, and the JSC system type to which each of the JSC DNN structures is associated, and the method further, Select the JSC DNN structure using the received JSC DNN identifier from the neural network table (1020), The method according to claim 23, comprising configuring the selected JSC DNN structure of the second device (112) in the selected JSC DNN structure of the one or more time slots.

25. The above method further, The method according to any one of claims 21 to 24, comprising co-training (1100) the JSC DNN structure (114) of the second device (112) and the JSC DNN structure (103) of the first device (101) via channel communication (108) with the one or more objects (109a to 109n) using a training dataset comprising known radar sensor data associated with one or more objects (109a to 109n) and a corresponding set of input communication data, wherein the weights of the JSC DNN structures (103, 114) of the first and second devices (101, 112) are updated based on a joint loss function which includes a combination of performance metrics associated with radar sensing information generated by the JSC DNN structure (103) of the first device (101) and reconstructed communication data generated by the JSC DNN structure (114) of the second device (112).

26. The method according to claim 25, wherein the performance metric associated with the generated radar sensing information is a radar sensing error rate derived from the radar sensing information generated by the JSC DNN structure (103) of the first device (101) and the corresponding known radar sensor data of the training dataset, and the performance metric associated with the reconstructed communication data is a block or bit error rate derived from the reconstructed communication data generated by the JSC DNN structure (114) of the second device (112) and the corresponding input communication data associated with the training dataset.

27. The JSC DNN structure (114) of the second device (312) includes a receiving-side communication and radar (CR) DNN (314a) and a transmitting-side radar feedback (RFB) DNN (314b), The receiving CR DNN (314a) is configured to process one or more received JSC signal waveforms (311a to 311n) of the time slot and to generate reconstructed communication data (315) associated with the input communication data (302) of the time slot and radar sensing feedback information (314d) of the time slot. The method according to any one of claims 21 to 26, wherein the transmitting RFB DNN (314b) is configured to process the generated radar sensing feedback information (314d) and to generate a radar sensing feedback signal for transmission to the first device (301) as a radar sensing feedback signal waveform (317).

28. The method according to claim 27, further comprising transmitting the generated radar sensing feedback information to the first device (301) in a conventional manner within one or more further time slots.

29. The method according to any one of claims 21 to 28, wherein the JSC DNN structure (114) of the second device (312) includes a receiving-side communication DNN (314c), the receiving-side communication DNN (314c) is configured to process one or more received JSC signal waveforms (311a to 311n) of the time slot, and is configured to generate reconstructed communication data (315) corresponding to the input communication data (302) of the time slot.

30. A method performed by a second device (112), This includes establishing the operation of an integrated sensing and communication (JSC) deep neural network (DNN) for one or more time slots of a communication session with a first device (101) (922), For each of the one or more time slots, the method further: The method further includes receiving (924) a JSC signal waveform transmitted from the first device (101) for each time slot via a communication channel, wherein the JSC signal waveform represents input communication data and radar signals, and the method further includes Based on the reflection of the transmitted JSC signal waveform from one or more objects in the communication channel, one or more additional JSC signal waveforms are received for each time slot (926), The method further includes processing the received JSC signal waveform and a further JSC signal waveform in the second device (112) using a receiving-side communication DNN structure (314c) (942), wherein the receiving-side communication DNN structure (314c) is configured to generate reconstructed communication data corresponding to the input communication data transmitted from the first device (101), and the method further includes A method comprising transmitting the reconstructed communication data to one or more upper-layer protocols of the protocol stack of the second device (944).

31. A method performed by a third device (343), This includes establishing the operation of an integrated sensing and communication (JSC) deep neural network (DNN) for one or more time slots of a communication session with a first device (101), For each of the one or more time slots, the method further: The method includes receiving a JSC signal waveform (307) transmitted from the first device (101) to the second device (112) for each time slot via a communication channel (308), wherein the JSC signal waveform (307) represents input communication data and radar signals of the second device (112), and the method further includes: Based on the reflection of the transmitted JSC signal waveform (307) from one or more objects (309a to 309n) in the communication channel, one or more further JSC signal waveforms (340a to 340n) are received for each time slot. The method includes processing the received JSC signal waveform (307) and further JSC signal waveforms (340a to 340n) in the third device (341) using a radar sensing DNN structure (343), wherein the radar sensing DNN structure (343) is configured to generate radar sensing feedback information relating to one or more objects (309a to 309n), and the method further includes, A method comprising transmitting a radar sensing feedback signal waveform (345) containing the generated radar sensing feedback information of one or more objects (309a to 309n) to the first device (101) for use by the first device (101) when generating radar sensing information for one or more objects (309a to 309n) in the time slot.

32. The method according to any one of the prior claims, wherein the first device is a base station and the second device is user equipment.

33. The method according to any one of claims 1 to 31, wherein the first device is a user device and the second device is a base station.

34. A computer program product comprising computer-readable instructions, wherein, when executed by a computer, the computer-readable instructions cause the computer to perform the method described in any of the prior claims.

35. The first device, A first device comprising one or more processors and memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the first device causes the first device to perform the method according to any one of claims 1 to 20.

36. The second device is A second device comprising one or more processors and memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the second device causes the second device to perform the method according to any one of claims 21 to 30.

37. A third device, A third device comprising one or more processors and memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the third device causes the third device to perform the method according to claim 31.

38. It is a device, One or more antennas, One or more processors, Equipped with memory, An apparatus wherein one or more processors are connected to the memory and one or more antennas, the memory further stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the apparatus causes the apparatus to perform the method according to any one of claims 1 to 33.

39. A first device configured according to claim 35, A second device configured according to claim 36, A communication system comprising the first device and the second device, which establishes JSC DNN operation for one or more time slots of a communication session.

40. The communication system according to claim 39, further comprising a third device configured according to claim 37, wherein the first device and the third device establish JSC DNN operation for one or more time slots of the communication session.

41. A computer-readable medium containing stored instructions, wherein, when executed, the instructions cause one or more processors to perform the method according to any one of claims 1 to 33.

Images