Deep learning assisted fingerprint based beam alignment

Abstract

Some implementations of a method may include: obtaining input data including user equipment location, number of user equipment, and desired received signal strength; processing the input data with a neural network having weights determined from a training phase to generate a set of one or more beampair indices; performing a beam search on at least one subset of the set of beampair indices; and receiving at least one beampair indices from the vehicle that provide the desired received signal strength.

Description

[0001] Cross-references to related applications

[0002] This application is a non-provisional filing of U.S. Provisional Patent Application Serial No. 62 / 894,666 entitled “Deep Learning Aided Fingerprintbased Beam Alignment”, filed on August 30, 2019, and claims the benefit of the patent application under 35 U.S.SC §119(e), the entire contents of which are incorporated herein by reference. Background Technology

[0003] Due to the high propagation loss observed at millimeter-wave frequencies, directional transmission is used. Aiming at the beam for both departure and arrival can be challenging due to the narrow beam and the high sensitivity of millimeter-wave frequencies to congestion.

[0004] Establishing good beam alignment utilizes knowledge of both the angle of arrival (AoA) and the angle of departure (AoD). The AoA-AoD pair can be determined as part of the channel estimation. However, performing channel estimation before beam alignment prevents the use of beamforming gain, which is crucial for reliable communication over millimeter-wave links. Therefore, to circumvent this problem, conventional beam alignment is performed using exhaustive beam scanning, where the base station and user terminal perform beam searches for all 360° × 360° beam pairs in all beam directions. This exhaustive search for all beam pairs can incur significant overhead due to the complexity of selecting the desired beam direction. To reduce the comprehensive search involved in beam scanning, hierarchical beam alignment has been proposed. Many such techniques rely on multi-resolution codebooks used at different levels. This technique, reminiscent of a bisection search algorithm, performs a beam search on a low-level codebook associated with a wider beam and another beam search on a high-level codebook, which is a subset of the wide beams selected from the low-level codebook. Unfortunately, many hierarchical codebook-based beam alignment techniques have not significantly reduced search complexity. Other operations aimed at reducing search complexity often involve optimization techniques. However, these methods may only be useful if the objective function exhibits smoothness. These techniques may only be applicable to objective functions without local optima.

[0005] To reduce the overhead and search complexity involved in beam alignment, blind beam control relying on accurate location information is proposed. A technique for inferring the line-of-sight (LOS) direction between communication devices is also presented. Furthermore, a beam-switching strategy is proposed that maximizes the rate by invoking a classical gradient descent method, assuming the channel always exhibits a single dominant LOS path. However, these methods may be impractical in practice when the LOS path no longer exists, often during heavy traffic when the dominant LOS path is blocked by obstacles. Summary of the Invention

[0006] Exemplary methods according to some implementations may include: obtaining input data including user equipment location, number of user equipments (UEs), and desired received signal strength; processing the input data with a neural network having weights determined from a training phase to generate a set of one or more beampair indices; performing a beam search on at least one subset of the set of beampair indices; and receiving from the user equipment at least one beampair indices providing the desired received signal strength.

[0007] For some implementations of the exemplary method, an out-of-band signaling channel can be used to transmit the location of the user equipment from the user equipment to the base station.

[0008] For some implementations of the exemplary method, the base station can know the number of UEs based on vehicle density.

[0009] For some implementations of the exemplary method, the softmax algorithm can be used to further generate a set of one or more beam pair indices.

[0010] For some implementations of the exemplary method, the softmax algorithm can generate probabilities associated with a set of beam pair indices.

[0011] For some implementations of the exemplary method, a set of one or more beam pair indexes may be stored in a database during the training phase.

[0012] For some implementations of the exemplary method, the training phase may include: obtaining training samples for each training location; initializing the weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: using the corresponding weight vector to compute the neuron output for each layer; applying the softmax function to obtain the class probability; compute the weight matrix and bias vector; and perform backpropagation of the error.

[0013] For some implementations of the exemplary method, processing the input data with a neural network may include: obtaining multiple fingerprints of different traffic conditions at the location of the user equipment; and using a neural network coupled to a softmax classifier to select one of the multiple fingerprints based on the traffic conditions at the location of the user equipment, wherein selecting one of the multiple fingerprints may generate a set of one or more beam pair indices, and the neural network may use weights determined from the training phase.

[0014] For some implementations of the exemplary method, traffic conditions may include the number of UEs at the user equipment location.

[0015] For some implementations of the exemplary method, the neural network can be a deep learning feedforward neural network.

[0016] For some implementations of the exemplary method, performing beam search may include: sending fingerprint information to a user equipment comprising a set of one or more beam pair indices; and performing a beam training process on at least a subset of the set of beam pair indices to select selected beam pairs from the set of one or more beam pair indices that satisfy the desired received signal strength.

[0017] Some implementations of the exemplary method may further include: obtaining training samples at at least one training location; initializing the weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: using the corresponding weight vector to compute the neuron outputs of at least one layer of the neural network; applying a softmax function to the output layer of the neural network to obtain class probabilities; updating the weight matrix and bias vector; and performing backpropagation of the error.

[0018] Some implementations of the exemplary method may also include determining a loss function between the predicted class probabilities and the true class probabilities, wherein a convergence metric threshold is reached if the loss function is less than the convergence metric threshold.

[0019] Some implementations of the exemplary method may also include selecting at least one beam pair index from a set of beam pair indexes for transmitting data to a receiver.

[0020] For some implementations of the exemplary method, selecting at least one beam pair index may include using a multi-functional beam transmission scheme.

[0021] In some implementations of the exemplary method, the selection of at least one beam pair index can be repeated periodically.

[0022] In some implementations of the exemplary method, the selection of at least one beam pair index can be performed when an event is triggered, and the triggering event can be the detection of a change in the parameters of the user equipment.

[0023] Exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, are operable to perform any of the exemplary methods.

[0024] Additional exemplary methods according to some implementation schemes may include: obtaining location information from a user equipment based on an initial network access procedure; processing the location information using a neural network to generate a fingerprint output having a set of associated beam pairs; and performing beam training using the set of beam pairs.

[0025] Some implementations of the additional exemplary method may also include notifying the user equipment of candidate beam pairs based on a set of beam pairs.

[0026] For some implementations of the additional exemplary method, candidate beam pairs may be a subset of associated beam pairs.

[0027] Additional exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, are operable to perform any of the additional exemplary methods.

[0028] Other exemplary methods according to some implementation schemes may include: adapting a location-specific beamforming fingerprint to traffic conditions at a base station; and using the beamforming fingerprint for beam training.

[0029] Another exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, are operable to perform the following operations: adapting a location-specific beam to a fingerprint at a base station to traffic conditions; and beam training the fingerprint using the beam.

[0030] Another additional exemplary method according to some implementations may include: obtaining input data including user equipment (UE) location, traffic density of the UE, and received signal strength (RSS) threshold; processing the information of the UE with a neural network having weights determined from a training phase to generate a set of one or more beam pair indices; transmitting the set of one or more beam pair indices to the UE; transmitting to the UE an instruction to perform beam training using the set of one or more beam pair indices; and receiving from the UE at least one beam pair index that satisfies the RSS threshold.

[0031] Another additional exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions operable, when executed by the processor, to perform the following operations: obtaining input data including user equipment (UE) location, traffic density of the UE, and a received signal strength (RSS) threshold; processing the UE information with a neural network having weights determined from a training phase to generate a set of one or more beampup indexes; transmitting the set of one or more beampup indexes to the UE; transmitting to the UE an instruction to perform beam training using the set of one or more beampup indexes; and receiving from the UE at least one beampup index that satisfies the RSS threshold. Attached Figure Description

[0032] Figure 1AThis is a system diagram illustrating an exemplary communication system according to some implementation schemes.

[0033] Figure 1B This illustrates the possibility of implementation according to some schemes. Figure 1A A system diagram of an exemplary wireless transmit / receive unit (WTRU) used within the communication system shown.

[0034] Figure 1C It is shown in some implementation schemes. Figure 1A A system diagram of an exemplary system of an exemplary radio access network (RAN) and an exemplary core network (CN) used within the communication system shown.

[0035] Figure 1D It is shown in some implementation schemes. Figure 1A The system diagram shows an exemplary system of another exemplary RAN and another exemplary CN used within the communication system shown.

[0036] Figure 2 This is a schematic diagram illustrating an exemplary fingerprint-based beam alignment according to some implementation schemes.

[0037] Figure 3 This is a schematic diagram illustrating an exemplary hybrid beamforming architecture according to some implementation schemes.

[0038] Figure 4 This is a schematic diagram illustrating exemplary knife-edge diffraction caused by vehicle obstacles according to some embodiments.

[0039] Figure 5A This is a schematic diagram illustrating an exemplary neural network model according to some implementation schemes.

[0040] Figure 5B This is a schematic diagram illustrating an exemplary softmax feedforward neural network model according to some implementation schemes.

[0041] Figure 6 This is a message timing diagram illustrating an exemplary process for beam alignment according to some implementation schemes.

[0042] Figure 7 This is a schematic diagram illustrating an exemplary selection of beam pairs to obtain diversity / multiplexing gain according to some implementation schemes.

[0043] Figure 8 This is a graph illustrating exemplary bit error rates for spatial diversity and spatial multiplexing auxiliary schemes according to some implementation schemes.

[0044] Figure 9 This is a schematic diagram illustrating an exemplary process for beam index modulation according to some implementation schemes.

[0045] Figure 10 This is a diagram illustrating an exemplary user distribution at four locations according to some implementation schemes.

[0046] Figure 11A This is a graph illustrating a first exemplary instantaneous RSS value for three fingerprints and three vehicle numbers according to some implementation schemes.

[0047] Figure 11B This is a graph illustrating a second exemplary instantaneous RSS value for three fingerprints and three vehicle numbers according to some implementation schemes.

[0048] Figure 12 This is an exemplary graph showing the cross-entropy versus the number of epochs of a network during training, according to some implementation schemes.

[0049] Figure 13 This is a diagram illustrating an example of RSS observed for a system using multiple fingerprints, some with learned fingerprints and some without, according to some implementation schemes.

[0050] Figure 14 This is a graph illustrating exemplary average RSS values ​​on each beam pair after adaptation in a selected fingerprint, according to some implementation schemes.

[0051] Figure 15A This is a graph illustrating an exemplary probability distribution function of fingerprints relative to traffic density according to some implementation schemes.

[0052] Figure 15B This is a graph illustrating an exemplary probability distribution function of fingerprints relative to traffic density according to some implementation schemes.

[0053] Figure 16 This is a diagram illustrating an exemplary average RSS observed at the receiver from a selected fingerprint pair according to some embodiments.

[0054] Figure 17 This is a 3D graph illustrating exemplary relationships between traffic density, SNR, and bits per channel unit according to some implementation schemes.

[0055] Figure 18 This is a flowchart illustrating an exemplary process for determining a beam pair that provides the desired RSS, according to some implementation schemes.

[0056] The entities, connections, arrangements, etc., shown in and described in conjunction with the various figures are presented by way of example rather than limitation. Therefore, any and all statements or other indications concerning what a particular figure “depicts,” what a particular element or entity in a particular figure “is” or “has,” and any and all similar states that may be interpreted in isolation and outside the context as absolute and therefore limiting, may only be appropriately interpreted as being preceded by a clause such as “In at least one embodiment, …”. For the sake of brevity and clarity, this implicit leading clause is not repeated in the detailed description.

[0057] Exemplary network for implementing the implementation scheme

[0058] In some implementations of this scheme described herein, the wireless transmit / receive unit (WTRU) can be used as, for example, a mobile phone, smartphone, mobile device, or user equipment (UE) device (which may be identified as a user).

[0059] Figure 1A This is a schematic diagram illustrating an exemplary communication system 100 that can be implemented in one or more of the disclosed embodiments. Communication system 100 can be a multiple access system providing content such as voice, data, video, messaging, and broadcasting to multiple wireless users. Communication system 100 enables multiple wireless users to access such content through the sharing of system resources (including wireless bandwidth). For example, communication system 100 may employ one or more channel access methods, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal FDMA (OFDMA), Single Carrier FDMA (SC-FDMA), Zero-Tail Unique Word DFT Extended OFDM (ZT UW DTS-s OFDM), Unique Word OFDM (UW-OFDM), Resource Block Filtered OFDM, Filter Bank Multicarrier (FBMC), etc.

[0060] like Figure 1AAs shown, the communication system 100 may include wireless transmit / receive units (WTRUs) 102a, 102b, 102c, 102d, RAN 104 / 113, CN 106, Public Switched Telephone Network (PSTN) 108, Internet 110, and other networks 112. However, it should be understood that the disclosed embodiments contemplate any number of WTRUs, base stations, networks, and / or network elements. Each of the WTRUs 102a, 102b, 102c, and 102d can be any type of device configured to operate and / or communicate in a wireless environment. As an example, WTRUs 102a, 102b, 102c, and 102d (any of which may be referred to as a “station” and / or “STA”) may be configured to transmit and / or receive wireless signals and may include user equipment (UE), mobile stations, fixed or mobile subscriber units, subscription-based units, pagers, cellular phones, personal digital assistants (PDAs), smartphones, laptops, netbooks, personal computers, wireless sensors, hotspots or Mi-Fi devices, Internet of Things (IoT) devices, watches or other wearable devices, head-mounted displays (HMDs), vehicles, drones, medical devices and applications (e.g., remote surgery), industrial devices and applications (e.g., robots and / or other wireless devices operating in industrial and / or automated processing chain environments), consumer electronics devices, devices operating on commercial and / or industrial wireless networks, etc. Any of WTRUs 102a, 102b, 102c, and 102d may be interchangeably referred to as a UE.

[0061] The communication system 100 may also include base station 114a and / or base station 114b. Each of base stations 114a and 114b may be any type of device configured to wirelessly interface with at least one of WTRUs 102a, 102b, 102c, and 102d to facilitate access to one or more communication networks, such as CN 106, Internet 110, and / or other networks 112. As an example, base stations 114a and 114b may be base transceiver stations (BTS), Node Bs, evolved Node Bs, home Node Bs, home evolved Node Bs, gNBs, NR Node Bs, site controllers, access points (APs), wireless routers, etc. Although base stations 114a and 114b are each depicted as a single element, it should be understood that base stations 114a and 114b may include any number of interconnected base stations and / or network elements.

[0062] Base station 114a may be part of RAN 104 / 113, which may also include other base stations and / or network elements (not shown), such as base station controllers (BSCs), radio network controllers (RNCs), relay nodes, etc. Base station 114a and / or base station 114b may be configured to transmit and / or receive radio signals on one or more carrier frequencies (which may be referred to as cells (not shown)). These frequencies may be in licensed spectrum, unlicensed spectrum, or a combination of licensed and unlicensed spectrum. A cell may provide coverage of radio services to a specific geographic area, which may be relatively fixed or changeable over time. A cell may be further divided into cell sectors. For example, the cell associated with base station 114a may be divided into three sectors. Thus, in one embodiment, base station 114a may include three transceivers, i.e., one transceiver for each sector of the cell. In one embodiment, base station 114a may employ multiple-input multiple-output (MIMO) technology and may utilize multiple transceivers for each sector of the cell. For example, beamforming may be used to transmit and / or receive signals in desired spatial directions.

[0063] Base stations 114a and 114b can communicate with one or more of WTRUs 102a, 102b, 102c, and 102d via air interface 116, which can be any suitable wireless communication link (e.g., radio frequency (RF), microwave, centimeter wave, micrometer wave, infrared (IR), ultraviolet (UV), visible light, etc.). Any suitable radio access technology (RAT) can be used to establish air interface 116.

[0064] More specifically, as noted above, the communication system 100 may be a multiple access system and may employ one or more channel access schemes, such as CDMA, TDMA, FDMA, OFDMA, SC-FDMA, etc. For example, base stations 114a and WTRUs 102a, 102b, and 102c in RAN 104 / 113 may implement radio technologies such as Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access (UTRA), which may use Wideband CDMA (WCDMA) to establish the air interface 116. WCDMA may include communication protocols such as High-Speed ​​Packet Access (HSPA) and / or evolved HSPA (HSPA+). HSPA may include High-Speed ​​Downlink (DL) Packet Access (HSDPA) and / or High-Speed ​​UL Packet Access (HSDPA).

[0065] In one implementation, base station 114a and WTRUs 102a, 102b, 102c can implement radio technologies such as evolved UMTS terrestrial radio access (E-UTRA), which can use Long Term Evolution (LTE) and / or Advanced LTE (LTE-A) and / or Advanced LTE Pro (LTE-A Pro) to establish air interface 116.

[0066] In one implementation, base station 114a and WTRUs 102a, 102b, 102c can implement radio technologies such as NR radio access, which can use New Radio (NR) to establish air interface 116.

[0067] In one implementation, base station 114a and WTRUs 102a, 102b, and 102c can implement multiple radio access technologies. For example, base station 114a and WTRUs 102a, 102b, and 102c can, for instance, use a dual connectivity (DC) principle to implement both LTE and NR radio access together. Therefore, the air interface used by WTRUs 102a, 102b, and 102c can be characterized by multiple types of radio access technologies and / or transmissions sent to / from multiple types of base stations (e.g., eNBs and gNBs).

[0068] In other implementations, base station 114a and WTRUs 102a, 102b, and 102c can implement radio technologies such as IEEE 802.11 (i.e., WiFi), IEEE 802.16 (i.e., WiMAX), CDMA2000, CDMA2000 1X, CDMA2000 EV-DO, Provisional Standard 2000 (IS-2000), Provisional Standard 95 (IS-95), Provisional Standard 856 (IS-856), Global System for Mobile Communications (GSM), GSM Enhanced Data Rate Evolution (EDGE), and GSM EDGE (GERAN).

[0069] Figure 1ABase station 114b can be, for example, a wireless router, a home node B, a home evolution node B, or an access point, and can utilize any suitable RAT to facilitate wireless connectivity in local areas such as commercial locations, homes, vehicles, campuses, industrial facilities, air corridors (e.g., for use by drones), roads, etc. In one embodiment, base station 114b and WTRUs 102c, 102d can implement radio technologies such as IEEE 802.11 to establish a wireless local area network (WLAN). In one embodiment, base station 114b and WTRUs 102c, 102d can implement radio technologies such as IEEE 802.15 to establish a wireless personal area network (WPAN). In yet another embodiment, base station 114b and WTRUs 102c, 102d can utilize cellular-based RATs (e.g., WCDMA, CDMA2000, GSM, LTE, LTE-A, LTE-A Pro, NR, etc.) to establish picocells or femtocells. Figure 1A As shown, base station 114b may have a direct connection to Internet 110. Therefore, base station 114b may not need to access Internet 110 via CN106.

[0070] RAN 104 / 113 can communicate with CN 106, which can be any type of network configured to provide voice, data, application, and / or Voice over Internet Protocol (VoIP) services to one or more of WTRUs 102a, 102b, 102c, and 102d. Data can have different Quality of Service (QoS) requirements, such as different throughput requirements, latency requirements, error tolerance requirements, reliability requirements, data throughput requirements, mobility requirements, etc. CN 106 can provide call control, billing services, location-based services, prepaid calling, internet connectivity, video distribution, etc., and / or perform advanced security functions such as user authentication. Although not explicitly stated... Figure 1A As shown, but it should be understood that RAN 104 / 113 and / or CN 106 can communicate directly or indirectly with other RANs that use the same RAT as RAN 104 / 113 or a different RAT. For example, in addition to being connected to RAN 104 / 113 which can utilize NR radio technology, CN 106 can also communicate with another RAN (not shown) that uses GSM, UMTS, CDMA 2000, WiMAX, E-UTRA or WiFi radio technology.

[0071] CN 106 may also act as a gateway for WTRUs 102a, 102b, 102c, and 102d to access PSTN 108, the Internet 110, and / or other networks 112. PSTN 108 may include a circuit-switched telephone network providing Common Old-Style Telephone Service (POTS). The Internet 110 may include a global system of interconnected computer networks and devices using common communication protocols such as Transmission Control Protocol (TCP), User Datagram Protocol (UDP), and / or Internet Protocol (IP) from the TCP / IP Internet Protocol suite. Network 112 may include wired and / or wireless communication networks owned and / or operated by other service providers. For example, network 112 may include another CN connected to one or more RANs, which may use the same RAT as RAN 104 / 113 or a different RAT.

[0072] Some or all of the WTRUs 102a, 102b, 102c, and 102d in the communication system 100 may include multi-mode capabilities (e.g., WTRUs 102a, 102b, 102c, and 102d may include multiple transceivers for communicating with different wireless networks via different wireless links). For example, Figure 1A The WTRU 102c shown can be configured to communicate with a base station 114a that can employ cellular-based radio technology and with a base station 114b that can employ IEEE 802 radio technology.

[0073] Figure 1B This is a system diagram illustrating an exemplary WTRU 102. (See diagram below.) Figure 1B As shown, WTRU 102 may include a processor 118, a transceiver 120, a transmitting / receiving element 122, a speaker / microphone 124, a keypad 126, a display / touchpad 128, non-removable memory 130, removable memory 132, a power supply 134, a Global Positioning System (GPS) chipset 136, and / or other peripheral devices 138, etc. It should be understood that WTRU 102 may include any sub-combination of the foregoing elements while remaining consistent with the implementation.

[0074] Processor 118 can be a general-purpose processor, a special-purpose processor, a conventional processor, a digital signal processor (DSP), multiple microprocessors, one or more microprocessors associated with a DSP core, a controller, a microcontroller, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) circuit, any other type of integrated circuit (IC), a state machine, etc. Processor 118 can perform signal encoding, data processing, power control, input / output processing, and / or any other functions that enable WTRU 102 to operate in a wireless environment. Processor 118 can be coupled to transceiver 120, which can be coupled to transmitting / receiving element 122. Although Figure 1B The processor 118 and transceiver 120 are depicted as separate components, but it should be understood that the processor 118 and transceiver 120 may be integrated together in an electronic package or chip.

[0075] Transmitting / receiving element 122 may be configured to transmit signals to or receive signals from a base station (e.g., base station 114a) via air interface 116. For example, in one embodiment, transmitting / receiving element 122 may be an antenna configured to transmit and / or receive RF signals. In one embodiment, transmitting / receiving element 122 may be a transmitter / detector configured to transmit and / or receive, for example, IR, UV, or visible light signals. In yet another embodiment, transmitting / receiving element 122 may be configured to transmit and / or receive RF and optical signals. It should be understood that transmitting / receiving element 122 may be configured to transmit and / or receive any combination of wireless signals.

[0076] Although the transmitting / receiving element 122 is in Figure 1B While depicted as a single element, WTRU 102 may include any number of transmitting / receiving elements 122. More specifically, WTRU 102 may employ MIMO technology. Therefore, in one embodiment, WTRU 102 may include two or more transmitting / receiving elements 122 (e.g., multiple antennas) for transmitting and receiving wireless signals via air interface 116.

[0077] Transceiver 120 can be configured to modulate signals transmitted by transmitting / receiving element 122 and demodulate signals received by transmitting / receiving element 122. As noted above, WTRU 102 may have multi-mode capability. Therefore, transceiver 120 may include multiple transceivers to enable WTRU 102 to communicate via various RATs (such as NR and IEEE 802.11).

[0078] The processor 118 of WTRU 102 may be coupled to a speaker / microphone 124, a keypad 126, and / or a display / touchpad 128 (e.g., a liquid crystal display (LCD) unit or an organic light-emitting diode (OLED) display unit) and may receive user input data therefrom. The processor 118 may also output user data to the speaker / microphone 124, keypad 126, and / or display / touchpad 128. Furthermore, the processor 118 may access information from any type of suitable memory (such as non-removable memory 130 and / or removable memory 132) and store data in any type of suitable memory. Non-removable memory 130 may include random access memory (RAM), read-only memory (ROM), a hard disk, or any other type of memory storage device. Removable memory 132 may include a user identity module (SIM) card, memory stick, secure digital storage (SD) card, etc. In other embodiments, the processor 118 may access information from memory that is not physically located on WTRU 102 (such as on a server or home computer (not shown)) and store data in that memory.

[0079] The processor 118 may receive power from the power supply 134 and may be configured to distribute and / or control power to other components in the WTRU 102. The power supply 134 may be any suitable device for powering the WTRU 102. For example, the power supply 134 may include one or more dry cell battery packs (e.g., nickel-cadmium (NiCd), nickel-zinc (NiZn), nickel metal hydride (NiMH), lithium-ion (Li-ion), etc.), solar cells, fuel cells, etc.

[0080] The processor 118 may also be coupled to a GPS chipset 136, which may be configured to provide location information (e.g., longitude and latitude) about the current location of the WTRU 102. In addition to or instead of the information from the GPS chipset 136, the WTRU 102 may receive location information from base stations (e.g., base stations 114a, 114b) via air interface 116 and / or determine its location based on the timing of signals received from two or more nearby base stations. It should be understood that, while remaining consistent with the implementation, the WTRU 102 may acquire location information using any suitable location determination method.

[0081] The processor 118 may also be coupled to other peripheral devices 138, which may include one or more software and / or hardware modules providing additional features, functions, and / or wired or wireless connectivity. For example, peripheral device 138 may include an accelerometer, electronic compass, satellite transceiver, digital camera (for photos and / or video), Universal Serial Bus (USB) port, vibration device, television transceiver, hands-free headset, etc. Modules, FM radio units, digital music players, media players, video game player modules, internet browsers, virtual reality and / or augmented reality (VR / AR) devices, activity trackers, etc. Peripheral devices 138 may include one or more sensors, which may be one or more of the following: gyroscopes, accelerometers, Hall effect sensors, magnetometers, orientation sensors, proximity sensors, temperature sensors, time sensors; geolocation sensors; altimeters, light sensors, touch sensors, magnetometers, barometers, gesture sensors, biometric sensors, and / or humidity sensors.

[0082] WTRU 102 may include a full-duplex radio for which the transmission and reception of some or all signals (e.g., associated with specific subframes for UL (e.g., for transmission) and downlink (e.g., for reception)) may be concurrent and / or simultaneous. The full-duplex radio may include an interference management unit for reducing and / or substantially eliminating self-interference through signal processing via hardware (e.g., a choke) or via a processor (e.g., a separate processor (not shown) or via processor 118). In one embodiment, WTRU 102 may include a full-duplex radio for which the transmission and reception of some or all signals (e.g., associated with specific subframes for UL (e.g., for transmission) and downlink (e.g., for reception)) may be concurrent and / or simultaneous.

[0083] Figure 1C This is a system diagram illustrating RAN 104 and CN 106 according to one implementation scheme. As noted above, RAN 104 can communicate with WTRUs 102a, 102b, and 102c via air interface 116 using E-UTRA radio technology. RAN 104 can also communicate with CN 106.

[0084] RAN 104 may include evolved Nodes B 160a, 160b, and 160c; however, it should be understood that RAN 104 may include any number of evolved Nodes B while remaining consistent with the implementation scheme. Evolved Nodes B 160a, 160b, and 160c may each include one or more transceivers for communicating with WTRUs 102a, 102b, and 102c via air interface 116. In one implementation, evolved Nodes B 160a, 160b, and 160c may implement MIMO technology. Therefore, evolved Node B 160a may, for example, use multiple antennas to transmit radio signals to and / or receive radio signals from WTRU 102a.

[0085] Each of the evolved nodes B 160a, 160b, and 160c can be associated with a specific cell (not shown) and can be configured to handle radio resource management decisions, handover decisions, and user scheduling in the UL and / or DL, etc. Figure 1C As shown, evolution nodes B 160a, 160b, and 160c can communicate with each other via the X2 interface.

[0086] Figure 1C The CN 106 shown may include a Mobility Management Entity (MME) 162, a Serving Gateway (SGW) 164, and a Packet Data Network (PDN) Gateway (or PGW) 166. While each of the foregoing elements is depicted as part of the CN 106, it should be understood that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0087] The MME 162 can connect to each of the evolved nodes B 162a, 162b, and 162c in RAN 104 via the S1 interface and can be used as a control node. For example, the MME 162 can be responsible for authenticating users of WTRUs 102a, 102b, and 102c, bearer activation / deactivation, selecting a specific serving gateway during the initial attachment of WTRUs 102a, 102b, and 102c, etc. The MME 162 can provide control plane functions for handover between RAN 104 and other RANs (not shown) employing other radio technologies such as GSM and / or WCDMA.

[0088] The SGW 164 can connect to each of the evolved Nodes B 160a, 160b, and 160c in RAN 104 via the S1 interface. The SGW 164 typically routes and forwards user data packets to and from WTRUs 102a, 102b, and 102c. The SGW 164 can perform other functions such as anchoring the user plane during inter-evolved Node B handovers, triggering paging when DL data is available for WTRUs 102a, 102b, and 102c, and managing and storing the context of WTRUs 102a, 102b, and 102c.

[0089] SGW 164 can be connected to PGW 166, which provides WTRU 102a, 102b, 102c with access to packet-switched networks (such as Internet 110) to facilitate communication between WTRU 102a, 102b, 102c and IP-enabled devices.

[0090] CN 106 can facilitate communication with other networks. For example, CN 106 can provide WTRUs 102a, 102b, and 102c with access to circuit-switched networks (such as PSTN 108) to facilitate communication between WTRUs 102a, 102b, and 102c and conventional landline communication equipment. For example, CN 106 may include an IP gateway (e.g., an IP Multimedia Subsystem (IMS) server) that serves as an interface between CN 106 and PSTN 108, or be able to communicate with such an IP gateway. Additionally, CN 106 can provide WTRUs 102a, 102b, and 102c with access to other networks 112, which may include other wired and / or wireless networks owned and / or operated by other service providers.

[0091] Despite WTRU in Figures 1A to 1D While described as a wireless terminal, it is conceivable that in some representative implementations, such a terminal may (e.g., temporarily or permanently) use a wired communication interface with a communication network.

[0092] In a representative implementation, the other network 112 may be a WLAN.

[0093] A WLAN in Basic Services Set (BSS) mode may have an access point (AP) for the BSS and one or more stations (STAs) associated with the AP. The AP may have access or an interface to a distribution system (DS) or another type of wired / wireless network that carries traffic to and / or carries traffic out of the BSS. Traffic originating outside the BSS and destined for a STA can reach and be delivered to the STA via the AP. Traffic originating from a STA and destined for a destination outside the BSS can be sent to the AP for delivery to the appropriate destination. Traffic between STAs within the BSS can be sent via the AP, for example, where a source STA can send traffic to the AP, and the AP can deliver the traffic to the destination STA. Traffic between STAs within the BSS can be considered and / or referred to as point-to-point traffic. Point-to-point traffic can be sent between source and destination STAs (e.g., directly between them) using Direct Link Establishment (DLS). In some representative implementations, the DLS may use 802.11e DLS or 802.11z Tunneled DLS (TDLS). WLANs using the Standalone BSS (IBSS) mode may not have an access point (AP), and STAs within the IBSS or using the IBSS (e.g., all STAs) can communicate directly with each other. The IBSS communication mode may sometimes be referred to as the "ad-hoc" communication mode in this document.

[0094] When operating in 802.11ac infrastructure mode or a similar mode, the AP can transmit beacons on a fixed channel, such as the primary channel. The primary channel can be of fixed width (e.g., a 20 MHz bandwidth) or dynamically set via signaling. The primary channel can be the operating channel of the BSS and can be used by the STA to establish a connection with the AP. In some representative implementations, Carrier Sense Multiple Access / Collision Avoidance (CSMA / CA) can be implemented, for example, in an 802.11 system. For CSMA / CA, each STA (including the AP) can listen to the primary channel. If the primary channel is listened to / detected and / or determined to be busy by a particular STA, that STA can back off. A single STA (e.g., only one station) can transmit in a given BSS at any given time.

[0095] High-throughput (HT) STAs can communicate using a 40MHz wide channel, for example, by combining a primary 20MHz channel with adjacent or non-adjacent 20MHz channels to form a 40MHz wide channel.

[0096] The Very High Throughput (VHT) STA supports channels with widths of 20MHz, 40MHz, 80MHz, and / or 160MHz. 40MHz and / or 80MHz channels can be formed by combining consecutive 20MHz channels. A 160MHz channel can be formed by combining eight consecutive 20MHz channels, or by combining two non-consecutive 80MHz channels (this can be referred to as an 80+80 configuration). For the 80+80 configuration, after channel coding, data can be split into two streams by a segment parser. Each stream can be processed individually using Inverse Fast Fourier Transform (IFFT) and time-domain processing. These streams can be mapped to two 80MHz channels, and data can be transmitted via the transmitting STA. At the receiver of the receiving STA, the operations described above for the 80+80 configuration can be reversed, and the combined data can be sent to Media Access Control (MAC).

[0097] 802.11af and 802.11ah support operating modes below 1 GHz. Compared to those used in 802.11n and 802.11ac, 802.11af and 802.11ah reduce channel operating bandwidth and carrier. 802.11af supports 5 MHz, 10 MHz, and 20 MHz bandwidths in the TV white space (TVWS) spectrum, while 802.11ah supports 1 MHz, 2 MHz, 4 MHz, 8 MHz, and 16 MHz bandwidths using non-TVWS spectrum. According to representative implementations, 802.11ah may support instrument-type control / machine-type communications, such as MTC devices in macro coverage areas. MTC devices may have certain capabilities, such as limited capabilities, including support (e.g., only support) certain bandwidths and / or limited bandwidths. MTC devices may include batteries with battery life above a threshold (e.g., to maintain a very long battery life).

[0098] WLAN systems such as 802.11n, 802.11ac, 802.11af, and 802.11ah, which support multiple channels and channel bandwidths, include a channel that can be designated as the primary channel. The primary channel can have a bandwidth equal to the maximum common operating bandwidth supported by all STAs in the BSS. The bandwidth of the primary channel can be set and / or limited by STAs operating in the BSS (each supporting a minimum bandwidth operating mode). In the 802.11ah example, for STAs supporting (e.g., only supporting) a 1MHz mode (e.g., MTC type devices), the primary channel can be 1MHz wide, even if the AP and other STAs in the BSS support 2MHz, 4MHz, 8MHz, 16MHz, and / or other channel bandwidth operating modes. Carrier Sense and / or Network Allocation Vector (NAV) settings can depend on the status of the primary channel. If the primary channel is busy, for example, because an STA (supporting only the 1MHz operating mode) is transmitting to the AP, the entire available band can be considered busy even if most of the band remains idle and potentially available.

[0099] In the United States, the available frequency bands for 802.11ah are 902MHz to 928MHz. In South Korea, the available frequency bands are 917.5MHz to 923.5MHz. In Japan, the available frequency bands are 916.5MHz to 927.5MHz. The total available bandwidth for 802.11ah ranges from 6MHz to 26MHz, depending on the country code.

[0100] Figure 1D This is a system diagram illustrating RAN 113 and CN 115 according to one implementation scheme. As noted above, RAN 113 can communicate with WTRUs 102a, 102b, and 102c via air interface 116 using NR radio technology. RAN 113 can also communicate with CN 115.

[0101] RAN 113 may include gNBs 180a, 180b, and 180c; however, it should be understood that RAN 113 may include any number of gNBs while remaining consistent with the implementation. Each of gNBs 180a, 180b, and 180c may include one or more transceivers for communication with WTRUs 102a, 102b, and 102c via air interface 116. In one implementation, gNBs 180a, 180b, and 180c may implement MIMO technology. For example, gNBs 180a and 180b may utilize beamforming to transmit signals to and / or receive signals from gNBs 180a, 180b, and 180c. Therefore, gNB 180a may, for example, use multiple antennas to transmit radio signals to and / or receive radio signals from WTRU 102a. In one implementation, gNBs 180a, 180b, and 180c may implement carrier aggregation technology. For example, gNB 180a may transmit multiple component carriers to WTRU 102a (not shown). A subset of these component carriers may be on unlicensed spectrum, while the remaining component carriers may be on licensed spectrum. In one implementation, gNBs 180a, 180b, and 180c may implement Cooperative Multipoint (CoMP) technology. For example, WTRU 102a may receive cooperative transmissions from gNBs 180a and 180b (and / or gNB 180c).

[0102] WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c using transmissions associated with scalable parameter sets. For example, OFDM symbol spacing and / or OFDM subcarrier spacing can vary depending on different transmissions, different cells, and / or different portions of the radio transmission spectrum. WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c using subframes or transmission time intervals (TTIs) of various or scalable lengths (e.g., containing different numbers of OFDM symbols and / or continuously varying absolute time lengths).

[0103] gNBs 180a, 180b, and 180c can be configured to communicate with WTRUs 102a, 102b, and 102c in standalone and / or non-standalone configurations. In standalone configuration, WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c without accessing other RANs (e.g., evolved Node Bs 160a, 160b, and 160c). In standalone configuration, WTRUs 102a, 102b, and 102c can use one or more of gNBs 180a, 180b, and 180c as mobility anchors. In standalone configuration, WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c using signals in unlicensed frequency bands. In a non-standalone configuration, WTRUs 102a, 102b, and 102c can communicate or connect to gNBs 180a, 180b, and 180c, and also communicate or connect to other RANs (such as evolved Node Bs 160a, 160b, and 160c). For example, WTRUs 102a, 102b, and 102c can implement DC principles to communicate substantially simultaneously with one or more gNBs 180a, 180b, and 180c and one or more evolved Node Bs 160a, 160b, and 160c. In a non-standalone configuration, evolved Node Bs 160a, 160b, and 160c can be used as mobility anchors for WTRUs 102a, 102b, and 102c, and gNBs 180a, 180b, and 180c can provide additional coverage and / or throughput for serving WTRUs 102a, 102b, and 102c.

[0104] Each of gNBs 180a, 180b, and 180c can be associated with a specific cell (not shown) and can be configured to handle radio resource management decisions, handover decisions, user scheduling in UL and / or DL, network slicing support, dual connectivity, interoperability between NR and E-UTRA, routing of user plane data to User Plane Functions (UPF) 184a and 184b, routing of control plane information to Access and Mobility Management Functions (AMF) 182a and 182b, etc. Figure 1D As shown, gNB 180a, 180b, and 180c can communicate with each other via the Xn interface.

[0105] Figure 1DThe CN 115 shown may include at least one AMF 182a, 182b, at least one UPF 184a, 184b, at least one Session Management Function (SMF) 183a, 183b, and possibly a Data Network (DN) 185a, 185b. Although each of the foregoing elements is depicted as part of the CN 115, it should be understood that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0106] AMF 182a and 182b can connect to one or more of gNBs 180a, 180b, and 180c via the N2 interface in RAN 113 and can be used as control nodes. For example, AMF 182a and 182b can be responsible for authenticating users of WTRU 102a, 102b, and 102c, supporting network slicing (e.g., handling different PDU sessions with different requirements), selecting specific SMF 183a and 183b, managing registration areas, terminating NAS signaling, mobility management, etc. AMF 182a and 182b can use network slicing to customize CN support for WTRU 102a, 102b, and 102c based on the type of service used by WTRU 102a, 102b, and 102c. For example, different network slices can be established for different use cases, such as services that rely on Ultra-Reliable Low Latency (URLLC) access, services that rely on Enhanced Mobile Broadband (eMBB) access, and services for Machine Type Communication (MTC) access. The AMF162 can provide control plane functions for handover between RAN 113 and other RANs (not shown) that employ other radio technologies, such as LTE, LTE-A, LTE-A Pro, and / or non-3GPP access technologies, such as WiFi.

[0107] SMFs 183a and 183b can connect to AMFs 182a and 182b in CN 115 via the N11 interface. SMFs 183a and 183b can also connect to UPFs 184a and 184b in CN 115 via the N4 interface. SMFs 183a and 183b can select and control UPFs 184a and 184b, and configure traffic routing through UPFs 184a and 184b. SMFs 183a and 183b can perform other functions, such as managing and allocating UE IP addresses, managing PDU sessions, controlling policy enforcement and QoS, and providing downlink data notifications. PDU session types can be IP-based, non-IP-based, Ethernet-based, etc.

[0108] UPF 184a and 184b can connect via the N3 interface to one or more of the gNBs 180a, 180b, and 180c in RAN 113. These gNBs can provide WTRU 102a, 102b, and 102c with access to packet-switched networks (such as Internet 110) to facilitate communication between WTRU 102a, 102b, 102c and IP-enabled devices. UPF 184 and 184b can perform other functions such as routing and forwarding packets, enforcing user plane policies, supporting multihomed PDU sessions, handling user plane QoS, buffering downlink packets, and providing mobility anchoring.

[0109] CN 115 may facilitate communication with other networks. For example, CN 115 may include an IP gateway (e.g., an IP Multimedia Subsystem (IMS) server) that serves as an interface between CN 115 and PSTN 108, or may communicate with such an IP gateway. Additionally, CN 115 may provide WTRUs 102a, 102b, and 102c with access to other networks 112, which may include other wired and / or wireless networks owned and / or operated by other service providers. In one embodiment, WTRUs 102a, 102b, and 102c may be connected to DNs 185a and 185b via UPFs 184a and 184b through their N3 interfaces and their N6 interfaces with local data networks (DNs) 185a and 185b.

[0110] Given Figures 1A to 1D as well as Figures 1A to 1D The corresponding descriptions herein refer to one or more of the functions described below, which may be performed by one or more emulation devices (not shown): WTRU102a-d, base station 114a-b, evolved Node B 160a-c, MME 162, SGW 164, PGW 166, gNB 180a-c, AMF182a-b, UPF 184a-b, SMF 183a-b, DN 185a-b, and / or any other device described herein. An emulation device may be one or more devices configured to mimic one or more of the functions described herein. For example, an emulation device may be used to test other devices and / or simulate network and / or WTRU functions.

[0111] Simulation devices can be designed to perform one or more tests on other devices in laboratory and / or carrier network environments. For example, the one or more simulation devices may perform one or more or all functions while being fully or partially implemented and / or deployed as part of a wired and / or wireless communication network to test other devices within the communication network. The one or more simulation devices may perform one or more or all functions while being temporarily implemented / deployed as part of a wired and / or wireless communication network. Simulation devices may be directly coupled to another device for testing purposes and / or may use over-the-air wireless communication to perform tests.

[0112] The one or more emulation devices may perform one or more (including all) functions without being implemented / deployed as part of a wired and / or wireless communication network. For example, the emulation devices may be used in test scenarios within a test laboratory and / or non-deployed (e.g., testing) wired and / or wireless communication networks to perform testing of one or more components. The one or more emulation devices may be test equipment. Direct RF coupling and / or wireless communication via RF circuitry (e.g., which may include one or more antennas) may be used by the emulation devices to transmit and / or receive data. Detailed Implementation

[0113] A system and method for an adaptive multi-fingerprint-based beam alignment scheme are described, which uses a deep learning neural network to adjust fingerprints to match current traffic conditions and location information. Multiple fingerprints are collected for different traffic conditions at a given location. The base station uses a deep learning feedforward neural network (such as a softmax classifier) ​​to adapt fingerprint selection to the current traffic conditions and location information. Weights can be designed and trained offline for fingerprint selection. During fingerprint selection, the base station can forward information about the selected fingerprints to the user terminal. The base station performs a training process to select beam pairs from the fingerprints that satisfy the target received signal power. If a beam pair meets a threshold, the user terminal reports the index of the beam pair from the selected fingerprints, thus significantly reducing search complexity.

[0114] A multi-functional beam transmission scheme is described to apply adaptive and intelligent multi-fingerprint-based beam alignment. Multiple beams can be selected to satisfy the target received signal power. Depending on the user terminal requirements, the base station can use additional beams to increase multiplexing and diversity gain. If the number of RF chains is less than the number of available beam pairs to satisfy the target received power, the base station can use beam indexing modulation to improve spectral efficiency.

[0115] To circumvent the necessity of a LOS path, a context-information-based beam alignment is proposed, where the beam alignment (BS) searches in a context-information-based direction. This scheme is then enhanced using sophisticated learning techniques. However, this scheme only works well with omnidirectional reception at the receiver, making it impractical for millimeter-wave communications. Furthermore, when the LOS is blocked, a set of alternative angles of arrival / departure (AoA / AoD) exists to guide the beam to an unobstructed desired location under given traffic conditions.

[0116] Extensive literature exists on fingerprint-based positioning technologies, where channel state information, or received signal strength (RSS), often referred to as a fingerprint, is typically used at predetermined locations to determine the location of a user terminal. Fingerprints are collected for different locations and stored in a database. In this context, beam pairs (AoA-AoD) are used as fingerprints to build the database for different locations. However, a limitation of many such technologies is that the fingerprints in the database are fixed for a given location. This assumption, however, generally does not hold in the case of beam alignment, as beam pairs can vary depending on traffic conditions. A fingerprint for a specific traffic density at a given location can differ for another traffic density at the same location.

[0117] Recently, machine learning-assisted wireless transmission has gained attention for its more accurate predictions and superior performance compared to conventional methods that abandon learning. More specifically, learning-based methods in localization may be more effective in minimizing localization errors.

[0118] As described in more detail below, a predetermined fingerprint for a given location may not be optimal due to environmental variations (such as traffic changes, mobility, and obstacles obstructing the RF path) and device characteristics (such as hardware defects). One scheme is described below that, for some implementations, adjusts the fingerprint selection for a given device at a specific location in real time from a set of predetermined fingerprints, given a traffic density.

[0119] For some implementations, the following process can be performed: A set of fingerprints is determined and these fingerprints are mapped to a series of parameters (which can be quantified) such as traffic density (congestion), location, device profile, angle of arrival (AoA) and angle of departure (AoD) of the RF signal between the user equipment and the base station, and RF interference. Table 1 below shows examples of multiple fingerprints for two such parameters (location and traffic density). Some implementations may use other parameters to map fingerprints. A learning process is used to select the fingerprint that is likely the best fingerprint for a given user at a given point in time. Some implementations may use such a learning process as a way to process the relationship between fingerprints and varying parameters that may affect the fingerprint selection for a given user at a given point in time. The selected fingerprints are transmitted to the user equipment apparatus, and the beam pairs in the fingerprint set are selected to achieve specific performance or complexity reduction objectives, such as a specific received signal strength (RSS) value.

[0120] System Model

[0121] Figure 2 This is a schematic diagram illustrating exemplary fingerprint-based beam alignment according to some implementation schemes. Consider a vehicle scenario where base station (BS) 202 serves vehicles (users or user equipment) 204, 210, and 216 in its cell. The number of vehicles N at any given time is... v All follow a Poisson distribution with mean λ and variance λ. Assume the serving cell is divided into N locations, where the BS is equipped with fingerprint (FP) knowledge of 208 and 214 for each location in its database, such as... Figure 2 As shown. The fingerprint used here is a beam pair, whose AoA-AoD values ​​can be predetermined offline. Fingerprints (AoA-AoD pairs) are typically obtained by using beam scanning, where the beam pairs are scanned at high resolution. However, considering the varying traffic density over time, corresponding to different numbers of vehicles at any given time point, a single predetermined fingerprint cannot provide improved performance while reducing search complexity. This is because AoA-AoD pairs providing high received signal power can be blocked / suppressed by obstacles 206, 212 (e.g., nearby vehicles). Therefore, a multi-fingerprint-based scheme is envisioned for different traffic densities at a given location. Equation 1 illustrates an exemplary multi-fingerprint database:

[0122]

[0123] Figure 2 The system model is shown, where the BS uses the fingerprint FPλ1(l j ) represents traffic density λ k The user at location l1 is provided with services, where entry f i (k，j)This represents the i-th fingerprint at the j-th location of traffic density k. Location information can be obtained from the Global Positioning System (GPS). Furthermore, due to the hardware complexity and power-intensive ADC / DAC at millimeter-wave frequencies, dedicating a dedicated RF chain to each antenna element may be impractical; and using only analog beamforming will result in poor angular resolution and inaccuracies.

[0124] Figure 3 This is a schematic diagram illustrating an exemplary hybrid beamforming architecture according to some implementation schemes. Therefore, hybrid beamforming can be used as a combination of digital and analog beamforming. In hybrid beamforming designs, the signal is digitally precoded in baseband and phase-shifted via analog phase shifters at the RF stage. Many hybrid beamforming architectures tend to rely on fully connected or subarray connected designs. In fully connected designs, each RF chain tends to be connected to all phase shifters of the antenna array, such as... Figure 3 As shown. In contrast, in a subarray interconnect design, the antenna array is divided into subarrays, where each RF chain is connected to only a subset of the phase shifters. For learning-aided schemes, the BS and user can achieve beamforming gain using either a fully connected or subarray interconnect design. However, for example, both the BS and user use... Figure 3 The fully connected architecture shown uses matrix F in the baseband for signal 302. BB 304 performs digital precoding, and then uses an analog RF beamforming matrix F RF Phase shifting is performed.

[0125] The BS (transmitter) is equipped with N t Transmitting antenna and N t RF The chain, while the user (receiver) is equipped with N r Receiving antenna and N r RF Chain. Furthermore, assume {(f 1 RF ,w 1 RF ),(f 2 RF ,w 2 RF ),…,(f N RF ,w N RF The beamforming vectors at the BS and user ends are selected respectively for the traffic density λ. The received signal vector y322 at the user end is given by Equation 2:

[0126]

[0127] Where F RF308 is the size of N at BS. t ×N t RF The transmit beamforming matrix, where N t RF Column 306 consists of the N of the transmitting antenna. t The possible AoD set of line 310 {(f 1 RF …f N RF} constitutes. Similarly, W RF 316 represents the size N at the user end. t ×N r RF The receiving beamforming matrix, where N r RF Column 318 consists of the N of the receiving antenna. r The possible set of AoA in line 314 {(w 1 RF …w N RF Composed of ) matrix W BB 320 is the receiver baseband weight matrix used in Equation 2. Furthermore, s is the transmitted symbol, and n is the number of symbols with distributions. The noise vectors of identical and independent distributed entries, where H is a vector of size N. r ×N t The statistical millimeter-wave channel model 312 is shown in Equation 3:

[0128]

[0129] and It is a complex-valued Gaussian random variable with Rayleigh and uniform distributions for amplitude and phase, respectively. For a variable with N... r and N t Uniform linear array (ULA) of antenna elements, response vector α r and α t As shown in equations 4 and 5:

[0130]

[0131]

[0132] Where Φ r and Φ t These are the arrival angle and departure angle, respectively.

[0133] Figure 4 This is a schematic diagram illustrating exemplary knife-edge diffraction caused by vehicle obstacles according to some embodiments. Received signal strength (RSS) can be used as a performance metric to determine the beam pair used for successful transmission.

[0134] Furthermore, the construction of the fingerprint database considers not only the path loss experienced by the millimeter-wave carrier but also the attenuation and congestion caused by adjacent vehicles. To achieve this, a multi-blade model can be used. For a given location, each fingerprint is constructed for each traffic density by considering the total attenuation caused by vehicles. The attenuation caused by a single blade for each vehicle is given by Equation 6:

[0135]

[0136] in The height 410 of the line connecting obstacle 402 and user 406 is the distance between obstacle 402 and the line connecting BS and user 406, and r f The Fresnel ellipsoid radius is given in Equation 7:

[0137]

[0138] in It is the wavelength, d 用户 408 is the distance between the transmitter and the user, while d 障碍物 404 is the distance between the launcher and the obstacle, such as... Figure 4 As shown.

[0139] For G t and G r The transmit and receive antenna gains, and the path loss experienced by the signal at a distance d are given by Equation 8:

[0140]

[0141] Where d0 is the near-range reference distance, S σ0 It is the shadowing factor, while PL0 is the free-space path loss. Furthermore, for a given wavelength... Free space path loss is expressed as Equation 9:

[0142]

[0143] Therefore, the total received power, taking into account path loss and attenuation caused by vehicles, is given by Equation 10:

[0144] P r =P t -PL [dB] -A Equation 10

[0145] Equation 11 is due to large-scale fading. However, considering small-scale fading as well as beamforming and combining effects, the net received power... It is represented as shown in Equation 11:

[0146]

[0147] The capacity is given by equation 12:

[0148]

[0149] in This is the noise variance. The net received power observed for the fingerprint structure is... Initially, beam scanning operations are performed at N locations with discrete traffic density, and then specific beam pairs achieving the target RSS are stored in a database. In other words, the AoA-AoD fingerprint is obtained through a high-resolution beam search activity by considering all congestion caused by obstacles. Fingerprint construction is typically performed offline via computer-generated environment simulation or in real time during BS installation.

[0150] Problems addressed in some implementation schemes

[0151] Achieving precise beam alignment in directional transmission systems is challenging, especially at millimeter-wave frequencies due to their high sensitivity to congestion. To circumvent this problem, fingerprint-based beam alignment techniques can be used. More broadly, accurate fingerprints can be considered auxiliary information that can be used to enhance system performance.

[0152] A fingerprint can comprise a set of selected beam pairs for a given location, on which communication links can be established. Typically, during network deployment, a fingerprint is constructed for each location by considering the surrounding environment, such as buildings, lampposts, and vehicles. Fingerprints of beam pairs providing beams with specific angle-of-arrival (AoA-AoD) values ​​(e.g., a set of possible beam pairs) are usually obtained by using beam scanning, enabling a high-resolution scan of beam pairs to be initially performed for N locations with discrete traffic densities. Specific beam pairs for achieving the target RSS are stored in a database. The AoA-AoD fingerprint can be obtained by performing a high-resolution beam search activity, taking into account all congestion caused by obstacles. Fingerprint construction is typically performed offline via computer-generated environment simulation or in real-time during network deployment.

[0153] Traffic variations among users and vehicles can severely limit fingerprint performance. This limitation stems from the fact that the beam orientation, or the number of available beam pairs, is highly dependent on the density and location of users and vehicles on the road. The beam orientation, or the number of available beam pairs, depends on traffic conditions that change over time. For example, traffic density in the morning differs from that in the afternoon or during special events. To respond to these environmental variations, multiple fingerprints can be used for a given location, and fingerprint selection can be dynamically adjusted to match traffic conditions at a given location at a given time. This paper addresses these challenges.

[0154] In some implementations, a multi-fingerprint-based beam alignment scheme is provided, where the base station (BS) adapts its fingerprints to traffic conditions at a given location with the help of learning. This multi-fingerprint-based beam alignment scheme can be used to design multi-functional beam transmissions, where multiple beam pairs satisfying an RSS threshold are selected to achieve higher multiplexing and diversity gain.

[0155] Learning-assisted multi-fingerprint-based beam alignment

[0156] Fingerprints can be built offline based on experience during the network deployment phase. Fingerprint construction depends on the static topology of road structures and buildings that will not change in the short term, but the topology can change if new buildings are constructed. Environmental changes can be attributed to constantly changing mobile user traffic. In the absence of traffic, fingerprints built during the network design phase can be applied indefinitely unless the regional topology changes. However, because mobile user traffic conditions change over time, the number of beam pairs in the option set can vary. Therefore, fingerprints can be built for different traffic densities by allowing user devices (or users) to identify possible beam pairs through beam scanning during the beam pair training phase. This process can be repeated for different densities or traffic conditions. Furthermore, the required number of fingerprints is location-specific and can be determined empirically.

[0157] Table 1 shows an exemplary database based on multiple fingerprints, such that BP2 represents a valid departure angle, e.g., AoD (which can be equal to 30°), and its corresponding arrival angle pair can be AoA (which can be equal to 60°), and can be indexed as beam pair 2 (AoD-AoA). Similarly, BP 10 This indicates a beam pair of 10, whose departure and arrival angles can be 110° and 270°, respectively. In summary, BP... X This represents any beam pair paired with the corresponding angle of arrival at any departure angle, denoted by index X. A database can be built by calculating the net received power at each location under different traffic conditions.

[0158] Table 1. Exemplary Multi-Fingerprint Databases

[0159]

[0160] This relationship can be represented by a lookup table for link adaptation, allowing the selection of the corresponding fingerprint based on traffic density and location. For example, if a user is at location L1 and the base station (BS) estimates the traffic density as λ2, then the BS can select a beam {BP}. 320 BP 210The fingerprint of BP3, ... is obtained. The BS shares this information with the user equipment, so the BS and the user equipment invoke beam search to identify the best beam pair among the available beam pairs for the selected fingerprint. This process significantly reduces the search space involved in beam alignment. To further reduce search complexity, an RSS threshold can be set so that the user equipment selects a specific beam pair whose observed RSS value is higher than the threshold. After selecting a beam pair, the user equipment forwards this information to the BS, thereby eliminating the need to search for consecutive beam pairs.

[0161] While the aforementioned multi-fingerprint adaptive scheme enhances performance compared to single-fingerprint-based beam alignment, the performance gain can become limited if thresholds (such as the RSS observed in the fingerprint at a given location L of the traffic density λ in the lookup table) become outdated. Therefore, achieving perfect beam alignment can be challenging. To address this challenge, learning-assisted multi-fingerprint-based beam alignment can be used. Using a learning-assisted scheme eliminates the dependence on traffic condition thresholds for fingerprint selection (listed in Table 1). This is because these values ​​can become outdated due to defects and impairments in the channel, such as problems in the ADC / DAC.

[0162] Neural networks can be used for intelligent adaptation among multiple fingerprints. Using neural networks can reduce complexity and provide superior performance. The learning-assisted scheme can have two phases: (i) a training phase and (ii) a testing phase. In the training phase, the weight vector of the network is computed using training samples, such that the inputs and outputs are known. This process is classified as a supervised learning technique, allowing for the use of supervised design to train the weights. The training weights can be computed offline, and these computations do not impose real-time overhead on the system.

[0163] Figure 5A This is a schematic diagram illustrating an exemplary neural network model according to some implementation schemes. Figure 5B This is a schematic diagram illustrating an exemplary softmax feedforward neural network model according to some implementation schemes. If the number of hidden layers 512, 562 is greater than 1, the neural network is called a deep neural network, such as... Figure 5A and Figure 5B As shown, each graph has 2 hidden layers. Figure 5A and Figure 5B As shown, training samples are passed to the input layer, and the weights are designed to minimize the error, which is the difference between the true output value and the predicted output value. Figure 5A and Figure 5BIn the matrix, W1 508, 558, W2 516, 566 and W3 522, 572 are weights, and b1 510, 560, b2 518, 568 and b3 524, 574 are the deviations from the input layers 502, 552 to the hidden layers 512, 562 to the output layer 526, respectively, such that W(p,q) represents the weights attached between nodes p and q. Matrix x i 504 and 554 are the network inputs. Matrix u i and v i is the neuron input matrix at the network node, where i represents the class. For Figure 5A y1 528 and y2 530 are the outputs of the neural network. For Figure 5B net1 576 and net2 578 are the outputs of the neural network, while y1 582 and y2 584 are the output probabilities of the softmax function 580.

[0164] In the learning-assisted scheme, each fingerprint corresponds to a category. Furthermore, the output of each hidden layer is determined by activation functions f() (506, 556, 514, 564, 520, 570) (also known as fractional functions), which determine the system's performance. The choice of activation function can depend on the tractability of the analysis, computational complexity, and the type of output signal. The loss function characterizes the error (loss) between the predicted result and the true result in the training samples.

[0165] In learning-assisted schemes, the selection of fingerprints (or categories) can be determined by the output probabilities associated with each fingerprint. Fingerprints with high probabilities can be selected. Because this scheme uses probabilities at the output, fingerprint selection can be interpreted as a logistic regression with multiple categories, such that each fingerprint constitutes a category.

[0166] Figure 5B The softmax function, used for the activation (or fractional) function to generate output probabilities, is shown. Sometimes, the resulting neural network is called a softmax neural network.

[0167] Figure 5A Each layer is assigned a score by an activation function. However, due to the probability of processing a specific outcome, the linear weights are converted into probabilities at the output by a softmax function, such as... Figure 5B As shown, and expressed as Equation 13:

[0168]

[0169] Where z is the fractional vector of the layer before calculating the output probability, given by Equation 14:

[0170] z = f(x) iW, b) = Wx i +b Equation 14

[0171] In Equation 14, W is the layer weight matrix, b is the bias vector, and x i It is the input of the layer, such as Figure 5B As shown. Consider, for example... Figure 5B In the output layer, the softmax function 580 is applied to scores net1 576 and net2 578, calculated using a weight matrix W3 572 and a bias vector b3 574. It's important to note that the softmax function 580 is only used in the output layer to generate the output probability 586; for all other layers, only the scores obtained using Equation 14 are calculated, such as... Figure 5B As shown.

[0172] In one implementation, the input vector x i It is a 3D vector storing location, traffic density, and RSS value, while the output represents the probability associated with each fingerprint. In other words, the output uses [0 ... 0 1 0 ... 0]. T The form is where 1 is the probability associated with that specific fingerprint. Location parameters can be specified based on latitude and longitude values ​​or angle and distance relative to the base station.

[0173] Initially, the weight matrix is ​​randomly selected from a distribution N(0,1), so the predictions at the output will be incorrect. Therefore, to improve fingerprint predictions given traffic density and location, a loss function is introduced, which is a measure of the difference between the predicted probability and the true probability associated with a given class. In other words, by taking the loss function into account, the weight matrix is ​​optimized to ensure that the loss is minimized. For some implementations, the weight matrix and bias vector are updated based on error backpropagation to reduce the loss function.

[0174] More specifically, for some implementations, the goal is to minimize or at least reduce the difference between the true probability distribution and the predicted probability distribution. This loss function can also be interpreted as the Kullback-Leibler divergence between the two distributions. Therefore, for distributions p and q, it is expressed as Equation 15:

[0175] D KL (p||q)=-∑ s p(i)log 10 (q(i)) Equation 15

[0176] In the example, p(i) is the probability of correctly classing i, i.e., p(i) = [0···1···0], and q is the function in Equation 13. Substituting Equation 13 into Equation 15, we get Equation 16:

[0177]

[0178] Where S is the number of training samples.

[0179] In addition, it has The cross-entropy, where the cross-entropy is maintained Because there is no uncertainty in the correct category, this loss function can also be called cross-entropy loss.

[0180] After defining the cross-entropy loss 3, the total loss function for all classes associated with the regularization penalty R(W) is shown in Equation 17:

[0181]

[0182] Among them have And S is the total number of training samples. The basic principle of adding a regularization term in Equation 17 is to ensure that it does not lead to overfitting.

[0183] The current aim is to compute the gradient with respect to the weight matrix W3 and Figure 5B The deviation b3 is used to minimize Equation 17. To achieve this, the gradient of each class is calculated, and its weights are now W. i 3 and biased to b i 3, where i represents the category. Note that z in Equation 17... i It is a function of the weight matrix W3 and the bias vector b3. After a series of steps, regarding W... i and b i The gradient is

[0184]

[0185]

[0186] Therefore, gradient descent is used to generate

[0187]

[0188]

[0189] Where α is the step size. Similarly, the weight matrices W1 and W2, and the bias vectors b1 and b2, are obtained by using the gradient of the loss function in Equation 18 with the corresponding matrices W and vectors b. This process is called error backpropagation. Example pseudocode for the learning-aided fingerprint-based algorithm is shown in Table 2. The network's weight matrix W and bias vector b are computed offline and stored in memory. This entire process is performed during the training phase.

[0190] Table 2 shows the pseudocode for the fingerprint-based learning-assisted method. The network's weight matrix W and bias vector b can be computed offline and stored in memory. This entire process can be performed during the training phase.

[0191] Table 2. Pseudocode for fingerprint-based methods used for learning assistance

[0192] Offline weight training process:

[0193] Input: Training samples for each training location

[0194] Output: Fingerprint

[0195] Initialize matrices W1, W2, and W3 with random values.

[0196] • Repeat until equation 18 converges:

[0197] Use Equation 14 to compute the output of f() for each layer with specified weights.

[0198] Use Equation 13 to apply the softmax function to obtain the probability of each class.

[0199] The weight matrix is ​​obtained through Equation 20, and the bias vector is obtained through Equation 21.

[0200] ο Backpropagation of execution errors

[0201] Online fingerprint calculation process:

[0202] • Enter the location, number of vehicles, and target RSS.

[0203] • Apply training weights determined by the offline training weight process.

[0204] • Output fingerprint

[0205] Figure 6 This is a message timing diagram illustrating an exemplary process for beam alignment according to some implementation schemes. Figure 6An example of the aforementioned real-time testing phase process is shown. During initial access, the BS (e.g., a gNB for 5G NR) 602 communicates with the user (e.g., using a lower frequency), where the BS 602 estimates the location of the user (or user equipment). For some implementations, the user equipment 604 may use the initial access to send location information 606. The BS 602 also has information 608 indicating the number of users (vehicles) in its cell, which can be used to estimate vehicle / traffic density (λ). Furthermore, the user equipment 604 forwards its RSS threshold requirement information to the BS 602. Using the number of users (vehicles), the RSS threshold, and location as input parameters, the BS 602 uses a softmax method to compute weights and biases during the training phase. This method is used to predict the probability of each category (or each fingerprint), and the BS 602 selects the fingerprint with the highest probability. When selecting a fingerprint, the BS 602 transmits a set of beam pair options 610 to the user (UE) 604. The user equipment 604 may send an acknowledgment 612. BS 602 can perform a 614-beam search on the selected beam pairs in the set. In some embodiments, the beam search is performed by the transmitter selecting a given beam (the transmit beam from the beam pair associated with the fingerprint) and transmitting using that beam. The receiver performs measurements of the signal received on that beam (using the corresponding receive beam from the given beam pair associated with the fingerprint) to see if the beam pair meets the desired quality. This process can be repeated for one or more beam pairs in the set. Repetition can be limited if the receiver sends feedback to the transmitter indicating that an acceptable beam pair has been identified, which can reduce the size of the search space. At the end of the beam search cycle, one or more beams (or one or more beam pair indices) can be selected to transmit data to the receiver. In some embodiments, the selection of one or more beams can be accomplished using a multi-functional beam transmission scheme. In some embodiments, the initial beam search process of identifying beam pairs associated with the fingerprint can be used for coarse beam tuning, and the beam search during final beam selection (and subsequent operations) can include further beam fine-tuning using baseband processing and filtering for further refinement. For some implementations, fine beam searches can be repeated until a given resolution level is reached. For some implementations, beam searches can be repeated periodically. For some implementations, beam searches can be repeated if a triggering event occurs, such as a change in user equipment parameters (e.g., a change in input parameters due to UE movement or traffic density changes) or a change in another parameter (which may not be part of the ANN's input parameters), such as a change in the priority of serving a given UE (receiver) with one or more selected beam sets. User 604 then returns an index of specific beam pairs that meet its post-processing RSS threshold to 618. Figure 6 The diagram shows a message sequence illustrating the method.

[0206] Multifunctional beam transmission

[0207] Figure 7 This is a schematic diagram illustrating an exemplary selection of beam pairs to obtain diversity / multiplexing gain according to some implementation schemes. Using the above scheme, multi-functional beam transmission can allow for some tolerance in terms of beam search complexity. Multi-functional beam transmission can be used to improve spectral efficiency and enhance performance. Such methods can use the assumption that multiple beam pairs exist that satisfy an RSS threshold. As previously described, BS 702 uses the selected fingerprint for beam search, causing the user equipment to select a specific beam pair that satisfies the target RSS and return the index of that specific beam pair to the BS.

[0208] In contrast, in multi-functional beam transmission, user 708 selects several beam pairs 704 and 706 that satisfy the RSS threshold at the cost of increased search complexity, such as... Figure 7 As shown. Although beam searching for consecutive beam pairs is added, there may be cases where no additional beam pairs exist. These additional beam pairs can be used to implement diversity and / or multiplexing gain. The number of beams that can be used may be limited by the number of RF chains.

[0209] Figure 8 This is a graph illustrating exemplary bit error rates for spatial diversity and spatial multiplexing auxiliary schemes according to some implementation schemes. Figure 8 The diagram shows the average BER 802 versus the average SNR 804 for different transmission schemes 806, 808, 810, and 812 communicating via a millimeter-wave channel using a 64×32 element MIMO scheme with two RF chains at the BS and receiver. Figure 8 In the diagram, two spatial streams are used for spatial multiplexing, while only a single spatial stream is used for diversity. The diagram shows that diversity-oriented schemes 806 and 808 perform better than spatial multiplexing schemes 810 and 812, which aim to achieve higher data rates.

[0210] Given the observed qualified beam pairs, the BS-user pair can use link adaptation based on the characteristics of the channel in each beam. Based on the post-processed SNR observed by the user equipment, the BS can use diversity or multiplexing. If the channel is in deep fading, the user can choose diversity and can choose other multiplexing methods. Furthermore, the BS can combine multiplexing and diversity-assisted transmission to adjust (or optimize for some implementations) the power allocation of each beam. After regular link adaptation, 10 is achieved. -3 The specific SNR threshold of the target BER is in Figure 8 As shown, each scheme has a vertical line.

[0211] After calculating the instantaneous post-processed SNR, the receiver determines the transmission scheme type and modulation mode by comparing the instantaneous post-processed SNR with a predefined threshold. For example, the receiver can compare the post-processed SNR with... Figure 8 The vertical lines are compared, and the requested pattern information is forwarded to the BS. The post-processed SNR value can be calculated offline and stored in memory, so the receiver does not have to perform such calculations in real time.

[0212] Figure 9 This is a schematic diagram illustrating an exemplary process for beam index modulation according to some embodiments. If the number of beam pairs 910 that meet the target RSS observed by the user is greater than the number of RF chains 902, the BS can use beam index modulation on the beam pairs reported by the user. Beam index modulation enables communication of additional implicit information by inferring which specific beam is activated at the receiver. Figure 9 An exemplary phase-shifting element (and / or phase / gain circuitry) 904, a combiner circuit / element 906, and an antenna transmitter 908 configured to perform hybrid beamforming in one embodiment are shown.

[0213] Figure 9 This illustrates a typical beam indexing modulation used by BS to increase the data rate. The beam pairs (N) satisfy the target RSS. b The number of 910 chains is higher than that of RF chains (N). RF The number of 902. This is similar to spatial modulation, where the antenna index carries this type of information. Compared to the antenna index in spatial modulation, the learning-assisted scheme can use the beam index to deliver this type of information. Therefore, the total number of bits that can be transmitted per user per channel per second is shown in Equation 22:

[0214] Number of bits = log2(M) + log2(N) b Equation 22 states that, in addition to selecting beam pairs for transmission based on the input signal stream during beam indexing modulation, the selected beam pairs can also obtain diversity or multiplexing gain depending on the nature of the channel in these beam pairs.

[0215] Simulation results characterizing the performance are shown in several figures. The performance of the multi-fingerprint-assisted beam alignment scheme, which relies on learning and benchmarks, is characterized. The number of vehicles at any time point follows a Poisson distribution ~Poisson(λ), with both mean and variance λ. Furthermore, vehicle-induced congestion is random, following the distribution u(0,N). v ), where N v Let be the number of vehicles following a Poisson distribution ~Poisson(λ). Assume the maximum congestion number equals the number of vehicles. Furthermore, from a database of vehicle dimensions, vehicle heights follow a normal distribution with an average value of μ. h The standard deviation is σh Table 3 shows the parameters used in the simulation. The neural network uses two hidden layers, each with 20 nodes, while the input and output layers each have 3 nodes. Furthermore, the activation function chosen for each hidden layer is the Tan-Sigmoid function, while the softmax function is used in the output layer. Multiple fingerprints were used for three different traffic densities.

[0216] Table 3. Simulation Parameters

[0217] parameter value <![CDATA[P t ]]> 20dBm <![CDATA[N t ]]> 32 <![CDATA[N r ]]> 8 <![CDATA[G t ]]> 10dBi <![CDATA[G r ]]> 5dBi Λ(28GHz) 0.0107 Number of vehicles (Nv) Poisson(λ) Vehicle height (meters) N(150，8.6) block <![CDATA[rand(0,N v )]]> <![CDATA[d 用户 ,d 障碍物 ]]> rand()

[0218] Figure 10 This is a diagram illustrating an exemplary user distribution at four locations according to some implementation schemes. Figure 10 An example is shown where BS 1012 is located at the center of the cell. Vehicles at location 1 (1006), location 2 (1008), and location 3 (1010) are shown as normalized horizontal distances of 1002 and vertical distances of 1004 from base station (BS) 1012. Furthermore, the cell is divided into four locations, each with its own fingerprint for different traffic densities. The number of vehicles at each location follows a Poisson distribution, while the distances from vehicles to the base station are uniformly distributed within each cell. Figure 10 It was observed that BS was providing service to two users, 1014 and 1016, in locations 2 and 3, while vehicles near the users were considered obstacles.

[0219] Figure 11A This is a graph illustrating a first exemplary instantaneous RSS value for three fingerprints and three vehicle numbers according to some implementation schemes. Figure 11A The relationship between the instantaneous RSS values ​​1102 of the three fingerprints 1106, 1108, and 1110 and the traffic density (λ) 1104 is shown when the number of vehicles follows a Poisson distribution. Figure 10 As shown, the average values ​​are 5, 20, and 45. Figure 11A The RSS values ​​of the scene are depicted, where the height difference between the obstacles and the line connecting the transmitter and receiver makes v in Equation 6 less than -0.7, and the vehicle attenuation is zero. Signal attenuation is due to path loss. However, in Figure 11A The fluctuations observed are due to fading introduced by the channel, as shown in Equation 11.

[0220] Figure 11B This is a graph illustrating a second exemplary instantaneous RSS value for three fingerprints and three vehicle numbers according to some implementation schemes. Figure 11BThe RSS values ​​of three fingerprints (1156, 1158, 1160) and traffic density (λ) (1154) are plotted, where the height difference from the obstacle to the line connecting the transmitter and receiver makes v in Equation 6 greater than -0.7. In other words, the signal is attenuated due to vehicle and path loss, which is clearly visible in the figure, where points scattered around have RSS values ​​as low as -125 dB. The physical meaning is that some of the available beam pairs in the fingerprint are blocked due to vehicle obstacles. Without these blockages, Figure 11B Showing and Figure 11A Similar behaviors, among which Figure 11A and Figure 11B All are drawn for non-adaptive systems.

[0221] Figure 12 This is a diagram illustrating an exemplary graph of the cross-entropy versus epochs of a network during training, according to some implementation schemes. To analyze the loss function of the softmax neural network presented in Equation 17, Figure 12 The diagram illustrates the relationship between the network's cross-entropy 1202 and the number of epochs 1204 during training. A network is said to have reached "1 epoch" when each data sample is used to design its parameters once. An epoch is approximately equal to the number of times the dataset is used to design parameters. Figure 12 The results show that the network reaches its best performance after approximately 60 epochs. Furthermore, the validation error trajectory 1208 is only slightly higher than the training error trajectory 1206, which means that the designed neural network weights do indeed provide a good fit in the mapping between input and output samples. Figure 12 It can be used to study the design of neural network parameters. In other words, if Figure 12 If the validation error is high while the training error is low, it means the network is overfitting, so the regularization parameter can be adjusted; on the other hand, if both the validation error and the training error are high, it means underfitting, so the number of neurons (nodes) can be adjusted.

[0222] Figure 13 This is a diagram illustrating an example of RSS observed for a system using multiple fingerprints, some with learned fingerprints and some without, according to some implementation schemes. Figure 13 The relationship between RSS 1320 and traffic density (λ) 1304 is shown for a system using multiple fingerprints with and without learning 1306, for a single fingerprint 1310, and for line-of-sight (LOS) 1312 propagation. In LOS-based beam alignment, the beam can be directed to the user's direction obtained from the location, regardless of congestion.

[0223] In this simulation, the target RSS is set to -82dBm. From Figure 13It can be seen that beam alignment based on multiple fingerprints offers superior performance, while the performance of beam alignment and line-of-sight (LOS) based on a single fingerprint deteriorates sharply with increasing traffic density. This is because the probability of LOS congestion increases with traffic density, leading to a low RSS. On the other hand, beam pairs based on a single fingerprint, designed for a given traffic density and using the same fingerprint for other traffic densities, experience congestion because a beam pair suitable for one setting is not suitable for another. In the simulation, a fingerprint with a traffic density λ of 5 was used. Furthermore, the expected AoA-AoD pair suitable for this particular setting may not exist in the fingerprint constructed for another setting. Therefore, performance is significantly affected. In contrast, the design based on multiple fingerprints offers better performance; however, it can be seen that adaptation based on multiple fingerprints without learning cannot consistently maintain the target RSS. Instead, it hovers around the target RSS due to constantly changing channel statistics caused by environmental variations. Therefore, employing learning in the design based on multiple fingerprints will intelligently adapt to the time-varying environment, ensuring that it always meets the target RSS, such as... Figure 13 As shown.

[0224] Figure 14 This is a graph illustrating exemplary average RSS values ​​on each beam pair after adaptation in a selected fingerprint, according to some implementation schemes. Figure 14 The relationship between the average RSS value of all beam pairs in the selected fingerprint after adaptation 1402 and the traffic density (λ) 1404 is shown. More specifically, Figure 14 It is a measure of the total average received power observed within a given fingerprint. If the transmit power is evenly distributed among the beam pairs in the fingerprint, it is the power observed after post-processing. The principle behind choosing this metric is to study the effect of the average RSS power on all beam pairs when the transmit power in each beam is constant. Figure 14 The illustration shows that fingerprint selection using learning 1408 is performed similarly to perfect beam alignment 1406, where an exhaustive beam scan is performed, and fingerprint adaptation without learning 1410 is significantly different from perfect beam alignment 1406. The physical implication is that the receiver is able to capture signals from all directions predicted by the learning algorithm. In other words, the learning accurately predicts fingerprints consisting of the maximum possible number of beam pairs to achieve a successful link connection, which is also observed in the average RSS. On the other hand, fingerprint selection assignment with learning is significantly affected by poor fingerprint selection, i.e., selecting fingerprints associated with incorrect beam pairs, such as... Figure 14 As shown. More specifically, the selected fingerprint does not contain beam pairs used for successful transmission, meaning that the transmission power allocated to these beam pairs experiences either jamming or deep fading, resulting in a low average RSS value at the user. It is important to emphasize that the design achieves perfect beam alignment performance while significantly reducing search complexity.

[0225] Figure 15A This is a graph illustrating an exemplary probability distribution function of fingerprints relative to traffic density according to some implementation schemes. Figure 15B This is a graph illustrating an exemplary probability distribution function of fingerprints relative to traffic density according to some implementation schemes. Figure 15A and Figure 15B The relationship between the probability distribution functions (PDFs) of fingerprints (1502, 1552) and traffic densities (1504, 1554) is shown. Figure 15A In the process, as the number of vehicles or traffic density increases, the PDF of fingerprint 1 (1506) begins to gradually decrease, while that of fingerprint 2 (1508) increases monotonically. Similarly, as traffic density increases further, the PDF of fingerprint 2 (1508) decreases, and the PDF of fingerprint 3 (1510) begins to increase. This means that the set of beam pairs in the fingerprints that provide successful alignment begins to decrease because increased traffic density leads to increased congestion. Therefore, fingerprints with beam pairs suitable for this environment are selected for successful transmission. For example, when considering a traffic density range of 5 to 20, the PDF of fingerprint 1 (1506) decreases because it selects fingerprint 2 (1508) whenever the beam set within it is blocked due to increased vehicle density, thanks to the learning model developed during the training phase, as it provides an alternative beam pair for link connection. Figure 15B The PDF shows fingerprints 1556, 1158, and 1560 when learning was abandoned. Figure 15B In the process of switching from one fingerprint to another... Figure 15A The difference observed is because the handover is determined solely by a lookup table, which may be outdated due to the time-varying nature of the channel.

[0226] Figure 16 This is a graph illustrating an exemplary average RSS observed at the receiver from a selected fingerprint pair according to some embodiments. Figure 16 and Figure 14 The difference is that, Figure 14 It is the average RSS observed across all beam pairs of the selected fingerprint. Figure 16 Two v values ​​(for different vehicle heights) were plotted for RSS 1602 and traffic density 1604 with and without learning, resulting in four trajectories 1606, 1608, 1610, and 1612. In this study, the target RSS was set to -82 dB. For both v cases, the learning-assisted design performed approximately 1 dB higher than the target RSS. In contrast, the lookup table-dependent design was significantly affected, especially in the traffic density regions between 10 and 15 and 30 and 40. This performance is similar to that of a conventional link adaptation design.

[0227] Figure 17 This is a 3D graph illustrating exemplary relationships between traffic density, SNR, and bits per channel unit according to some implementation schemes. Figure 17 The 3D trajectories for bit rate 1702, SNR 1704, and traffic density 1706 are shown with and without learning 1708. Figure 17 As observed, the rate (bpcu) 1702 without learning 1710 is lower than that with learning-assisted beam alignment 1708, while the design with learning 1708 provides a maximum rate of ~2 bpcu, independent of vehicle density 1706. This is because the design without learning 1710 relies on a lookup table threshold that experiences congestion in traffic density areas between 10 and 15 and 30 and 40.

[0228] Complexity

[0229] This section will discuss the complexity of the presented design. The conceptually simple beam scanning technique exhibits considerable complexity during beam searching. Consider the complexity at angle Θ. d Signals ∈ (0, 360°) leaving the transmitter are received at angle Θ. d The signal is received within the range of (0, 360°). Assume the half-power beamwidth (HPBW) of the signal ray is β. Then, beam alignment based on beam scanning must be performed... Beam pair combinations perform an exhaustive search. On the other hand, fingerprint-based beam alignment is much less complex because the number of beam pairs is significantly reduced. It should be noted that the complexity of the proposed design mainly occurs during the offline learning phase. In other words, Figure 5B The weights are designed before any communication between the BS and the user, and then stored in memory, with a storage complexity of approximately [value missing] between the two layers of the network. Where n i and n j These represent the number of neurons (or nodes) between layers i and j. During the training phase of the neural network, a sufficient number of training samples are stored. Approximately 1000 training samples are used in the simulation. Once the training weights are determined, the training samples can be discarded because only the training weights are used during the testing phase.

[0230] The complexity depends on the amount of computation the design performs when applying weights in real time. To illustrate further, consider... Figure 5B And assume that each hidden layer has n i =n j = n neurons, and input vector x i It also has n dimensions. Then the total number of calculations (additions and multiplications) used for h hidden layers is approximately... Furthermore, the complexity of any learning algorithm depends on the dimension of the input (input vector x). i In machine learning, this is called a feature set; however, in this design, it is only three-dimensional.

[0231] Given the constantly changing traffic density, a multi-fingerprint-based database is proposed, which intelligently adapts among different fingerprints through learning. Furthermore, as an extension of the proposed design, an application is presented to improve spectral efficiency and the performance of multi-functional beam transmission, where beam pairs satisfying the required received signal strength participate in improving spectral efficiency. Beam alignment based on multiple fingerprints can provide superior performance compared to beam alignment based on a single fingerprint. Moreover, a learning-assisted multi-fingerprint design is shown to offer better fidelity than a scheme employing multiple fingerprints but abandoning learning. Additionally, the proposed learning-assisted beam alignment design is performed similarly to beam scanning-based beam alignment, where exhaustive beam search is performed with reduced search complexity. More specifically, our design is able to maintain target RSS in dense vehicular environments, while beam alignment based on a single fingerprint and line-of-sight (LOS) experiences congestion.

[0232] Figure 18 This is a flowchart illustrating an exemplary process for determining beam pairs that provide a desired RSS, according to some embodiments. For some embodiments, exemplary process 1800 may include obtaining 1802 input data, which includes user equipment location, number of user equipments, and desired received signal strength. For some embodiments, exemplary process 1800 may also include processing 1804 the input data with a neural network having weights determined from a training phase to generate a set of one or more beam pair indices. For some embodiments, exemplary process 1800 may also include performing 1806 beam search on at least a subset of the set of beam pair indices. For some embodiments, exemplary process 1800 may also include receiving 1808 at least one beam pair index that provides the desired received signal strength (RSS) from a user equipment.

[0233] Exemplary methods according to some implementations may include: obtaining input data including user equipment location, number of user equipment, and desired received signal strength; processing the input data with a neural network having weights determined from a training phase to generate a set of one or more beampair indices; performing a beam search on at least one subset of the set of beampair indices; and receiving from the user equipment at least one beampair indices providing the desired received signal strength.

[0234] For some implementations of the exemplary method, an out-of-band signaling channel can be used to transmit the location of the user equipment from the user equipment to the base station.

[0235] In some implementations of the exemplary method, the base station can know the number of user equipment based on vehicle density.

[0236] For some implementations of the exemplary method, the softmax algorithm can be used to further generate a set of one or more beam pair indices.

[0237] For some implementations of the exemplary method, the softmax algorithm can generate probabilities associated with a set of beam pair indices.

[0238] For some implementations of the exemplary method, a set of one or more beam pair indexes may be stored in a database during the training phase.

[0239] For some implementations of the exemplary method, the training phase may include: obtaining training samples for each training location; initializing the weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: using the corresponding weight vector to compute the neuron output for each layer; applying the softmax function to obtain the class probability; compute the weight matrix and bias vector; and perform backpropagation of the error.

[0240] Additional exemplary methods according to some implementation schemes may include: obtaining location information from a user equipment based on an initial network access procedure; processing the location information using a neural network to generate a fingerprint output having a set of associated beam pairs; and performing beam training using the set of beam pairs.

[0241] Some implementations of the additional exemplary method may also include notifying the user equipment of candidate beam pairs based on a set of beam pairs.

[0242] For some implementations of the additional exemplary method, candidate beam pairs may be a subset of associated beam pairs.

[0243] Exemplary methods / apparatus according to some implementations may include: adapting a location-specific beamforming fingerprint to traffic conditions at a base station; and using the beamforming fingerprint for beam training.

[0244] Exemplary methods according to some implementations may include: obtaining input data including user equipment location, number of user equipments (UEs), and desired received signal strength; processing the input data with a neural network having weights determined from a training phase to generate a set of one or more beampair indices; performing a beam search on at least one subset of the set of beampair indices; and receiving from the user equipment at least one beampair indices providing the desired received signal strength.

[0245] For some implementations of the exemplary method, an out-of-band signaling channel can be used to transmit the location of the user equipment from the user equipment to the base station.

[0246] For some implementations of the exemplary method, the base station can know the number of UEs based on vehicle density.

[0247] For some implementations of the exemplary method, the softmax algorithm can be used to further generate a set of one or more beam pair indices.

[0248] For some implementations of the exemplary method, the softmax algorithm can generate probabilities associated with a set of beam pair indices.

[0249] For some implementations of the exemplary method, a set of one or more beam pair indexes may be stored in a database during the training phase.

[0250] For some implementations of the exemplary method, the training phase may include: obtaining training samples for each training location; initializing the weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: using the corresponding weight vector to compute the neuron output for each layer; applying the softmax function to obtain the class probability; compute the weight matrix and bias vector; and perform backpropagation of the error.

[0251] For some implementations of the exemplary method, processing the input data with a neural network may include: obtaining multiple fingerprints of different traffic conditions at the location of the user equipment; and using a neural network coupled with a softmax classifier to select one of the multiple fingerprints based on the traffic conditions at the location of the user equipment, wherein selecting one of the multiple fingerprints may generate a set of one or more beam pair indices, and the neural network may use weights determined from the training phase.

[0252] For some implementations of the exemplary method, traffic conditions may include the number of UEs at the user equipment location.

[0253] For some implementations of the exemplary method, the neural network can be a deep learning feedforward neural network.

[0254] For some implementations of the exemplary method, performing beam search may include: sending fingerprint information to a user equipment comprising a set of one or more beam pair indices; and performing a beam training process on at least a subset of the set of beam pair indices to select selected beam pairs from the set of one or more beam pair indices that satisfy the desired received signal strength.

[0255] Some implementations of the exemplary method may further include: obtaining training samples at at least one training location; initializing the weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: using the corresponding weight vector to compute the neuron outputs of at least one layer of the neural network; applying a softmax function to the output layer of the neural network to obtain class probabilities; updating the weight matrix and bias vector; and performing backpropagation of the error.

[0256] Some implementations of the exemplary method may also include determining a loss function between the predicted class probabilities and the true class probabilities, wherein a convergence metric threshold is reached if the loss function is less than the convergence metric threshold.

[0257] Some implementations of the exemplary method may also include selecting at least one beam pair index from a set of beam pair indexes for transmitting data to a receiver.

[0258] For some implementations of the exemplary method, selecting at least one beam pair index may include using a multi-functional beam transmission scheme.

[0259] In some implementations of the exemplary method, the selection of at least one beam pair index can be repeated periodically.

[0260] In some implementations of the exemplary method, the selection of at least one beam pair index can be performed when an event is triggered, and the triggering event can be the detection of a change in the parameters of the user equipment.

[0261] Exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, are operable to perform any of the exemplary methods.

[0262] Additional exemplary methods according to some implementation schemes may include: obtaining location information from a user equipment based on an initial network access procedure; processing the location information using a neural network to generate a fingerprint output having a set of associated beam pairs; and performing beam training using the set of beam pairs.

[0263] Some implementations of the additional exemplary method may also include notifying the user equipment of candidate beam pairs based on a set of beam pairs.

[0264] For some implementations of the additional exemplary method, candidate beam pairs may be a subset of associated beam pairs.

[0265] Additional exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, are operable to perform any of the additional exemplary methods.

[0266] Other exemplary methods according to some implementation schemes may include: adapting a location-specific beamforming fingerprint to traffic conditions at a base station; and using the beamforming fingerprint for beam training.

[0267] Another exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions that, when executed by the processor, are operable to perform the following operations: adapting a location-specific beam to a fingerprint at a base station to traffic conditions; and beam training the fingerprint using the beam.

[0268] Another additional exemplary method according to some implementations may include: obtaining input data including user equipment (UE) location, traffic density of the UE, and received signal strength (RSS) threshold; processing the information of the UE with a neural network having weights determined from a training phase to generate a set of one or more beam pair indices; transmitting the set of one or more beam pair indices to the UE; transmitting to the UE an instruction to perform beam training using the set of one or more beam pair indices; and receiving from the UE at least one beam pair index that satisfies the RSS threshold.

[0269] Another additional exemplary apparatus according to some embodiments may include: a processor; and a non-transitory computer-readable medium storing instructions operable, when executed by the processor, to perform the following operations: obtaining input data including user equipment (UE) location, traffic density of the UE, and a received signal strength (RSS) threshold; processing the UE information with a neural network having weights determined from a training phase to generate a set of one or more beampup indexes; transmitting the set of one or more beampup indexes to the UE; transmitting to the UE an instruction to perform beam training using the set of one or more beampup indexes; and receiving from the UE at least one beampup index that satisfies the RSS threshold.

[0270] It should be noted that the various hardware elements of one or more embodiments described in the implementation schemes are referred to as “modules” that perform (i.e., execute, implement, etc.) the various functions described herein in conjunction with the respective modules. As used herein, a module includes hardware (e.g., one or more processors, one or more microprocessors, one or more microcontrollers, one or more microchips, one or more application-specific integrated circuits (ASICs), one or more field-programmable gate arrays (FPGAs), one or more memory devices) that a person skilled in the art would consider suitable for a given specific implementation. Each of the said modules may also include executable instructions for performing one or more functions described as being performed by the respective module, and it should be noted that these instructions may take the form of or include the following instructions: hardware (i.e., hardwired) instructions, firmware instructions, software instructions, etc., and may be stored in any suitable one or more non-transitory computer-readable media (such as commonly referred to as RAM, ROM, etc.).

[0271] Although features and elements have been described above in specific combinations, those skilled in the art will understand that each feature or element may be used alone or in any combination with other features and elements. Furthermore, the methods described herein may be implemented in a computer program, software, or firmware incorporated in a computer-readable medium for execution by a computer or processor. Examples of computer-readable storage media include, but are not limited to, read-only memory (ROM), random access memory (RAM), registers, cache memory, semiconductor memory devices, magnetic media (such as internal hard disks and removable disks), magneto-optical media, and optical media (such as CD-ROM disks and digital versatile optical discs (DVDs)). A processor associated with the software may be used to implement a radio frequency transceiver for a WTRU, UE, terminal, base station, RNC, or any host computer.

Claims

1. A method comprising: obtaining input data comprising a wireless transmit / receive unit (WTRU) location, a number of WTRUs, and a desired received signal strength; processing the input data with a trained neural network to generate a set of one or more beam pairs, wherein each beam pair is associated with the WTRU location, the number of WTRUs, and the desired received signal strength; transmitting information indicative of one or more beam pair options to a WTRU, wherein the one or more beam pair options are based on the set of one or more beam pairs; transmitting a signal using a transmit beam of at least one of the one or more beam pair options for at least one of the one or more beam pair options; and receiving information from the WTRU indicative of at least one suitable beam pair from the one or more beam pair options, wherein the suitable beam pair provides the desired received signal strength.

2. The method of claim 1, wherein the WTRU location is received from the WTRU based on an initial network access procedure.

3. The method of any one of claims 1-2, wherein the network node knows the number of WTRUs according to a vehicle density.

4. The method of claim 3, wherein the trained neural network uses weights determined from a training phase, and further comprising: obtaining training samples of at least one training location; initializing a weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: computing neuron outputs of at least one layer of the trained neural network using a respective weight vector; applying a softmax function to an output layer of the trained neural network to obtain class probabilities; updating weight matrices and bias vectors; and performing backpropagation of errors.

5. The method of any one of claims 1-2, wherein the generated set of one or more beam pairs is further generated using a softmax algorithm.

6. The method of claim 5, wherein the softmax algorithm generates probabilities associated with a set of beam pairs.

7. The method of claim 1, wherein the generated set of one or more beam pairs is stored in a database during a training phase.

8. The method of claim 7, wherein the trained neural network uses weights determined from the training phase, and wherein the training phase comprises: obtaining training samples of each training location; initializing a weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: computing neuron outputs of each layer using a respective weight vector; applying a softmax function to obtain class probabilities; computing weight matrices and bias vectors; and performing backpropagation of errors.

9. The method of claim 1, wherein processing the input data with the trained neural network comprises: obtaining a plurality of sets of one or more beam pairs at different vehicle densities at the WTRU location; and and ​ ​ selecting, using the trained neural network comprising a softmax classifier, a set of one or more beam pairs from a plurality of sets of one or more beam pairs based on the vehicle density at the WTRU location, and wherein the trained neural network uses weights determined from a training phase.

10. The method of claim 9, wherein the vehicle density comprises a number of the WTRUs at the WTRU location.

11. The method of claim 9, wherein the trained neural network is a deep learning feedforward neural network.

12. The method of any of claims 1-2, 6-11, wherein the trained neural network uses weights determined from a training phase, and further comprising: obtaining training samples of at least one training location; initializing a weight vector to random values; and iteratively performing the following steps until a convergence metric threshold is reached: computing neuron outputs of at least one layer of the trained neural network using a respective weight vector; applying a softmax function to an output layer of the trained neural network to obtain class probabilities; updating weight matrices and bias vectors; and performing error backpropagation.

13. The method of claim 12, further comprising: determining a loss function between predicted class probabilities and true class probabilities, wherein the convergence metric threshold is reached if the loss function is less than the convergence metric threshold.

14. The method of claim 1, further comprising selecting the at least one suitable beam pair to transmit data to a receiver.

15. The method of claim 14, wherein selecting the at least one suitable beam pair comprises using a multi-function beam transmission scheme.

16. The method of claim 14, wherein selecting the at least one suitable beam pair is repeated periodically.

17. The method of claim 14, wherein selecting the at least one suitable beam pair is performed upon detecting a trigger event, and wherein the trigger event is a change in a parameter of the WTRU.

18. The method of claim 1, wherein the network node is a base station.

19. An apparatus comprising: a processor; and a non-transitory computer-readable medium storing instructions operative, when executed by the processor, to: obtain input data comprising a wireless transmit / receive unit (WTRU) location, a number of WTRUs, and an expected received signal strength; process the input data with a trained neural network to generate a set of one or more beam pairs, wherein each beam pair is associated with the WTRU location, the number of WTRUs, and the expected received signal strength; transmit information indicating one or more beam pair options to a WTRU, wherein the one or more beam pair options are based on the set of one or more beam pairs. ​ transmitting a signal using a transmit beam of the at least one of the one or more beam pair options; and receiving information from the WTRU indicating at least one suitable beam pair from the one or more beam pair options, wherein the suitable beam pair provides the desired received signal strength.

20. The apparatus of claim 19, wherein the WTRU location is received from the WTRU based on an initial network access procedure.

21. The apparatus of any one of claims 19-20, wherein the network node knows a number of WTRUs according to a vehicle density.

22. The apparatus of claim 21, wherein the trained neural network uses weights determined from a training phase, and is further configured to: obtain training samples for at least one training location; initializing the weight vector to random values; and iteratively perform the following steps until a convergence measure threshold is reached: compute neuron outputs for at least one layer of the trained neural network using a respective weight vector; apply a softmax function to an output layer of the trained neural network to obtain class probabilities; update weight matrices and bias vectors; and perform error backpropagation.

23. The apparatus of any one of claims 19-20, wherein the generated set of one or more beam pairs is further generated using a softmax algorithm.

24. The apparatus of claim 23, wherein the softmax algorithm generates probabilities associated with the set of beam pairs.

25. The apparatus of claim 19, wherein the generated set of one or more beam pairs is stored in a database during a training phase.

26. The apparatus of claim 25, wherein the trained neural network uses weights determined from the training phase, and wherein the training phase comprises: obtaining training samples for each training location; initializing weight vectors to random values; and iteratively performing the following steps until a convergence measure threshold is reached: computing neuron outputs for each layer using a respective weight vector; applying a softmax function to obtain class probabilities; computing weight matrices and bias vectors; and performing error backpropagation.

27. The apparatus of claim 19, wherein in processing the input data with the trained neural network, the apparatus is further configured to: obtain multiple sets of one or more beam pairs for different vehicle densities at the WTRU location; and select the set of one or more beam pairs from the multiple sets of one or more beam pairs based on the vehicle density at the WTRU location using the trained neural network comprising a softmax classifier, and wherein the trained neural network uses weights determined from a training phase.

28. The apparatus of claim 27, wherein the vehicle density comprises a number of WTRUs at the WTRU location. ​ 29. The apparatus of claim 27, wherein the trained neural network is a deep learning feedforward neural network.

30. The apparatus of any of claims 19-20, 24-29, wherein the trained neural network uses weights determined from a training phase, and is further configured to: obtain training samples of at least one training position; and iteratively perform the following steps until a convergence metric threshold is reached: compute neuron outputs of at least one layer of the trained neural network using a corresponding weight vector; apply a softmax function to an output layer of the trained neural network to obtain class probabilities; update weight matrices and bias vectors; and perform error backpropagation.

31. The apparatus of claim 30, further configured to: determine a loss function between predicted class probabilities and true class probabilities, wherein the convergence metric threshold is reached if the loss function is less than the convergence metric threshold. initializing the weight vector to random values; 32. The apparatus of claim 19, further configured to select the at least one suitable beam pair to transmit data to a receiver.

33. The apparatus of claim 32, wherein selecting the at least one suitable beam pair comprises using a multi-function beam transmission scheme.

34. The apparatus of claim 32, wherein selecting the at least one suitable beam pair is repeated periodically.

35. The apparatus of claim 32, wherein the at least one suitable beam pair is selected upon detection of a trigger event, and wherein the trigger event is a change in a parameter of the WTRU.

36. The apparatus of claim 19, wherein the network node is a base station. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​

Images