Multi-dimensional geometric wireless channel rendering using machine learning models

Abstract

Certain aspects of the present disclosure provide techniques and apparatus for training and using a machine learning model to estimate a layout of a spatial region. An example method generally includes estimating a representation of a channel using a machine learning model trained to generate an estimated representation of the channel based on a position of a transmitter in a spatial environment, a position of a receiver in the spatial environment, and a three-dimensional representation of the spatial environment. One or more actions are taken based on the estimated representation of the channel.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. patent application serial number 17 / 935,046, filed on September 23, 2022, which is hereby incorporated by reference into this document.

[0003] introduction

[0004] Aspects of the present disclosure relate to using a machine learning model to estimate a wireless channel.

[0005] In a wireless communication system, a transmitter and a receiver communicate by sending signals to each other over a wireless channel, which can be represented as a component received at a receiver over a given frequency band. Information about the wireless channel can be used for various purposes. For example, information about the wireless channel can be used to identify various parameters used for communication between the transmitter and the receiver, such as beamforming parameters, directional beam selection, etc. Information about the wireless channel can also be used to determine the layout of the spatial environment in which the transmitter and the receiver are located, which in turn can be used for various purposes, such as detecting devices entering and leaving different areas (e.g., based on a radius from a given device). Layout information and position estimates can also be used for many other purposes, such as emergency management within a spatial region, spatial optimization, etc.

[0006] The state of a wireless channel may generally depend on various factors in the spatial environment. For example, a wireless channel may be affected by sources of radio frequency interference, such as interfering network entities. In another example, hard surfaces such as walls, support columns, and materials in these environments may introduce attenuation and reflection of radio waves in radio frequency measurements obtained in the spatial region. Because the state of a wireless channel may depend on many factors that vary across different environments, it may be difficult to estimate the state of a wireless channel.

[0007] Therefore, there is a need for improved techniques for estimating wireless channels in spatial environments. Summary of the invention

[0008] One aspect provides a method for estimating a wireless channel using a machine learning model. An example method generally includes estimating a representation of the channel using a machine learning model, the machine learning model being trained to generate an estimated representation of the channel based on a location of a transmitter in a spatial environment, a location of a receiver in the spatial environment, and a three-dimensional representation of the spatial environment. One or more actions are taken based on the estimated representation of the channel.

[0009] Another aspect provides a method for training a machine learning model to estimate a wireless channel. An example method generally includes generating a training data set including a plurality of channel representations. Each channel representation generally includes information for a plurality of rays corresponding to a path between a transmitter and a receiver in a spatial environment. The machine learning model is trained to generate an estimated representation of the channel based on a three-dimensional representation of the spatial environment and the training data set.

[0010] Other aspects provide: a processing system configured to perform the aforementioned methods and those described herein; a non-transitory computer-readable medium comprising instructions that, when executed by one or more processors of the processing system, cause the processing system to perform the aforementioned methods and those described herein; a computer program product embodied on a computer-readable storage medium comprising code for performing the aforementioned methods and those further described herein; and a processing system comprising components for performing the aforementioned methods and those further described herein.

[0011] The following description and the associated drawings set forth in detail certain illustrative features of one or more implementations. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings depict certain aspects of the one or more embodiments and therefore are not to be considered limiting of the scope of the present disclosure.

[0013] Figure 1 Depicted are signals transmitted between a transmitter and a plurality of receivers in a space environment.

[0014] Figure 2 Depicted is an example pipeline for estimating a wireless channel using a machine learning model in accordance with aspects of the present disclosure.

[0015] Figure 3A and Figure 3B Depicted are examples of a ray propagation estimation stage including one or more machine learning models that estimate ray propagation in a wireless channel and reception or termination of propagated rays in the wireless channel in accordance with aspects of the present disclosure.

[0016] Figure 4 Depicted are examples of ray-surface interaction blocks for modeling interactions between rays and surfaces during ray propagation in accordance with aspects of the present disclosure.

[0017] Figure 5 Depicted are examples of blindly transmitting a set of rays from a transmitter, the set of rays including a subset of rays that contribute to signal reception at a receiver, in accordance with aspects of the present disclosure.

[0018] Figure 6 A ray launching model for launching rays that facilitate signal reception at a receiver in accordance with aspects of the present disclosure is depicted.

[0019] Figure 7 Depicted are example operations for estimating a layout of a spatial region based on a machine learning model and an input data set of multi-dimensional samples from the spatial region in accordance with aspects of the present disclosure.

[0020] Figure 8 Depicted are example operations for training a machine learning model to estimate a layout of a spatial region based on an input data set of multi-dimensional samples from the spatial region in accordance with aspects of the present disclosure.

[0021] Fig. 9 Depicted are example implementations of a processing system on which a machine learning model is trained to estimate the layout of a spatial region in accordance with aspects of the present disclosure.

[0022] Fig.10 Depicted are example implementations of a processing system on which a machine learning model is used to estimate a layout of a spatial region in accordance with aspects of the present disclosure.

[0023] To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the figures. It is contemplated that elements and features of one embodiment may be beneficially incorporated in other embodiments without further recitation. DETAILED DESCRIPTION

[0024] Aspects of the present disclosure provide techniques for estimating a channel between a transmitter and a receiver based on positioning information for the transmitter and the receiver and a layout of the spatial region in which the transmitter and the receiver operate. Such estimation can be performed, for example, using a machine learning model that is trained to generate an estimate of the channel based on positioning information for the transmitter and the receiver and a layout of the spatial region in which the transmitter and the receiver operate.

[0025] Information about the channel, such as an estimate of the channel at the receiver, can be used for various tasks within a wireless communication system and / or within a system that uses devices communicating within the wireless communication system. For example, information about the channel can be used for signal management, such as beamforming or beam selection, for communication between a transmitter and a receiver. In another example, information about the channel can be used for various sensing tasks, such as position estimation, floor map estimation, etc. This information can in turn be used to determine how a spatial region will be used, to generate virtual reality or extended reality scenes in a spatial region, for traffic management within a spatial region, for position estimation, etc. In yet another example, the learned positioning of signal reflectors in a spatial environment and the signal reflection characteristics of these reflectors can be used to perform various tasks in a dynamic environment, such as autonomous driving or other autonomous operations in which the state of the environment changes continuously.

[0026] Given information about the location of the transmitter and the location of the receiver in the spatial environment, various techniques can be used to estimate information about the channel. For example, a neural statistical model implemented as a deep neural network (DNN) can be used to generate an estimate of the channel based on various field measurements for the transmitter and the receiver. These models can allow the channels in various general scenarios to be estimated with minimal or at least relatively very low computational costs. However, the channel estimates generated by these models may reflect channel estimates in a general environment and may not take into account environment-specific characteristics that affect the state of the channel at the receiver. For example, these general models may not take into account reflections of signals from surfaces within the spatial environment, signal attenuation due to these reflections, signal attenuation and reflection characteristics of different materials used in the spatial environment, etc.

[0027] Various aspects of the present disclosure provide techniques that allow for estimating a channel based on a spatially consistent machine learning model that takes into account information about the environment in which the transmitter and receiver operate when estimating information about the channel. By doing so, various aspects of the present disclosure can generate accurate estimates of the channel between the transmitter and receiver in any given environment. These estimates of the channel can take into account different components of a wireless signal received at the receiver at different times due to different propagation paths in the spatial environment, and can ignore signal components that have minimal impact on the overall received channel at the receiver.

[0028] By accurately estimating channels in a spatial environment, aspects of the present disclosure may allow for improved accuracy of applications that use estimates of channels to perform various tasks. For example, accurate channel estimates generated by the machine learning models discussed herein may allow for improved communications in various spatial environments and at various frequencies at which transmitters and receivers communicate, as these accurate channel estimates may be used to identify obstructions within the spatial environment that degrade wireless signals, estimate optimal beams to be used in high-frequency (e.g., millimeter wave (mmWave)) communications between a transmitter and a receiver, and so forth. Additionally, the machine learning models discussed herein may allow for modeling of the propagation of signals between a transmitter and a receiver, which may allow information about the spatial environment, such as the presence of reflective or refractive surfaces in the spatial environment and the RF reflectivity and attenuation characteristics of these surfaces, to be obtained from the estimates of the channels generated by these machine learning models.

[0029] Example RF signal in a spatial region

[0030] Figure 1 Example signals transmitted between a transmitter and various receivers within a spatial environment are illustrated.

[0031] As illustrated, within the spatial environment 100, a transmitter 110 transmits signals to receivers 112, 114, and 116. The transmitter 110 may be, for example, a gNodeB in a cellular telecommunication system, and the receivers 112, 114, and 116 may be user equipment (UE) served by the gNodeB.

[0032] Because the space environment 100 includes various surfaces from which signals can be reflected, such as walls, floors, ceilings, or other surfaces, signaling sent between the transmitter 110 and the receivers 112, 114, and 116 may include a line-of-sight (LOS) component and, in some cases, one or more non-LOS components. For example, as illustrated, the signal received at the receiver 112 includes: a LOS component corresponding to path 1 102; a first non-LOS component corresponding to path 2 104, where the receiver 112 receives a reflection of the signal from a wall in the space environment 100; and a second non-LOS component corresponding to path 3 106, where the receiver 112 receives a reflection of the signal from the floor in the space environment 100. Similarly, the signal received at the receiver 114 includes a LOS component and a non-LOS component, where the receiver 114 receives a reflection of the signal from a wall in the space environment 100.

[0033] Because the receiver may receive signals that include both LOS components and non-LOS components, the total impulse response (or signal strength) f for any signal transmitted by transmitter 110 θAn impulse response for a LOS component and an impulse response for a non-LOS component may thus be included. As illustrated in impulse response curve 120 for receiver 112 and impulse response curve 130 for receiver 114, the LOS component may have the highest impulse response and the shortest delay between transmission by transmitter 110 and reception by an appropriate receiver, and the non-LOS components may have a lower impulse response and a longer delay between transmission and reception relative to the LOS components. These non-LOS components may have a lower impulse response due to various factors, including: signal attenuation from surfaces from which these signals reflect, as different surfaces may introduce different amounts of signal attenuation into the non-LOS components; and the overall distance traversed between the transmitter and receiver, as signal strength generally decreases as a function of distance.

[0034] In general, the global representation of a wireless channel can be modeled by the following equation:

[0035]

[0036] where y(t) is the global representation of the channel at time t, and a i represents the gain for the i-th path, and τ i represents the delay for the i-th path. This global representation illustrates that the wireless channel is an aggregation of the signal received at the receiver for each path from the transmitter to the receiver (e.g., for both LOS and non-LOS components). However, this global representation may not allow modeling of the propagation characteristics of the signal on any given path. More specifically, for a given path, the global representation may not allow modeling of reflections and signal attenuation for any non-LOS component of the signal.

[0037] In order to account for various characteristics of a three-dimensional spatial environment that affect signal reflections and signal attenuation, and to allow channel estimation within various spatial environments based on signal measurements obtained for received signals within the spatial environment, various aspects of the present disclosure provide techniques for estimating a representation of a channel using a neural network based on the locations of a transmitter and a receiver within the spatial environment and a three-dimensional representation of the spatial environment. As discussed in further detail herein, the neural network can estimate the channel based on estimates of signal attenuation due to propagation distance and signal attenuation due to ray-surface interactions (e.g., signal absorption, reflection angle, etc.). The entire channel can be rendered based on these estimates so that the channel includes signal components that are considered to have been received by the receiver (e.g., propagated to a point within a defined receiving ellipsoid for the receiver) and does not include signal components that have an insignificant effect on the entire channel due to attenuation below a threshold signal strength.

[0038] Example estimation of a wireless channel using a machine learning model

[0039] In a wireless communication system, a signal may be represented as a series of rays emitted from a location within a spatial environment of a transmitter and received when the signal reaches a receiving ellipsoid (e.g., a sphere or a non-spherical three-dimensional figure having an ellipsoidal shape), which is defined as the location of the receiver within the spatial environment and a radius from this location. In order to model the signal, the surface of the ellipsoid surrounding the transmitter may be tessellated into a set of surfaces (e.g., arcs, circles, polygons, etc.), and rays may be emitted omnidirectionally from the transmitter, where each ray corresponds to a specific surface from the set of surfaces. As discussed in further detail, each ray may be tracked until the ray satisfies a specified termination criterion (e.g., leaves the spatial environment, reaches a defined maximum number of reflections or diffractions, etc.), the signal strength of the ray drops below a threshold intensity, or until the ray reaches the receiving ellipsoid, which may be defined as having a radius where α represents the angle between adjacent rays sent by the receiver, and d represents the ray length. In general, the choice of α and the corresponding sphere radius represents a tradeoff between accuracy and complexity; smaller values ​​of α may result in more rays being generated to trace, with correspondingly increased accuracy and complexity, while larger values ​​of α may result in fewer rays being generated to trace, with correspondingly reduced accuracy and complexity.

[0040] In order to parameterize the ray and allow the machine learning model to be trained to make the ray defined as u k The ray "propagates", the impulse response f θ It can be defined according to the following expression: θ : Where F represents the spatial environment, and Represents the state of the ray after r interactions within the spatial environment. The environment F can be represented in various ways, such as a series of vertices V, a series of polygonal (e.g., triangular) faces F, a three-dimensional mesh (defined as a collection of vertices V and faces F), etc. In general, the spatial environment F may correspond to a specific geometric shape and may not embed information about the materials within the spatial environment.

[0041] ray can be represented as a data set, including geometry, signal characteristics, and state information. The geometry information may include information about: the position x of the ray at the rth interaction (r) , from the origin (e.g. the location of the transmitter) to x (r) The distance traveled d, the position where the signal intersects the receiving ellipsoid t rx , the position where the signal reaches the reflection point t s and receiving ellipsoid rx r The signal characteristics may include gain information a, flight time information τ for the kth rayk , the departure angle representing the direction in which the ray departs from the transmitter and the angle of arrival, which represents the direction in which the ray reaches the receiver Finally, the state information may include the state of the signal at the rth interaction and the state of the receiver

[0042] Figure 2 An example pipeline 200 for estimating a wireless channel using a machine learning model in accordance with aspects of the present disclosure is illustrated. As illustrated, the pipeline 200 includes a ray emission stage 210, a ray propagation estimation stage 220, a filtering and aggregation stage 230, and a processing stage 240.

[0043] At the ray emission stage 210, the ray emission of the transmitter is represented by x tx The information about the position of can be used to launch multiple rays. In some aspects, from the position x tx The emitted rays may be emitted blindly, where the transmitter is approximated as an icosahedral sphere (or a geodesic polyhedron, where an ellipsoid (such as a sphere) corresponding to the transmitter is modeled as a set of polygonal (e.g., triangular) faces from which rays are emitted. In some aspects, rays may be emitted from the transmitter based on a learned emission scheme at the ray emission stage 210, where one or more models are used to identify a relevant set of rays to emit based on the likelihood that the rays will reach a receiving ellipsoid with sufficient received signal strength.

[0044] For each ray emitted at the ray emission stage 210, the ray propagation estimation stage 220 propagates the ray within the spatial environment F via the ray caster 222, and determines at the receive / terminate block 224 whether the ray has reached the receiving ellipsoid or reached a termination condition indicating that the ray will not meaningfully contribute to an impulse response to a received signal at the receiver.

[0045] Subsequently, at the filtering and focusing stage 230, the rays may be filtered based on whether they are received at the receiving ellipsoid. The filtering may be performed based on the received power of the ray when the ray enters the receiving ellipsoid and based on the number of ray-surface interactions (such as reflections and / or refractions from surfaces in the space environment F). In general, if the ray enters the receiving ellipsoid of the receiver with a signal power exceeding a threshold level, the ray may be considered to have been received by the receiver, and if the ray enters the receiving ellipsoid of the receiver with a signal power falling below this threshold, the ray may be considered to not meaningfully contribute to an impulse response to the received signal and may be ignored. In addition, because each ray-surface interaction (e.g., reflections and / or refractions caused by surfaces in the space environment F) attenuates the signal, a maximum number of ray-surface interactions may be defined for the signal that meaningfully contributes to an impulse response to the received signal. If the number of ray-surface interactions estimated for a ray exceeds a threshold number of ray-surface interactions, the ray may be ignored when generating a representation of the channel.

[0046] At processing stage 240, the rays identified at filtering and gathering stage 230 as being relevant to the impulse response at the receiver may be gathered. This gathering may be for the rays within the spatial environment F for the rays located at x tx The transmitter at x rx The estimate of the channel at the receiver at is denoted h.

[0047] To generate this estimated representation of the channel, the renderer implemented via pipeline 200 can be trained to minimize the point-wise channel loss. For example, the machine learning model used to implement pipeline 200 can be trained according to the following expression:

[0048]

[0049] where at a given receiver position x rx , the transmitter's position x tx and the spatial environment layout F input, h i indicates the actual state of the channel, and represents the estimated (rendered) state of the channel. The loss function to be minimized may be, for example, a chamfer distance or other distance that measures the difference between the actual state of the channel and the estimated (rendered) state.

[0050] In order to train the machine learning model implemented via pipeline 200, a data set including multiple channel representations can be generated. Each channel representation can include multiple sets of vectors, where each set of vectors from the multiple sets corresponds to a specific path between a transmitter and a receiver in the spatial environment F. In order to reduce the size of the data set, a mask can be used to remove irrelevant vectors from each set of vectors. In general, irrelevant vectors, and therefore vectors that can be masked before training the machine learning model, can correspond to points where rays have signal strengths below a threshold power and therefore the rays are considered to no longer meaningfully contribute to the channel impulse response at the receiver.

[0051] Figure 3A An example of a ray propagation estimation stage 220 including one or more machine learning models that estimate ray propagation in a wireless channel and reception or termination of propagated rays in the wireless channel is depicted in accordance with aspects of the present disclosure.

[0052] As illustrated, to estimate the propagation of the ray, it will be represented as With gain The kth ray of and the layout F of the spatial environment are input into the ray-triangle intersection block 310. The ray-triangle intersection block 310 determines the next position of the ray given its current direction and current position relative to a particular face within the spatial environment, and if so, how the ray interacts with that face. The faces in F can be defined as Generally speaking, when determining whether a ray interacts with a particular face within a spatial environment, the ray-triangle intersection block 310 may solve the equation for t min :

[0053] The new position of the ray and the distance from the previous ray position are then input into the ray-surface interaction block 320. The ray-surface interaction block 320 generally uses the estimated direction and position information to determine whether the ray is expected to interact with a surface (e.g., reflect and / or refract from a surface). In general, the output of the ray-surface interaction block 320 may be a new direction. and gain To generate new directions and gain The ray-surface interaction block 320 may use the spatial model and the directional model to estimate how rays will reflect or refract relative to surfaces in the spatial environment F and the signal attenuation imposed by reflection or refraction relative to the surfaces.

[0054] After propagation and determining at the ray-surface interaction block 320 whether the ray interacts with a surface in the spatial environment F, the receive / termination block 224 determines whether the beam has been received or terminated. In general, a beam is considered to have been received if the estimated position of the beam is on the surface of the receive ellipsoid or is inside the receive ellipsoid. To determine whether a beam is terminated, as discussed above, the receive / termination block 224 may use a threshold number of ray-surface interactions and / or a threshold signal strength (or amount of gain) to determine whether the beam meaningfully contributes to an impulse response for a channel received at a receiver. In general, if the estimated signal strength (or gain) for a beam drops below a threshold, or if the signal has interacted with a surface more than a threshold number of times, it may be determined that the beam does not substantially contribute to an overall impulse response for the signal at the receiver, and therefore the beam may be ignored moving forward.

[0055] At free space loss block 330, signal attenuation due to propagation distance (independent of reflections from surfaces within the space environment) may be modeled to generate an estimate of the state of the ray at the receiver. Free space loss may be represented by Friis' Equation:

[0056]

[0057] In the above equation, K, d 0 and λ can be learnable parameters. K generally corresponds to a learned constant representing the antenna gain, d 0 corresponds to the learned reference distance, and λ corresponds to the learned signal wavelength. r represents the power P input to the transmitting antenna t and the power received at the receiving antenna for a given ray as a function of the distance d travelled by the given ray.

[0058] Figure 3B Another example of a ray propagation estimation stage 220 including one or more machine learning models that estimate ray propagation in a wireless channel and reception or termination of a propagated ray in the wireless channel according to aspects of the present disclosure is illustrated. In this example, the next position of the ray generated by the ray-triangle intersection block 310 and the distance from the previous ray position can be input into the ray-surface interaction block 320. At the same time, the carrier wavelength λ and timing information t generated by the ray-triangle intersection block 310 are k This can be input to the free space loss block 330. The ray-surface interaction block 320 generates and the angle of incidence and surface attenuation s 1 Three-dimensional rotation with condition The surface attenuation corresponds to the amount of attenuation caused by the surface with which the ray interacts. Meanwhile, the free space loss block 330 uses the carrier wavelength λ and the timing information t k To calculate for a given ray and two points The effects of free space antenna gain, path loss, and phase variation on the distance between 2 In general, the free space loss block 330 may use a complementary neural network that learns the free space effect on the ray to calculate s 2 , these free space effects include free space antenna gain, path loss, and the phase change characteristics of rays in a given space environment. This complementary neural network can be represented by the following expression: As shown, the rotation A, surface attenuation s 1 and free space effects 2 An input may be made to the receive / terminate block 224 to determine whether a ray is received at a receiver (e.g., enters a receive ellipsoid defined for the receiver in a spatial environment) or is terminated (e.g., has a signal strength below a threshold level, indicating that the ray will not meaningfully contribute to an impulse response at the receiver).

[0059] Figure 4 4 illustrates an example of a ray-surface interaction block 320 for modeling the interaction between a ray and a surface in a spatial environment during the propagation of the ray according to aspects of the present disclosure. As illustrated, the ray-surface interaction block 320 includes a spatial multilayer perceptron (MLP) 410 and a directional MLP 420. The ray-surface interaction network is generally directed to modeling the interaction between a ray and a surface (or surfaces) in a spatial environment. ) The interaction originates from In and in the direction The outgoing ray is estimated by traveling up a given ray. The outgoing ray can be modeled as the new direction of the ray and an estimate of the corresponding gain change caused by colliding with a surface in a spatial environment. Because the direction and gain change of a ray after interacting with a surface generally depends on the electromagnetic properties (reflection, refraction, scattering, absorption, etc.) of the material constituting the surface, aspects of the present disclosure can learn these properties by associating spatial regions in an environment with specific electromagnetic properties.

[0060] The spatial MLP 410 and the directional MLP 420 are generally configured to obtain spatial characteristics (e.g., corresponding to positions) by The spatial coordinates are mapped to electromagnetic properties of a specific surface in the spatial environment by mapping the spatial coordinates to properties such as gain factors. As illustrated, the spatial MLP 410 can encode electromagnetic properties that are specific to a spatial region but independent of the incident direction, while the directional MLP 420 predicts electromagnetic properties that are specific to a spatial region but independent of the incident direction. Incident on surface f iThe spatial MLP 410 can be based on the point The surface f i One-hot encoding, point The surface normal vector n at i and signed distance The three-dimensional vector input is used to generate the space vector v i , where sdf represents the signed distance function between coordinate x and face f. The output of spatial MLP 410 can be used together with the direction together are provided as input to estimate the gain factor s i and rotation Directed MLP 420. Rays The resulting complex gain can be modeled as Predicted rotation Can be based on The rotation is generated by the Euler-Rodrigues parameterization with the rotation axis encoded as the angle θ around which it is rotated. This rotation can be represented by the matrix A, so that the new direction of the ray k is defined as

[0061] Example raycasts for estimating wireless channels using a machine learning model

[0062] In some aspects, to transmit rays for estimating a wireless channel using a machine learning model, the ray emission stage 210 may use a "blind" emission scheme. In a "blind" emission scheme, as discussed above, the transmitter may be approximated as an icosahedral sphere, and rays may be generated and transmitted from the transmitter along vertex normals associated with each face of the icosahedral sphere. The "blind" emission scheme generally causes some beams to arrive at a receiving ellipsoid representing a receiver in a wireless communication system. However, many of the beams generated in the "blind" emission scheme may not contribute, or at least may not meaningfully contribute, to a channel impulse response for a received signal at the receiver.

[0063] Figure 5 An example 500 of "blindly" transmitting rays from a transmitter to a receiver in a space environment is illustrated. It should be noted that Figure 5A simplified example is illustrated in two dimensions, and the techniques discussed herein can be used to transmit rays in three-dimensional space. As illustrated in example 500, the transmitter transmits multiple rays omnidirectionally, and only a subset of these rays can actually contribute to the channel impulse response for the received signal at the receiver. In general, rays that contribute to the channel impulse response for the received signal at the receiver can be rays that enter the receiving ellipsoid 510 with a signal strength above a defined threshold, while rays that do not meaningfully contribute to the channel impulse response for the received signal at the receiver can be rays that enter the receiving ellipsoid 510 with a signal strength below the defined threshold, rays that reflect from a surface in the spatial environment F more than a threshold number of times, rays that leave the area of ​​interest, etc.

[0064] As illustrated, the rays that contribute to the channel impulse response for the received signal at the receiver may include rays corresponding to LOS components and non-LOS components. For example, the first ray 522 may be a LOS component that directly enters the receiving ellipsoid 510 without reflection or other interaction with the surface in the space environment F. Meanwhile, the second ray 524 may be a non-LOS component that enters the receiving ellipsoid 510 after reflecting from the surface 520 in the space environment F.

[0065] like Figure 5 As illustrated in , because a small number of rays relative to the total number of rays emitted from the transmitter can contribute to the channel impulse response for the received signal at the receiver, computational resource utilization can be reduced by sending rays along paths that meaningfully contribute to the channel impulse response for the received signal. By doing so, computational resources such as processing time and memory are not wasted on estimating the propagation of a large number of rays that will not contribute to the channel impulse response for the received signal. To allow rays to be emitted along paths that meaningfully contribute to the channel impulse response for the received signal, given the location x of the transmitter tx , the receiver's position x rx and a three-dimensional layout of a spatial environment F, aspects of the present disclosure may use a ray-casting neural network trained to learn which directions meaningfully contribute to the channel impulse response.

[0066] Figure 6 An example of a ray emission model 600 for emitting rays that contribute to signal reception at a receiver in accordance with aspects of the present disclosure is illustrated. In some aspects, the ray emission model 600 may correspond to Figure 2 The ray emission stage 210 is illustrated in FIG. Given the three-dimensional layout of the spatial environment F, the position x of the transmitter tx and the receiver position x rxGiven an input of , the ray launching model 600 can generally be trained to produce a distribution (which can be visualized as a heat map) that prioritizes ray directions that meaningfully contribute to the channel impulse response of the received signal at the receiver. Because the ray launching model 600 launches rays that contribute to signal reception at the receiver, the ray launching model 600 can be trained to de-prioritize ray directions that do not meaningfully contribute to the channel impulse response of the received signal. As illustrated, the ray launching model 600 can be constructed as a neural network including a trainable embedding module 610, an image prior module 620, and a trainable ray launching module 630.

[0067] The embedding module 610 is generally trained to transform x tx and x rx The inputs (e.g., the locations of the receiver and the transmitter in the spatial environment) are encoded into a high-dimensional embedding 612. In general, the embedding module 610 models the effects of the environment on the interactions between the receiver and the transmitter in the spatial environment. As discussed, these interactions can include obstacles between the transmitter and the receiver that block LOS components and non-LOS components from entering the receiving ellipsoid associated with the receiver, reflection points (and how these reflection points reflect signals), scattering due to various surfaces in the spatial environment, etc. In some aspects, the embedding module 610 can be a network, such as a multi-layer perceptron (MLP) network, which captures high-frequency features of the interactions between the transmitter and the receiver.

[0068] The image prior module 620 incorporates domain information into the ray-casting neural network. The image prior module 620 can use image method techniques for path finding in ray tracing systems. Given the input of the positions of the transmitter and the receiver, the image prior module 620 can generate an output prior 622, which can represent a LOS prior associated with the LOS direction between the transmitter and the receiver and can represent first-order virtual anchor positions and second-order virtual anchor positions in some aspects. For example, these virtual anchors can represent transmitters outside the spatial environment that act as virtual sources of reflected signals when the reflected signals are modeled as LOS components at the receiver.

[0069] In general, the output prior 622 of the image prior module 620 can serve as a prior for the ray launch module 630, which uses the combination of the prior 622 from the image prior module 620 and the embedding 612 from the embedding module 610 generated by the combiner 626, and learns the deviations around the prior, such as obstacles, scattering, reflections, etc. In some aspects, the ray launch module 630 can be a U-Net or other convolutional network. In general, by fusing the embedding 612 and the prior 622, the ray launch module 630 can model complex interactions (such as scattering and obstacles) without considering its own simpler interactions with the spatial environment, such as LOS components (whether real or virtual). The output 640 of the ray launch module 630 can be a probability heat map generated in the directions along which rays can be sent (or emitted) from the transmitter, which illustrates the correlation of each ray with the channel impulse response at the receiving ellipsoid corresponding to the receiver. This probability heat map (which may be in two-dimensional space) may be the result of flattening a spherical probability heat map rendered on a transmission ellipsoid associated with the transmitter. However, it should be appreciated that other techniques for representing spherical signals (such as spherical harmonics, spherical convolutional neural networks (CNNs), etc.) may also or alternatively be used to model the probability heat map rendered on a transmission ellipsoid associated with the transmitter.

[0070] In general, a plurality of ray-shooting directions may be sampled from the output 640 of the ray-shooting module 630 using various techniques such as the Gumbel-Softmax technique. These sampled ray-shooting directions may then be provided as input to a ray tracer (e.g., Figure 2 ), as discussed above, the ray tracer estimates the propagation of rays associated with each sampled ray emission direction and uses the estimated propagation to estimate the channel given the locations of the transmitter and receiver in the spatial environment and the three-dimensional layout of the spatial environment.

[0071] Example channel rendering using a machine learning model

[0072] Figure 7 Illustrated are example operations 700 for estimating a layout of a spatial region based on a machine learning model and an input data set of multi-dimensional samples from the spatial region in accordance with aspects of the present disclosure.

[0073] As illustrated, operations 700 begin at block 710, where a representation of a channel is estimated using a machine learning model. In general, the machine learning model can be trained to generate an estimated representation of the channel based on a transmitter's location in a spatial environment, a receiver's location in the spatial environment, and a three-dimensional representation of the spatial environment.

[0074] In some aspects, the machine learning model is trained to generate an estimated representation of the channel based on "blind" transmission of rays. As discussed, the "blind" transmission of rays generally includes the transmission of rays uniformly sampled on an ellipsoid (e.g., a sphere) representing the transmitter. The uniform sampling can be, for example, uniform sampling of faces on an icosahedral sphere that approximates the ellipsoid representing the transmitter.

[0075] In some aspects, the machine learning model is trained to generate an estimated representation of the channel based on the position of the transmitter and the position of the receiver in the spatial environment. To this end, the machine learning model can use a ray emission model configured to estimate a set of rays. This ray emission model generally includes an embedding model, an image prior model, and a ray emission model. The embedding model is generally trained to generate an embedding representing the relationship between the position of the transmitter and the position of the receiver in the spatial environment. The image prior model generally encodes the position of the transmitter and the position of the receiver into line of sight information. Finally, the ray emission model is generally trained to identify multiple rays to be emitted toward the receiver based on the embedding representing the relationship between the positions of the transmitter and the receiver and the line of sight information generated by the image prior model. In some aspects, the ray emission model can be a model that is trained to generate a probability heat map flattened on a two-dimensional grid on an ellipsoid around the position of the transmitter.

[0076] In some aspects, the machine learning model includes: a first sub-model configured to estimate signal attenuation due to surface reflections in a spatial environment; and a second sub-model configured to estimate signal attenuation due to free space in the spatial environment. The first sub-model can be further configured to estimate the distance the signal has traveled relative to a transmission source or reflection point in the spatial environment.

[0077] In some aspects, estimating a representation of a channel includes estimating a vector set representing each ray in a plurality of rays in the channel. Each vector in the vector set may represent the propagation of a ray over time. In general, each vector includes positioning information, direction information, gain information, flight time information, and a validity score indicating whether the ray contributes to the channel. A corresponding vector set representing a corresponding ray in the plurality of rays may be terminated based on the position of the ray intersecting the position of the receiver or the power of the ray dropping below a threshold. As discussed, the termination of the vector set may indicate that the signal no longer has sufficient strength to meaningfully contribute to an impulse response for the channel at a receiving ellipsoid around the receiver.

[0078] In some aspects, the machine learning model comprises a model trained to minimize the loss over time between an estimated state of a channel and an actual state of the channel.

[0079] At box 720, operation 700 continues to take one or more actions based on the estimated representation of the channel. In some aspects, taking one or more actions includes generating a graphical rendering of the estimated representation of the channel in a three-dimensional representation of the spatial environment. In some aspects, taking one or more actions includes selecting one or more beams for communication between a transmitter and a receiver based on the estimated representation of the channel. In some aspects, taking one or more actions includes detecting the presence and position of an object in the spatial environment based on the estimated representation of the channel. In some aspects, taking one or more actions includes identifying material properties of the spatial environment based on the estimated representation of the channel.

[0080] Figure 8 Depicted are example operations 800 for training a machine learning model to estimate a layout of a spatial region based on an input data set of multi-dimensional samples from the spatial region in accordance with aspects of the present disclosure.

[0081] As illustrated, operations 800 begin at block 810, where a training data set including a plurality of channel representations is generated. Each channel representation generally includes information for a plurality of rays corresponding to a path between a transmitter and a receiver in a spatial environment.

[0082] In some aspects, each ray in the plurality of rays is represented by a set of vectors representing the propagation of the ray over time. Each vector generally includes one or more of positioning information, direction information, gain information, time of flight information, or an effectiveness score indicating whether the ray contributes to the channel. In some aspects, generating the training data set includes generating a matrix on the plurality of rays using a mask configured to mask vectors in the matrix associated with effectiveness scores indicating that the ray does not contribute to the channel.

[0083] In some aspects, the training dataset includes channel representations associated with spatial environments having different surface attenuation and reflection characteristics. By including information from spatial environments having different surface attenuation and reflection characteristics, the resulting model can be used in a variety of spatial environments and can be not tightly coupled to a specific layout of the spatial environment or specific materials used to construct the spatial environment.

[0084] At block 820 , operation 800 continues by training the machine learning model to generate an estimated representation of the channel based on the three-dimensional representation of the spatial environment and the training data set.

[0085] In some aspects, the machine learning model includes: a first sub-model configured to estimate signal attenuation due to surface reflections in a spatial environment; and a second sub-model configured to estimate signal attenuation due to free space in the spatial environment. The first sub-model can be further configured to estimate the distance the signal propagates relative to a transmission source or reflection point in the spatial environment. In some aspects, the machine learning model can be further trained to generate a graphical rendering of an estimated representation of the channel in a three-dimensional representation of the spatial environment.

[0086] In some aspects, the estimated representation of the channel may include a set of vectors associated with each ray of the plurality of rays whose contribution to the channel exceeds a threshold contribution.

[0087] In some aspects, the machine learning model is trained based on minimizing the loss over time between the estimated state of the channel and the actual state of the channel. This loss can be, for example, a chamfer distance or other distance between the estimated state of the channel and the actual state of the channel.

[0088] In some aspects, the training data set includes a training data set of rays uniformly sampled on an ellipsoid (eg, a sphere) representing the transmitter.

[0089] In some aspects, operation 800 also includes training a ray emission model to generate a set of rays based on the position of the transmitter and the position of the receiver in the spatial environment. This ray emission model generally includes an embedding model, an image prior model, and a ray emission model. The embedding model is generally trained to generate an embedding representing the relationship between the position of the transmitter and the position of the receiver in the spatial environment. The image prior model generally encodes the position of the transmitter and the position of the receiver into line of sight information. Finally, the ray emission model is generally trained to identify multiple rays to be emitted toward the receiver based on the embedding representing the relationship between the positions of the transmitter and the receiver and the line of sight information generated by the image prior model. In some aspects, the ray emission model can be a model trained to generate a probability heat map flattened on a two-dimensional grid on an ellipsoid (e.g., a sphere) around the position of the transmitter. In some aspects, the ray emission model can be trained based on the aggregation of rays determined to have propagated to a position defined by a center point and a radius from the center point within a receiving ellipsoid (e.g., a sphere).

[0090] Example processing system for channel rendering using machine learning models

[0091] Fig. 9 Describes, for example, Figure 7 An example processing system 900 is described for estimating channels in a spatial environment based on a machine learning model.

[0092] The processing system 900 includes a central processing unit (CPU) 902, which may be a multi-core CPU in some examples. Instructions executed at the CPU 902 may be loaded, for example, from a program memory associated with the CPU 902 or may be loaded from the memory 924.

[0093] The processing system 900 also includes additional processing components customized for specific functions, such as a graphics processing unit (GPU) 904 , a digital signal processor (DSP) 906 , a neural processing unit (NPU) 908 , a multimedia processing unit 910 , and a wireless connectivity component 912 .

[0094] An NPU, such as NPU 908, is generally a dedicated circuit configured to implement all necessary control and arithmetic logic for executing machine learning algorithms, such as algorithms for processing artificial neural networks (ANNs), deep neural networks (DNNs), random forests (RFs), etc. An NPU is sometimes alternatively referred to as a neural signal processor (NSP), a tensor processing unit (TPU), a neural network processor (NNP), an intelligence processing unit (IPU), a vision processing unit (VPU), or a graphics processing unit.

[0095] NPUs, such as NPU 908, are configured to accelerate the execution of common machine learning tasks, such as image classification, machine translation, object detection, and various other estimation models. In some examples, multiple NPUs may be instantiated on a single chip, such as a system on a chip (SoC), while in other examples, multiple NPUs may be part of a dedicated neural network accelerator.

[0096] The NPU can be optimized for training or inference, or in some cases configured to balance performance between the two. For NPUs that can perform both training and inference, the two tasks can generally still be performed independently.

[0097] NPUs designed to accelerate training are generally configured to accelerate the optimization of new models, which is a highly computationally intensive operation that involves inputting an existing data set (usually labeled or tagged), iterating on the data set, and then adjusting model parameters (such as weights and biases) to improve model performance. Generally speaking, optimization based on error estimates involves passing back through the layers of the model and determining the gradient to reduce the estimation error.

[0098] NPUs designed to accelerate inference are generally configured to operate on complete models. Thus, such NPUs can be configured to input new data slices and quickly process them through the trained model to generate model outputs (e.g., inferences).

[0099] In one implementation, the NPU 908 is part of one or more of the CPU 902 , the GPU 904 , and / or the DSP 906 .

[0100] In some examples, the wireless connectivity component 912 may include, for example, subcomponents for third generation (3G) connectivity, fourth generation (4G) connectivity (e.g., 4G LTE), fifth generation connectivity (e.g., 5G or NR), Wi-Fi connectivity, Bluetooth connectivity, and other wireless data transmission standards. The wireless connectivity component 912 is further connected to one or more antennas 914.

[0101] The processing system 900 may also include one or more sensor processing units 916 associated with any manner of sensors, one or more image signal processors (ISPs) 918 associated with any manner of image sensors, and / or a navigation processor 920, which may include satellite-based positioning system components (e.g., GPS or GLONASS) and inertial positioning system components.

[0102] The processing system 900 may also include one or more input and / or output devices 922 , such as a screen, a touch-sensitive surface (including a touch-sensitive display), physical buttons, speakers, microphones, and the like.

[0103] In some examples, one or more of the processors of processing system 900 may be based on the ARM or RISC-V instruction set.

[0104] The processing system 900 also includes a memory 924, which represents one or more static and / or dynamic memories, such as dynamic random access memory, flash-based static memory, etc. In this example, the memory 924 includes computer-executable components that can be executed by one or more of the aforementioned processors of the processing system 900.

[0105] In particular, in this example, memory 924 includes a channel representation estimation component 924A, an action taking component 924B, and a machine learning model component 924C. The depicted components, as well as other components not depicted, may be configured to perform various aspects of the methods described herein.

[0106] In general, the processing system 900 and / or its components may be configured to perform the methods described herein.

[0107] Notably, in other aspects, aspects of the processing system 900 may be omitted, such as where the processing system 900 is a server computer, etc. For example, in other aspects, the multimedia processing unit 910, the wireless connectivity component 912, the sensor processing unit 916, the ISP 918, and / or the navigation processor 920 may be omitted. Furthermore, aspects of the processing system 900 may be distributed, such as training a model and using the model to generate inferences, such as user-validated predictions.

[0108] Fig.10 Describes, for example, Figure 8 An example processing system 1000 is described for training a machine learning model to estimate channels in a spatial environment.

[0109] Processing system 1000 includes a central processing unit (CPU) 1002, which in some examples may be a multi-core CPU. Processing system 1000 also includes additional processing components tailored for specific functions, such as a graphics processing unit (GPU) 1004, a digital signal processor (DSP) 1006, and a neural processing unit (NPU) 1008. CPU 1002, GPU 1004, DSP 1006, NPU 1008, and input and / or output devices 1022 may be similar to those described above with respect to Figure 8 The CPU 1002 , GPU 1004 , DSP 1006 , NPU 1008 , and input and / or output devices 1022 discussed are similar.

[0110] In some examples, wireless connectivity component 1012 may include, for example, subcomponents for third generation (3G) connectivity, fourth generation (4G) connectivity (e.g., 4G LTE), fifth generation connectivity (e.g., 5G or NR), Wi-Fi connectivity, Bluetooth connectivity, and other wireless data transmission standards. Wireless connectivity component 1012 may be further connected to one or more antennas (not shown).

[0111] In some examples, one or more of the processors of processing system 1000 may be based on the ARM or RISC-V instruction set.

[0112] The processing system 1000 also includes a memory 1024, which represents one or more static and / or dynamic memories, such as dynamic random access memory, flash-based static memory, etc. In this example, the memory 1024 includes computer-executable components that can be executed by one or more of the aforementioned processors of the processing system 1000.

[0113] In particular, in this example, memory 1024 includes a dataset generation component 1024A and a model training component 1024B. The depicted components, as well as other components not depicted, may be configured to perform various aspects of the methods described herein.

[0114] In general, the processing system 1000 and / or its components may be configured to perform the methods described herein.

[0115] Sample Clauses

[0116] Specific implementation details for each aspect are described in the following numbered clauses.

[0117] Clause 1: A computer-implemented method comprising: estimating a representation of a channel using a machine learning model, wherein the machine learning model is trained to generate an estimated representation of the channel based on a position of a transmitter in a spatial environment, a position of a receiver in the spatial environment, and a three-dimensional representation of the spatial environment; and taking one or more actions based on the estimated representation of the channel.

[0118] Clause 2: A method according to clause 1, wherein the machine learning model is trained to generate an estimated representation of the channel based on the transmission of rays uniformly sampled on an ellipsoid representing the transmitter.

[0119] Clause 3: A method according to clause 1 or 2, wherein the machine learning model is trained to generate an estimated representation of the channel based on a set of rays, wherein the set of rays is selected based on the position of the transmitter and the position of the receiver in the spatial environment.

[0120] Clause 4: A method according to clause 3, wherein: the machine learning model includes a ray emission model configured to estimate the set of rays, and the ray emission model includes: an embedding model, which is trained to generate an embedding representing the relationship between the position of the transmitter and the position of the receiver in the spatial environment; an image prior model, which encodes the position of the transmitter and the position of the receiver in the spatial environment into line of sight information; and a ray emission module, which is trained to identify multiple rays to be emitted toward the receiver based on the embedding representing the relationship between the position of the transmitter and the position of the receiver in the spatial environment and the line of sight information.

[0121] Clause 5: A method according to clause 3 or 4, wherein the ray shooting model comprises a model trained to generate a probability heat map flattened on a two-dimensional grid on an ellipsoid around the location of the transmitter.

[0122] Clause 6: A method according to any one of clauses 1 to 5, wherein the machine learning model comprises: a first sub-model, the first sub-model being configured to estimate signal attenuation due to surface reflections in the spatial environment; and a second sub-model, the second sub-model being configured to estimate signal attenuation due to free space in the spatial environment.

[0123] Clause 7: The method of clause 6, wherein the first sub-model is further configured to estimate a distance that a signal propagates relative to a transmission source or a reflection point in the spatial environment.

[0124] Clause 8: A method according to any one of clauses 1 to 7, wherein: using the machine learning model to estimate the representation of the channel includes estimating a set of vectors representing each of a plurality of rays in the channel, each vector in the set of vectors representing the propagation of the ray over time, and each vector including positioning information, direction information, gain information, flight time information, and a validity score indicating whether the ray contributes to the channel.

[0125] Clause 9: The method of clause 8, wherein the corresponding set of vectors representing corresponding rays of the plurality of rays is terminated based on one of: the position of the ray intersects the position of the receiver or the power of the ray drops below a threshold.

[0126] Clause 10: A method as described in any of clauses 1 to 9, wherein the machine learning model comprises a model trained to minimize the LOS between the estimated state of the channel and the actual state of the channel over time.

[0127] Clause 11: The method of any of clauses 1 to 10, wherein taking one or more actions comprises generating a graphical rendering of the estimated representation of the channel in the three-dimensional representation of the spatial environment.

[0128] Clause 12: The method of any of clauses 1 to 11, wherein taking one or more actions comprises selecting one or more beams for communication between the transmitter and the receiver based on the estimated representation of the channel.

[0129] Clause 13: The method of any of clauses 1 to 12, wherein taking one or more actions comprises detecting the presence and location of an object in the spatial environment based on the estimated representation of the channel.

[0130] Clause 14: The method of any of clauses 1 to 13, wherein taking one or more actions comprises identifying material properties of the spatial environment based on the estimated representation of the channel.

[0131] Clause 15: A computer-implemented method comprising: generating a training data set comprising a plurality of channel representations, each channel representation comprising information for a plurality of rays corresponding to a path between a transmitter and a receiver in a spatial environment; and training a machine learning model to generate an estimated representation of the channel based on a three-dimensional representation of the spatial environment and the training data set.

[0132] Clause 16: A method according to clause 15, wherein: each of the multiple rays is represented by a set of vectors representing the propagation of the ray over time, and each vector includes one or more of positioning information, direction information, gain information, flight time information, or a validity score indicating whether the ray contributes to the channel.

[0133] Clause 17: The method of clause 16, wherein generating the training data set comprises generating a matrix over the plurality of rays using a mask configured to mask out vectors in the matrix associated with a validity score indicating that the ray does not contribute to the channel.

[0134] Clause 18: A method according to any one of clauses 15 to 17, wherein the machine learning model comprises: a first sub-model, the first sub-model being configured to estimate signal attenuation due to free space in the spatial environment; and a second sub-model, the second sub-model being configured to estimate signal attenuation due to reflection from a surface in the spatial environment.

[0135] Clause 19: A method according to any of clauses 15 to 18, wherein the machine learning model is also trained to generate a graphical rendering of the estimated representation of the channel in the three-dimensional representation of the spatial environment.

[0136] Clause 20: A method as recited in any one of clauses 15 to 19, wherein the estimated representation of the channel comprises a set of vectors associated with each ray of a plurality of rays whose contribution to the channel exceeds a threshold contribution.

[0137] Clause 21: A method as described in any of clauses 15 to 20, wherein training the machine learning model includes training the machine learning model to minimize the LOS between the estimated state of the channel and the actual state of the channel over time.

[0138] Clause 22: The method of clause 21, wherein the loss comprises a chamfer distance over time between an estimated state of the channel and an actual state of the channel.

[0139] Clause 23: A method according to any one of clauses 15 to 22, wherein the training data set comprises a training data set of rays uniformly sampled on an ellipsoid representing the transmitter.

[0140] Clause 24: The method of any one of clauses 15 to 23, further comprising: training a ray emission model to generate a set of rays based on a position of the transmitter and a position of the receiver in the spatial environment.

[0141] Clause 25: A method according to clause 24, wherein the ray emission model comprises: an embedding model, which is trained to generate an embedding representing the relationship between the position of the transmitter and the position of the receiver in the spatial environment; an image prior model, which encodes the position of the transmitter and the position of the receiver in the spatial environment into line of sight information; and a ray emission module, which is trained to identify the multiple rays to be emitted toward the receiver based on the embedding representing the relationship between the position of the transmitter and the position of the receiver in the spatial environment and the line of sight information.

[0142] Clause 26: The method of clause 25, wherein the ray shooting model comprises a model trained to generate a probability heat map flattened on a two-dimensional grid on an ellipsoid around the location of the transmitter.

[0143] Clause 27: A method according to clause 25 or 26, wherein the ray launching model comprises a model trained based on a collection of rays determined to have propagated to a location within a receiving sphere defined by a center point and a radius from the center point.

[0144] Clause 28: A method according to any one of clauses 15 to 27, wherein the training data set comprises channel representations associated with spatial environments having different surface attenuation and reflection characteristics.

[0145] Clause 29: A processing system comprising: a memory including computer executable instructions; and one or more processors configured to execute the computer executable instructions and cause the processing system to perform a method according to any one of clauses 1 to 28.

[0146] Clause 30: A processing system comprising: means for performing the method according to any one of clauses 1 to 28.

[0147] Clause 31: A non-transitory computer readable medium comprising computer executable instructions which, when executed by one or more processors of a processing system, cause the processing system to perform the method of any of clauses 1 to 28.

[0148] Clause 32: A computer program product embodied on a computer readable storage medium, comprising: code for performing the method according to any one of clauses 1 to 28.

[0149] Additional considerations

[0150] The previous description is provided to enable any person skilled in the art to practice the various embodiments described herein. The examples discussed herein are not limitations on the scope, applicability, or embodiments set forth in the claims. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the universal principles defined herein may be applied to other embodiments. For example, without departing from the scope of the present disclosure, the functions and arrangements of the elements discussed may be changed. Various examples may omit, replace, or add various procedures or components as appropriate. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. In addition, the features described for some examples may be combined in some other examples. For example, any number of aspects set forth herein may be used to implement a device or practice method. In addition, the scope of the present disclosure is intended to cover such devices or methods practiced using other structures, functionality, or structures and functionality that are supplementary or alternative to the various aspects of the present disclosure set forth herein. It should be understood that any aspect of the disclosure disclosed herein may be embodied by one or more elements of the claims.

[0151] As used herein, the word “exemplary” means “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects.

[0152] As used herein, a phrase referring to "at least one of" a list of items refers to any combination of those items (including single members). As an example, "at least one of a, b, or c" is intended to cover: a, b, c, ab, ac, bc, and abc, as well as any combination with multiple identical elements (e.g., aa, aaa, aab, aac, abb, acc, bb, bbb, bbc, cc, and ccc, or any other ordering of a, b, and c).

[0153] As used herein, the term "determining" encompasses a wide variety of actions. For example, "determining" may include calculating, computing, processing, deriving, investigating, searching (e.g., searching in a table, database, or other data structure), ascertaining, and the like. Furthermore, "determining" may include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), and the like. Furthermore, "determining" may include resolving, selecting, choosing, establishing, and the like.

[0154] The method disclosed herein includes one or more steps or actions for implementing the method. The steps and / or actions of the method can be interchangeable with each other without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and / or use of specific steps and / or actions can be modified without departing from the scope of the claims. In addition, the various operations of the method described above can be performed by any appropriate component capable of performing the corresponding function. The component may include various hardware and / or software components and / or modules, including but not limited to circuits, application specific integrated circuits (ASICs) or processors. Typically, in the case of operations illustrated in the accompanying drawings, those operations may have corresponding corresponding components with similar numbers plus functional components.

[0155] The following claims are not intended to be limited to the embodiments shown herein, but should be granted the full scope consistent with the language of the claims. Within the claims, unless specifically stated otherwise, the reference to the singular element is not intended to mean "one and only one", but "one or more". Unless otherwise specifically stated, the term "some" refers to one or more. No claim element should be interpreted according to the provisions of 35 U.S.C. § 112 (f), unless the phrase "part for..." is used to explicitly record the element, or in the case of a method claim, the phrase "step for..." is used to record the element. All structural and functional equivalents of the elements of various aspects described throughout the present disclosure that are known or will be known later to a person of ordinary skill in the art are expressly incorporated herein by reference and are intended to be covered by the claims. In addition, nothing disclosed herein is intended to be dedicated to the public, regardless of whether such disclosure is explicitly recorded in the claims.

Claims

1. A computer-implemented method comprising: estimating a representation of a channel using a machine learning model trained to generate an estimated representation of the channel based on a location of a transmitter in a spatial environment, a location of a receiver in the spatial environment, and a three-dimensional representation of the spatial environment; as well as One or more actions are taken based on the estimated representation of the channel.

2. The method of claim 1 , wherein the machine learning model is trained to generate an estimated representation of the channel based on transmitting rays uniformly sampled on an ellipsoid representing the transmitter.

3. The method of claim 1 , wherein the machine learning model is trained to generate an estimated representation of the channel based on a set of rays selected based on the position of the transmitter in the spatial environment and the position of the receiver in the spatial environment.

4. The method according to claim 3, wherein: The machine learning model includes a ray shooting model configured to estimate the set of rays, and The ray emission model includes: an embedding model trained to generate an embedding representing a relationship between the position of the transmitter in the spatial environment and the position of the receiver in the spatial environment; an image prior model, the image prior model encoding the position of the transmitter in the spatial environment and the position of the receiver in the spatial environment into line of sight information; and A ray launching module is trained to identify a plurality of rays to be launched toward the receiver based on the embedding representing the relationship between the position of the transmitter in the spatial environment and the position of the receiver in the spatial environment and the line of sight information.

5. The method of claim 3, wherein the ray-shooting model comprises a model trained to generate a probability heat map associated with an ellipsoid around the location of the transmitter flattened on a two-dimensional grid.

6. The method of claim 1, wherein the machine learning model comprises: a first sub-model configured to estimate signal attenuation due to surface reflections in the spatial environment; and a second sub-model configured to estimate signal attenuation due to free space in the spatial environment.

7. The method of claim 6, wherein the first sub-model is further configured to estimate a distance that a signal propagates relative to a transmission source or a reflection point in the spatial environment.

8. The method according to claim 1, wherein: Using the machine learning model to estimate the representation of the channel includes estimating a set of vectors representing each of a plurality of rays in the channel, each set of vectors representing propagation of the ray over time; and Each vector includes positioning information, direction information, gain information, time of flight information, and a validity score indicating whether the ray contributes to the channel.

9. The method of claim 8, wherein the corresponding set of vectors representing corresponding rays of the plurality of rays is terminated based on one of: the position of the ray intersects the position of the receiver, or the power of the ray drops below a threshold.

10. The method of claim 1, wherein the machine learning model is trained to minimize the LOS between an estimated state of the channel and an actual state of the channel over time.

11. The method of claim 1 , wherein taking the one or more actions comprises generating a graphical rendering of the estimated representation of the channel in the three-dimensional representation of the spatial environment.

12. The method of claim 1, wherein taking the one or more actions comprises selecting one or more beams for communication between the transmitter and the receiver based on the estimated representation of the channel.

13. The method of claim 1, wherein taking the one or more actions comprises detecting a presence and a location of an object in the spatial environment based on the estimated representation of the channel.

14. The method of claim 1, wherein taking the one or more actions comprises identifying material properties of the spatial environment based on the estimated representation of the channel.

15. A computer-implemented method comprising: generating a training data set including a plurality of channel representations, each channel representation including information for a plurality of rays corresponding to a path between a transmitter and a receiver in a spatial environment; as well as A machine learning model is trained to generate an estimated representation of a channel based on the three-dimensional representation of the spatial environment and the training data set.

16. The method of claim 15, wherein: Each ray of the plurality of rays is represented by a set of vectors representing the propagation of the ray over time, and Each vector includes one or more of positioning information, direction information, gain information, time of flight information, or a significance score indicating whether the ray contributes to the channel.

17. The method of claim 16, wherein generating the training data set comprises generating a matrix over the plurality of rays using a mask configured to mask vectors in the matrix associated with validity scores indicating that the ray does not contribute to the channel.

18. The method of claim 15, wherein the machine learning model comprises: a first sub-model configured to estimate signal attenuation due to free space in the spatial environment; and a second sub-model configured to estimate signal attenuation due to reflections from surfaces in the spatial environment.

19. The method of claim 15, wherein the machine learning model is further trained to generate a graphical rendering of the estimated representation of the channel in the three-dimensional representation of the spatial environment.

20. The method of claim 15, wherein the estimated representation of the channel comprises a set of vectors associated with each ray of the plurality of rays whose contribution to the channel exceeds a threshold contribution.

21. The method of claim 15, wherein training the machine learning model comprises training the machine learning model to minimize the LOS over time between an estimated state of the channel and an actual state of the channel.

22. The method of claim 21, wherein loss comprises a chamfer distance over time between the estimated state of the channel and the actual state of the channel.

23. The method of claim 15, wherein the training data set comprises a training data set of rays uniformly sampled on an ellipsoid representing the transmitter.

24. The method of claim 15, further comprising: A ray casting model is trained to generate a set of rays based on a position of the transmitter in the spatial environment and a position of the receiver in the spatial environment.

25. The method of claim 24, wherein the ray-casting model comprises: an embedding model trained to generate an embedding representing a relationship between the position of the transmitter in the spatial environment and the position of the receiver in the spatial environment; an image prior model, wherein the image prior model encodes the position of the transmitter in the spatial environment and the position of the receiver in the spatial environment into line of sight information; as well as A ray launching module is trained to identify the plurality of rays to be launched toward the receiver based on the embedding representing the relationship between the position of the transmitter in the spatial environment and the position of the receiver in the spatial environment and the line of sight information.

26. The method of claim 25, wherein the ray-shooting model comprises a model trained to generate a probability heat map associated with an ellipsoid around the location of the transmitter flattened on a two-dimensional grid.

27. The method of claim 25, wherein the ray launching model comprises a model trained based on a collection of rays determined to have propagated to a location within a receiving ellipsoid defined by a center point and a radius from the center point.

28. The method of claim 15, wherein the training data set includes channel representations associated with spatial environments having different surface attenuation and reflection characteristics.

29. A processing system comprising: A memory, wherein executable instructions are stored in the memory; and a processor configured to execute the executable instructions to cause the processing system to: estimating a representation of a channel using a machine learning model trained to generate an estimated representation of the channel based on a location of a transmitter in a spatial environment, a location of a receiver in the spatial environment, and a three-dimensional representation of the spatial environment; as well as One or more actions are taken based on the estimated representation of the channel.

30. A processing system comprising: A memory, wherein executable instructions are stored in the memory; and a processor configured to execute the executable instructions to cause the processing system to: generating a training data set including a plurality of channel representations, each channel representation including information for a plurality of rays corresponding to a path between a transmitter and a receiver in a spatial environment; as well as A machine learning model is trained to generate an estimated representation of a channel based on the three-dimensional representation of the spatial environment and the training data set.

Images