ANGLE INFORMATION ESTIMATION OF ULTRA-WIDEBAND WIRELESS SIGNALS

Abstract

The invention relates to a method and device for estimating angle information (50) of a received ultra-wideband wireless signal. Upon reception of a wireless signal emitted from a transmitting device (20) with known sounding sequence, the receiving device (10) estimates the channel impulse response (CIR), selects a portion of the channel impulse response (CIR), and estimates angle information (50) given the angle-dependent antenna transfer functions of either the transmitting device (20), the receiving device (10), or both. For this, the selected portion of the channel impulse response of the signal is fed into a neural network (73) which outputs an angle information probability distribution for the ultra-wideband wireless signal (50).

Claims

1. (canceled)

2. A method comprising: receiving, from at least one receiver antenna of a receiving device, a radio signal; deriving a channel impulse response of a propagation channel of the received radio signal or an envelope function indicative of an envelope of the channel impulse response; deriving state information of the receiving device; and deriving, by inputting the channel impulse response or the envelope function into an angle estimator, an angle information probability distribution for the radio signal, wherein the state information is also input into the angle estimator to improve the accuracy of the angle information probability distribution.

3. The method of claim 2, wherein the angle estimator comprises a neural network.

4. The method of claim 2, wherein the state information is derived using sensor information received from a sensor.

5. The method of claim 2, wherein the radio signal comprises an ultra-wideband radio signal.

6. The method of claim 2, further comprising deriving a path location within the channel impulse response or the envelope function, wherein deriving the angle information probability distribution is based on the path location.

7. The method of claim 2, further comprising deriving a path window of the channel impulse response or the envelope function, wherein deriving the angle information probability distribution is based on the path window.

8. The method of claim 2, wherein the at least one receiver antenna has an angle-dependent transfer function.

9. The method of claim 2, further comprising: deriving timing information from the radio signal; deriving, using the timing information, a time-of-flight estimate of the received radio signal; and inputting the time-of-flight estimate into the angle estimator for deriving the angle information probability distribution.

10. The method of claim 2, wherein the radio signal comprises data indicative of a transmitter antenna transfer function of a transmitting device of the radio signal, the method further comprising: inputting the data indicative of said transmitter antenna transfer function into the angle estimator for deriving the angle information probability distribution.

11. The method of claim 2, wherein: the radio signal is received by at least two antennas of the receiving device; and the method comprises: measuring a phase difference of the radio signal as received by the at least two antennas; and combining the measured phase difference with the derived angle information probability distribution.

12. An apparatus comprising: a transceiver with at least one receiver antenna structured for receiving a radio signal; a control unit structured for: deriving a channel impulse response of a propagation channel of the received radio signal or an envelope function indicative of an envelope of the channel impulse response; deriving state information of the apparatus; and deriving, by inputting the channel impulse response or the envelope function into an angle estimator, an angle information probability distribution for the radio signal, wherein the state information is also input into the angle estimator to improve the accuracy of the angle information probability distribution.

13. The apparatus of claim 12, wherein the angle estimator comprises a neural network.

14. The apparatus of claim 12, further comprising a sensor structured to provide sensor information, wherein the state information is derived using the sensor information.

15. The apparatus of claim 12, wherein the radio signal comprises an ultra-wideband radio signal.

16. The apparatus of claim 12, wherein: the control unit is further structured to derive a path location within the channel impulse response or the envelope function; and deriving the angle information probability distribution is based on the path location.

17. The apparatus of claim 12, wherein: the control unit is further structured to derive a path window of the channel impulse response or the envelope function; and deriving the angle information probability distribution is based on the path window.

18. The apparatus of claim 12, wherein the at least one receiver antenna has an angle-dependent transfer function.

19. The apparatus of claim 12, wherein the control unit is further structured for: deriving timing information from the radio signal; deriving, using the timing information, a time-of-flight estimate of the received radio signal; and inputting the time-of-flight estimate into the angle estimator for deriving the angle information probability distribution.

20. The apparatus of claim 12, wherein: the radio signal comprises data indicative of a transmitter antenna transfer function of a transmitting device of the radio signal; and the control unit further structured for inputting the data indicative of said transmitter antenna transfer function into the angle estimator for deriving the angle information probability distribution.

21. The apparatus of claim 12, wherein: the radio signal is received by at least two antennas of the apparatus; and the control unit is further configured for: measuring a phase difference of the radio signal as received by the at least two antennas; and combining the measured phase difference with the derived angle information probability distribution.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The invention will be better understood and objects other than those set forth above will become apparent when consideration is given to the following detailed description thereof. This description makes reference to the annexed drawings, wherein:

[0059] FIG. 1 shows the composition of an example channel impulse response (CIR) 61 with marked first path location 62 and first path window 63.

[0060] FIG. 2 shows a block diagram of an ultra-wideband wireless (UWB) angle estimator 40 according to the invention comprising a receiver antenna 11 and a receiving device 10.

[0061] FIG. 3 shows a flow chart depicting the operation of the angle estimator 40 of FIG. 2 to estimate the angle information 50.

[0062] FIG. 4 shows a block diagram of the a-posteriori angle information probability estimator 72 as implemented in the angle information estimator 45 of FIG. 3, and how it can be trained using a supervised learning approach 74.

[0063] FIG. 5 shows an implementation of the neural network 73 of FIG. 4, having the first path window 63 of the envelope function 64 as input, and the predicted angle information probability distribution 50 as output.

[0064] FIG. 6 shows on the left a receiving device 10 according to the invention equipped with two receiver antennas 11, 12, with a distance of d=, where is the wavelength of the carrier of the UWB signal 60. The measured phase difference corresponds to several possible angle of arrivals, which can be disambiguated by the estimated angle information 50 as shown on the right.

[0065] FIG. 7 shows block diagrams of four different wireless networks, each comprising at least one receiving device 10, 30, at least one transmitting device 20, 30, or at least one receiving and transmitting device 30 according to the invention.

[0066] FIG. 8 shows two block diagrams of radar applications comprising a receiving device 10 and a transmitting device 20 or a receiving and transmitting devices 30, respectively, according to the invention.

DETAILED DESCRIPTION

[0067] Assuming a cascaded, linear and time invariant model for a wireless ultra-wideband propagation channel with N.sub.MP multipath components, its channel impulse response (CIR) 61 in time domain h.sub.CIR can be written as

[00001]$\begin{array}{cc}{h}_{\mathrm{CIR}}={.Math.}_{n=1}^{N_{MP}}{h}_{\mathrm{tx}}\left({}_{\mathrm{tx},n}\right)*h_{env,n}*h_{rx}\left({}_{\mathrm{rx},n}\right)& \left[\mathrm{Eq}.1\right]\end{array}$

where h.sub.tx(.sub.tx,n) and h.sub.rx(.sub.rx,n) are the AOD .sub.tx,n, AOA .sub.rx,n dependent impulse responses of the transmitter 21 and receiver antenna 11, respectively, where h.sub.env,n is the impulse response of the environment for the n-th multipath component, and where * is the convolution operator. This is visualized in FIG. 1.

[0068] The effects of electromagnetic wave interactions with material belonging to the same rigid body as the antennas are included in the impulse response of the transmitter 21 or receiver 11 antenna to simplify terminology. Hence an AoA- or AoD-dependent antenna transfer function can be a result of either a non-isotropic antenna or by material belonging to the same rigid body as the antenna, which acts similarly to parasitic elements in multi-element antennas (as for example used in multi-element Yagi antennas).

[0069] So far, the focus of antenna engineers has been on designing either directional or omnidirectional antennas showing a similar transfer function for all angles with a significant gain. Furthermore, the antennas have been mounted on devices such that their radiation characteristics were not influenced by the devices. This allows to achieve a constant antenna transfer functions for all angles with a high antenna gain. Instead, the method presented herein suggests to amplify the angle dependency of the antenna transfer function in order to infer the AoA, AoD from the CIR 61 estimate.

[0070] From Eq. 1 it is visible that in case of a multi-path transmission channel, the measured CIR 61 is dependent on the transfer function of multiple AoA and AoD as shown in FIG. 1 for a propagation channel having a direct and an indirect path. In order to limit the effect of AoA or AoD transfer functions not corresponding to the desired first path (note that for radar applications the first path corresponds to the path of the signal directly reflected back from the target object), only a first path window 63 of the CIR 61 containing the desired first path location 62 is taken into consideration as highlighted in FIG. 1. This window size must be adapted to the pulse duration, the center frequency of the ultra-wideband (UWB) signal, and the environment for best performance.

[0071] Antenna transfer function including the effects of objects in its vicinity are difficult to physically model and the transfer function of the environment is often unknown.

[0072] Additionally, the antenna transfer function might be similar for two different angles, therefore it is difficult to find an analytic formula mapping a measured CIR to an AoA, AoD. Instead machine learning tools enable to acquire data-driven models providing probability distributions as their outputs, as will be shown below.

[0073] In consideration of this knowledge and the complex, costly and power-hungry hardware of state of the art AoA, AoD estimation approaches, the present invention provides an AoA, AoD estimation method having a simpler hardware configuration, enabling smaller, low-power devices to estimate the angle information using less actively mechanically or electronically controlled elements.

[0074] The angle estimator 40 according to the invention to estimate angle information from a UWB signal comprises the following components, which are shown in FIG. 2: [0075] At least one receiver antenna 11 with an angle-dependent transfer function. Alternatively, electromagnetic wave reflective or absorptive material can be placed on the same rigid body and in the vicinity of the antenna 11 (as is often the case when the antenna is integrated in a device or mounted on a device). This material will also lead to angle-dependent antenna transfer function. [0076] A transceiver 41 in a receiving device 10 reading the signal from the receiver antenna 11. [0077] A control unit 43 comprising memory and a processor for storing and executing software in the memory and controlling the angle estimator 40 accordingly. [0078] A CIR estimator 44 for deriving a CIR response 61 of a propagation channel of said received ultra-wideband radio signal 60 or an envelope function 64 indicative of an envelope of said channel impulse response 61. [0079] An angle information estimator 45 deriving, using said CIR 61 or said envelope function 64, an angle information probability distribution 50 for said ultra-wideband radio signal 60.

[0080] These core components together with components used in other realizations will be discussed in detail in the following paragraphs and in FIG. 3 and FIG. 4. Note that each estimator and control module as shown in FIG. 2 contains a database that may store configuration information, and logging data necessary for its operation.

[0081] Also note that at least some or all of the components as shown and discussed herein separately can be implemented in software running on the control unit 43.

[0082] The transmitter antenna 21 emits a sounding sequence contained in the UWB radio signal 60 and known to the receiving device with a certain angle-dependent transfer function. This signal excites the receiver antenna 11 with a certain angle-dependent transfer function. The transceiver 41 samples this excitation, which allows the CIR estimator 44 to estimate the CIR 61 oralternatively or in additionan envelope function 64 of the CIR. This CIR estimate 61 is then fed to the angle information estimator 45 which is shown in FIG. 3. A leading-edge detections algorithm implemented in the first path detector 70 finds the first path location 62 in the CIR 61. Samples in a window 63 around this first path location 62 are fed to an a-posteriori angle information probability distribution estimator 72. This estimator outputs the a-posteriori angle-information probability distribution 50, that the signal was received with a certain AoA, AoD

p(.sub.tx,.sub.rx|h.sub.CIR)

[0083] Note that in case only the AoA is to be estimated, i.e. the probability distribution p, it is beneficial to use a transmitter antenna 21 with an isotropic transfer function on the transmitting device 20, while a receiver antenna 11 with an angle-dependent transfer function must be used on the receiving device 10.

[0084] Note that in case only the AoD is to be estimated, i.e. the probability distribution p, it is beneficial to use an receiver antenna 11 with an isotropic transfer function on the receiving device 10, while a transmitter antenna 21 with angle-dependent transfer function must be used on the transmitting device 20.

[0085] Note that in case that either a state information estimator 47 or a timing information estimator 46 is able to provide an estimate of the time-of-flight (ToF) 51 of the received signal, it is beneficial to condition the probability distribution not only on the window 63 of the estimated CIR, but also on the estimated ToF t 51, i.e.

p(.sub.tx,.sub.rx|h.sub.CIR,T).

[0086] Note that in case transmit antenna transfer function information 56 is transmitted with the ultra-wideband radio signal 60, it is beneficial to condition the probability distribution not only on the window 63 of the estimated CIR, and the estimated ToF 51 (if available), but also on the antenna transfer function information 56 indicative of the transmitter antenna impulse response function h.sub.tx (.sub.tx), i.e.

p(.sub.tx,.sub.rx|h.sub.CIR,,h.sub.tx(.sub.tx)),

or by deconvoluting the channel impulse response 61 first with the transmitter antenna transfer function information.

[0087] The a-posteriori angle information probability distribution estimator 72 is based on a neural network 73 and outputs either a discretized probability distribution, or the parameters of a continuous probability distribution as visualized in FIG. 4. The inputs to the neural network 73 are the CIR samples contained in the first path window 63, as well as the ToF 51 and/or state information 53 based on sensor information 54, if available.

[0088] If a discretized probability distribution is used, the angular information probability distribution is represented by N.sub.bin bins, each representing the probability that the AoA, AoD take a certain value. Denoting with zR.sup.N.sup.bin the unnormalized log probabilities of each bin outputted by the neural network 73, the normalized probability that a signal's angle information belongs to bin i{1, 2, . . . , N.sub.bin} is

[00002]$p\left(\mathrm{bin}=i|h_{\mathrm{CIR}},\right)=\exp \left(z\left[i\right]\right)/\left({.Math.}_{j=1}^{N_{bin}}\exp \left(z\left[j\right]\right)\right).$

[0089] Such a neural network 73 outputting a discretized probability distribution is visualized in FIG. 5. The samples of the envelope function 64 falling into the first path window 63 are fed as magnitude and phase to the input layer of the neural network 73, and the information is then passed through three hidden layers comprising 800, 400, and 400 neurons, each, to the output layer of the neural network 73.

[0090] A person with ordinary skill in the art will understand that the visualized fully connected neural network can be replaced convolutional neural network, a complex neural network, or any other parametrization suitable for training with a supervised learning algorithm.

[0091] If a continuous probability distribution is used, the neural network 73 is trained to output a parameterization of a selected distribution. If either the AoA or the AoD are estimated, a mixture of K von Mises-Fisher distributions is used. In case both AoA and AoD are estimated, a mixture of K bivariate von Mises-Fisher distributions is used. K must be chosen based on the selected antennas and the environment.

[0092] The neural network 73 is trained using a supervised learning algorithm 74. The training data consist of angle information-CIR window pairs and are augmented by ToF data in case these are available. Denoting with {acute over ()}.sub.rx the target AoA 55, the cross-entropy loss J per training data which the supervised learning framework is trying to minimize over all training data is given as

[00003]$J=-\log p\left({}_{rx}={\stackrel{}{}}_{rx}|h_{\mathrm{CIR}},\right).$

[0093] This is done accordingly also for the case where the AoD, or both the AoA and AoD distributions are estimated and similarly if the probability distribution is also conditioned on the ToF or the transmitter antenna transfer function information.

[0094] Note that this supervised learning algorithm 74 can also be run during operation if the target angle information 55 is supplied by a state information estimator 47 in form of the predicted angle information.

[0095] Even though the timing information estimator 46 is optional for the device to estimate the angle information 50 from a wireless UWB signal 60, it enhances its capabilities. By estimating the time of arrival (ToA) of a received signal and by scheduling the time of departure (ToD) of a signal which is to be transmitted, the ToF between two devices can be estimated. To this end, the transceiver control software 48 must control the transceiver in accordance with a two-way ranging protocol if the clocks on the transceivers are not synchronized. In case of radar applications, these ToD and ToA estimates enable to estimate the distance of the transmitter-object-receiver path. These ToF estimates can be used as inputs to the angle information estimator 45, and as inputs to the state information estimator 47. Even if only (ToA) estimates are at hand, they can still be used as inputs to the state information estimator 47 which in turn can also provide a ToF estimate to the angle information estimator 45.

[0096] The state information estimator 47 estimates the pose and velocity of the device by fusing [0097] angle information 50 obtained from multiple transmitters with known locations (and orientation in case the AoD information is used), [0098] timing information in case a timing information estimator 46 is embodied in the device, [0099] and information from any other sensor 54 embodied in the device, e.g. an inertial-measurement unit [0100] with a motion model (if one is available) to produce an estimate of the pose and velocity of the device.

[0101] In one embodiment of the method, different center frequencies of the ultra-wideband communication channel are employed by the transmitting 20 and receiving device 10 and an a-posteriori angle information probability distribution estimator 72 is trained for each. By fusing the a-posteriori angle information probability distributions 50, ambiguities due to similar antenna transfer function for two different angles for a certain center frequency can be resolved.

[0102] In another embodiment of the method, not the channel impulse response 61, but its complex envelope 64 is fed either as complex numbers, or as magnitude and phase to the neural network 73. A person with ordinary skill in the art will understand that other characteristics of the estimated CIR 61 can be used as inputs to the neural network 73 as long as they are correlating with the antenna transfer function for different angles.

[0103] In case that a multi-band communication channel is used for communication, the necessary bandwidth to extract angle information 50 from the measured signal can be achieved by combining the CIR estimates 61 of each band.

[0104] In case that the transmitter and the receiver employ incoherent clocks, the CIR estimates 61 over different signals can be acquired. Upon each signal reception, the CIR 61 is sampled at a different time. This enables the accumulation of a CIR with a higher resolution. The performance of the presented device can be increased if such high-resolution CIR estimates are used as inputs to the a-posteriori angle information probability distribution estimator. Alternatively, instead of accumulating a high-resolution CIR over different signals, also the a-posteriori probability distributions can be accumulated and fused into one final a-posteriori distribution, which also leads to an increased performance of the presented device.

[0105] In case the receiving device 10 has multiple receiving antennas 11, 12, the corresponding angle probability distributions 50 can be combined with phase difference measurements of the different antennas. This allows to resolve phase ambiguities as illustrated in FIG. 6, where a setup with two receiver antennas 11, 12 is shown, the first receiver antenna 11 being placed one wavelength (of the carrier of the UWB signal 60) apart from the second receiver antenna 12.

[0106] The presented method to estimate angle information can be used in multiple setups as visualized in FIG. 7. A receiving device 10 can estimate its relative orientation with respect to a transmitting device 20, or multiple receiving devices 10 can localize a transmitting device 20 by fusing the estimated angle information 50

[0107] In another implementation, a receiving device 10 can localize itself by fusing angle information 50 acquired from multiple transmitting devices 20. In case all modules in a networked system employ the proposed method and device, the complete network structure can be estimated.

[0108] In another embodiment of the invention, the AoA of a reflected UWB wireless signal is estimated. This enables to localize objects in radar applications if combined with ToF measurements, as visualized in FIG. 8.

[0109] The present invention may be practiced as a method or device adapted to practice the method. It is understood that the examples in this application are intended in an illustrative rather than in a limiting sense. In accordance with the present disclosure, limitations of current systems for localizing have been reduced or eliminated. While certain aspects of the present invention have been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the following claims. It will also be understood that the components of the present disclosure may comprise hardware components or a combination of hardware and software components. The hardware components may comprise any suitable tangible components that are structured or arranged to operate as described herein. Some of the hardware components may comprise processing circuitry (e.g., a processor or a group of processors) to perform the operations described herein. The software components may comprise code recorded on tangible computer-readable medium. The processing circuitry may be configured by the software components to perform the described operations. It is therefore desired that the present embodiments be considered in all respects as illustrative and not restrictive.