METHOD FOR CHANNEL ESTIMATING BETWEEN A TRANSCEIVER AND A MOBILE COMMUNICATION DEVICE

Description

[0001] The field of the invention is that of telecommunications between communicating entities in relative movement.

[0002] More particularly, the present invention relates to radio communications between a transmitter / receiver, such as a base station, and a connected moving vehicle.

[0003] Generally speaking, a radio signal transmitted by an antenna of a communicating entity, called an antenna signal, undergoes deformations depending on the propagation conditions between an origin point defined at the output of the origin antenna and a destination point defined at the input of an antenna of the recipient communicating entity. In order to limit these deformations, the antenna signal is previously distorted by applying pre-equalization coefficients depending on the characteristics of the propagation channel between these two antennas. It is therefore necessary to characterize this propagation channel.

[0004] Estimates of a radio communication channel are generally obtained by having one of the communicating entities send known reference signals, recording the signals received by the other communicating entity, and then calculating an estimate of the channel at the time and frequency of the reference signal. For example, in communication systems using OFDM modulation (initials of the English words « Orthogonal Frequency Division Multiplexing » meaning "Orthogonal Frequency Division Multiplexing"), the channel estimates take the form of complex gains for each subcarrier of the OFDM symbol on which the reference signals were transmitted.

[0005] Furthermore, communications with an increasing number of connected vehicles, particularly public transport vehicles, will become very expensive in terms of power and bandwidth, unless advanced communication methods are used. These will rely on channel estimations at the transmitter level (" Channel State Information at the Transmitter », or CSIT in English).

[0006] A first problem related to communications with connected vehicles concerns the power losses during the penetration of waves from the outside to the inside of the vehicle; in practice, these losses significantly affect the system's performance. More precisely, measurements have shown that penetration losses in a vehicle can be as high as 25 dB for user equipment receiving waves at 2.4 GHz frequency inside a van. Even higher losses are expected for higher frequencies, for example, for the bands around 3.5 GHz allocated to new generation mobile communication systems, and for certain vehicles well insulated from electromagnetic waves.

[0007] To solve this problem, it has been proposed (see the article by Y. Sui, A. Papadogiannis and T. Svensson entitled "The Potential of Moving Relays - A performance Analysis", IEEE Vehicular Technology Conference, Yokohama, May 2012) to equip the vehicle, for example a bus or a tram, with a mobile relay located outside the vehicle. This mobile relay uses an external antenna (typically, on the roof of the vehicle), which is connected to an antenna located inside the vehicle. The external antenna communicates with the base station (link " backhaut » in English), while the indoor antenna communicates with user devices using a local communication network (e.g. a WiFi network). The effectiveness of such a communication system then relies on the performance and reliability of the link between the relay and the base station.

[0008] A second problem with communications with connected vehicles is that the channel varies depending on the position of the external antenna. After an external antenna of the vehicle transmits a reference signal, the base station transmits the useful signal pre-equalized accordingly with a certain processing delay Δ t (of the order of 5 ms in the case of LTE (Long Term Evolution) systems for example); however, during this delay, the vehicle has moved, so that the channel between the base station and the external antenna is no longer the same as when the reference signal was transmitted. However, in the state of the art, it is possible to pre-equalize the useful signal for a given movement of the vehicle antenna, up to a certain maximum Δ x (typically, of the order of 0.3 carrier wavelengths), based on present and previous channel estimates and taking into account the classic phenomenon of rapid fading (" fast-fading » in English), that is, variations in the amplitude of the received signal caused by reflections of electromagnetic waves on objects close to the receiver. Unfortunately, at typical speeds in urban areas (around 50 km / h) and at typical carrier frequencies (around GHz), the vehicle moves more than Δ x in Δ t.

[0009] To solve this second problem, a process called "Predictive Antenna" has been proposed (see the article by M. Sternad, M. Grieger, R. Apelfröjd, T.

[0010] Svensson, D. Aronsson and A. Belén Martinez entitled “Using predictor antennas for long-range prediction of fast fading for moving relays”, IEEE Wireless Communications and Networking Conference, Paris, April 2012), which uses an additional antenna outside the vehicle, called a “Predictive Antenna”, placed in front of the external antenna, called the “main antenna”, dedicated to the exchange of useful data (“ payload » in English) with the base station. Estimating the baseband radio channel at the current position of the Predictive Antenna makes it possible to predict the channels that will concern the main antenna (therefore placed at the rear of the Predictive Antenna) when the main antenna reaches the current position of the Predictive Antenna.

[0011] Bjorsell et al., "Predictor antennas in action," 2017 IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), IEEE, October 8, 2017, presents an evaluation of the predictive antenna concept, used outside a vehicle, which solves the channel fading problem. This consists of an additional antenna in front of the others, whose estimated channel is a scaled prediction of the channels encountered by the rear antennas when they reach this position. This evaluation was carried out on a large set of noisy channel measurements in an urban environment. The predictive antenna concept allows an extension of the achievable prediction horizon, compared to Kalman or Wiener extrapolation based on past channel measurements.It represents an accuracy that would enable, for example, accurate multiple input - multiple output (MIMO) in the downlink direction for vehicles. The results confirm that the distribution of the NMSE (Normalized Mean Squared Error), over all propagation environments studied, is close to that obtained by correlation-based models and outperforms the use of past channel measurements.

[0012] The Predictive Antenna method thus advantageously makes it possible to predict channel estimates over durations that are an order of magnitude greater than those applicable on the basis of extrapolation of channel measurements in the past. This method is therefore of great interest for application to connected vehicles served by an advanced radio infrastructure.

[0013] For signal processing, as for the evaluation of a statistical model of the channel evolution in time and frequency, as well as for channel prediction, it is highly desirable to have a series of time samples that are regularly spaced in time. This is particularly the case for the Predictive Antenna method. However, the positioning of the reference signals with respect to time, frequency and antenna port is determined by the communication standard used, and by implementation choices. Unfortunately, in many communication systems, this positioning is not regular in time. This is particularly the case for Time Separation Duplex transmissions (" Time-Division Duplex » , or TDD in English), where the base station's estimation of the channels on the uplink (" uplink » in English) using reference signals also on the uplink is interrupted by transmission intervals on the downlink (' downlink » in English), during which the base station has no reference signal. It is therefore necessary, in these communication systems, to implement an interpolation method to obtain, from series of channel estimates spaced irregularly in time, a series of channel estimates located in time according to the needs, for example regularly spaced in time.

[0014] Various interpolation methods are known that are suitable for this purpose, for example those implementing high-order polynomial interpolation. However, the known methods have suboptimal and unsatisfactory properties when applied to channel estimations with estimation errors, i.e. to "noisy" data sets, as is the case for communications between a base station and a vehicle in an urban area.

[0015] The present invention therefore relates, according to a first aspect, to a channel estimation method, comprising, for at least one time difference observed between two sub-sequences of channel measurements, or channel estimations, consisting of scalars or complex vectors, the following steps: a first extrapolation is performed based on channel measurements or channel estimates of the subsequence preceding said time gap, moving forward in time, a second extrapolation is performed based on channel measurements or channel estimates of the subsequence following the time gap, moving backward in time, and a weighted average of said measurements or estimates extrapolated forward in time and said measurements or estimates extrapolated backward in time is calculated, to obtain channel measurements or channel estimates regularly spaced in the time gap.

[0016] Thus, the present invention proposes, for the purpose of performing channel estimation, a novel method for interpolating time series of channel measurements or channel estimations, for the purpose of filling abnormally long time gaps between transmissions of reference signals. The vector case concerns the application of the invention to carrier bytes, in which case the invention makes it possible to filter noise for a plurality of frequencies.

[0017] With these provisions, noisy and time-irregular series of channel measurements and channel estimates can be processed very efficiently, and channel predictions can be derived over considerably longer durations than the state of the art.

[0018] The method according to the invention may advantageously be implemented during a communication between, on the one hand, a network entity, and on the other hand a mobile communicating object comprising at least one Predictive Antenna and at least one main antenna.

[0019] According to particular characteristics, said first extrapolation, said second extrapolation and said weighted average are carried out by means of Kalman smoothing.

[0020] Thanks to these provisions, the final estimates obtained are optimal in the sense of minimizing the mean square error ( « Mean Square Error », or MSE in English) on the estimated quantities. Indeed, on the one hand, extrapolation forward in time, extrapolation backward in time, and the weighted average are all optimal in the MSE sense, and on the other hand the two extrapolations are statistically independent.

[0021] Correlatively, according to a second aspect, the invention relates to a network entity for channel estimation, comprising, for a given time difference between two sub-sequences of channel measurements, or channel estimations, consisting of scalars or complex vectors, means for: performing a first extrapolation based on channel measurements or channel estimates of the subsequence preceding said time gap, moving forward in time, performing a second extrapolation based on channel measurements or channel estimates of the subsequence following the time gap, moving backward in time, and calculating a weighted average of said measurements or estimates extrapolated forward in time and said measurements or estimates extrapolated backward in time, to obtain channel measurements or channel estimates regularly spaced in the time gap.

[0022] This network entity can advantageously be hosted in a physical or virtual network node (in the « cloud "), responsible for managing transmissions downlink, for example in a base station of a cellular communication network.

[0023] According to particular characteristics, said network entity further comprises means for carrying out said first extrapolation, said second extrapolation and said weighted average by means of Kalman smoothing.

[0024] The benefits offered by these network entities are essentially the same as those offered by the correlative processes briefly outlined above.

[0025] It should be noted that it is possible to implement these devices in the context of software instructions and / or in the context of electronic circuits.

[0026] According to a third aspect, the invention relates to a communications network. Said communications network is characterized in that it comprises: at least one network entity as briefly described above, and at least one mobile communicating object, equipped with at least one Predictive Antenna and at least one main antenna.

[0027] The advantages offered by this communication network are essentially the same as those offered by the correlative processes briefly set out above.

[0028] The invention also relates to a computer program downloadable from a communications network and / or stored on a computer-readable medium and / or executable by a microprocessor. This computer program is remarkable in that it comprises instructions for executing the steps of the channel estimation method briefly described above, when it is executed on a computer.

[0029] The advantages offered by this computer program are essentially the same as those offered by the said method.

[0030] Other aspects and advantages of the invention will become apparent upon reading the detailed description below of particular embodiments, given as non-limiting examples. The description refers to the accompanying figures, in which: there figure 1 schematically represents a classic channel estimation algorithm, and the figure 2 schematically represents an interpolation method according to the invention.

[0031] We will now describe an embodiment of the invention in which the algorithm described in the article by J. Björsell, M. Sternad and M. Grieger entitled "Predictor Antennas in Action" (IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, Montreal, October 2017) is used. This algorithm applies to the "Predictive Antenna" method described briefly above, and performs interpolation and noise filtering on measurements made on the basis of aperiodic signals (the reference signals).

[0032] The description below will refer to the figure 1 , which is adapted from a figure in this article.

[0033] For simplicity, we will assume here that we use a single Predictive Antenna, as well as a single main antenna (for the exchange of useful data with the base station), but this algorithm can easily be generalized to the case of a plurality of predictive and / or main antennas.

[0034] We will first summarize the main steps of this algorithm.

[0035] The value that the complex scalar channel will take h m seen by the main antenna τ seconds later is predicted by multiplying the channel h p previously estimated, and suitably delayed, seen by the Predictive Antenna, by a coefficient with complex values a h according to the following equation: h ^ m t + τ = a h h ^ p t + τ − Δ t , Or t is the current moment, ĥ p ( t + τ - Δ t ) is a smoothed estimate of h p obtained from noisy measurements made before the instant t, and Δ tis the time difference between the passages of the Predictive Antenna and the main antenna through the same spatial position.

[0036] The algorithm consists of two parts that interact with each other: a part that operates with a long time scale, called the "long-scale part," and a part that operates with a short time scale, called the "short-scale part." The short-scale part generates a new prediction at successive time intervals called "prediction intervals," with a short period. The long-scale part generates parameters at successive time intervals with a long period, a multiple of the short period.

[0037] The long scale part estimates the coefficient a h and speed v of the vehicle from buffered channel measurements of the Predictive Antenna and the main antenna. More precisely, the coefficient a h is evaluated from the maximum correlation between the antennas, and the speed v is evaluated from the value of the time shift between the antennas that maximizes the correlation between the antennas (the distance between the antennas being known in advance). In addition, the long-scale part determines the coefficients of a finite impulse response low-pass filter (" Finite Impulse Response ", or FIR in English); this FIR filter is applied in the short scale part to remove high frequency estimation errors on h m And h p . The values ​​of a h , v and FIR filter coefficients are used by the short-scale part during the subsequent prediction interval, after which new parameters are estimated by the long-scale part.

[0038] The short-scale part interpolates and filters the channels of all antennas, and then predicts the channel of the main antenna. Indeed, interpolation is necessary because it is assumed that the time series of channel estimates includes subsequences in which at least two channel estimates are available, but with temporal gaps between the subsequences. The interpolation provides a dense and time-uniform sequence of samples to which the last obtained FIR filter can be applied. The filtered channel estimates of the Predictive Antenna are then inserted into equation (1) to predict the value that the main channel will take τ seconds later. Finally, the filtered measurements for all antennas are copied into the buffer for use in the next iteration of the long scale part.

[0039] It should be noted, to complete this summary, that this algorithm requires an initiation phase (not shown in the figure 1 ), during which a long scale iteration is implemented to generate the first estimates based on unfiltered measurements extracted from the buffer.

[0040] We will now describe, with reference to the figure 2 , an interpolation method implemented in the present embodiment of the invention.

[0041] For a given time gap between two subsequences of channel estimates, we first extrapolate forward in time based on the channel estimates of the subsequence preceding the gap. We then extrapolate backward in time based on the channel estimates of the subsequence following the gap. Finally, we calculate a weighted average of the forward-extrapolated estimates and the backward-extrapolated estimates to obtain channel estimates that are evenly spaced across the time gap.

[0042] In a TDD system, clocked by time slots of equal duration, we can estimate the propagation channels relative to the time slots on the uplink ( uplink ) based on measurements (indicated by “x”s on the figure 2 ). Using channel reciprocity, we can then extrapolate the propagation channels relative to the time slots on the downlink ( downlink ) based on channel estimates of time slots uplink.

[0043] According to a variant, said extrapolations are carried out by means of a Kalman filter.

[0044] In this regard, it is recalled (see Wikipedia) that a filter that estimates the states of a dynamic system from a series of incomplete or noisy measurements is called a "Kalman filter". In the case of successive measurements, the Kalman filter is a recursive estimator with a time step: this means that, to estimate the current state, only the results of current measurements and the estimate of the previous state, carried out at the previous time step, are necessary; in other words, the history of measurements and estimates is not required. The Kalman filter uses the dynamics of the system considered, which defines its evolution over time: better data is thus obtained by eliminating the effect of noise. These data can be calculated for the past ("smoothing"), that is, for a past time step from current measurements and estimates carried out during a future time step, provided that these measurements and estimates are available.These data can also be calculated for the future ("prediction"), that is, for a future time step, from current measurements and estimates made during a past time step.

[0045] The state of the filter is represented by two variables which are updated at each iteration of the filter, that is to say at each time step: the estimate of the current state; and the error covariance matrix, which measures the accuracy of the estimated state.

[0046] The Kalman filter consists of two distinct phases: a prediction phase and an update phase. The prediction phase uses the estimated state at the previous time step, or the next time step, to produce an estimate of the current state. In the update stage, the current measurements are used to correct the predicted state, with the aim of obtaining a more accurate estimate.

[0047] It can be shown that, when the noise is Gaussian, the Kalman filter minimizes the mean square error on the estimated quantities; this is why the Kalman filter is said to be "optimal".

[0048] Note that the use of a Kalman filter to make a prediction on channel estimates is known per se (see for example the article by S. Kashyap, C. Mollén, E. Björnson and EG Larsson entitled “Performance Analysis of (TDD) Massive MIMO with Kalman Channel Prediction”, IEEE International Conference on Acoustics, Speech, and Signal Processing, New Orleans, March 2017).

[0049] In the present variant of the invention, the forward extrapolation in time is carried out by means of a Kalman filter; it is therefore an autoregressive time series model describing the evolution of the channel forward in time (it is recalled that in an autoregressive series, the current estimate of a quantity depends linearly on the previous estimates of this same quantity and on a stochastic term). The backward extrapolation in time is also carried out by means of a Kalman filter; it is therefore an autoregressive time series model describing the evolution of the channel backward in time. Finally, the weighted average of the two extrapolations is calculated by carrying out an optimal mean square error weighting, which takes into account the variances of the errors in the two extrapolations.

[0050] In a TDD system, time slots uplink And downlink constitute the time steps of the Kalman filter, whether it is used to predict (forward in time) or to smooth (backward in time). More precisely, for each time slot uplink, the channel is estimated as the optimal weighted sum in the MSE sense of the following two terms: the channel predicted by the Kalman filter on the basis of the estimate at the previous time step and the current measurement. We then calculate a first estimate of the channel of the m-th (where m ≥ 1) time slot downlink as the channel predicted by the Kalman filter after m time steps based on the last channel estimate uplink.

[0051] Thanks to the Predictive Antenna, we also have access to measurements and estimates in the future. With Kalman smoothing, the channel of each time slot uplink is estimated as the optimal weighted sum in the MSE sense of the following two terms: the current channel measurement, and the extrapolation one time step back by the Kalman filter based on the channel estimated at the subsequent time step. Let us denote by N the number of time slots downlink. We then calculate a second estimate of the channel of the m-th time slot downlink as the channel extrapolated back by the Kalman filter at the end of ( N - m ) backward time steps based on the first channel estimate uplink in the future.

[0052] Finally, the final channel estimate for the m -th time slot downlink is obtained by calculating the optimal weighted sum in the MSE sense of the first estimate and the second estimate of the channel m -th time slot downlink.

[0053] The invention can be implemented within entities, for example base stations, of communication networks, by means of software and / or hardware components.

[0054] The software components may be integrated into a conventional computer program for managing a network node. Therefore, as indicated above, the present invention also relates to a computer system. This computer system conventionally comprises a central processing unit controlling a memory by signals, as well as an input unit and an output unit. In addition, this computer system may be used to execute a computer program comprising instructions for implementing any of the channel estimation methods according to the invention.

[0055] Indeed, the invention also relates to a computer program downloadable from a communication network comprising instructions for executing the steps of a channel estimation method according to the invention, when it is executed on a computer. This computer program can be stored on a computer-readable medium and can be executable by a microprocessor.

[0056] This program may use any programming language, and may be in the form of source code, object code, or code intermediate between source code and object code, such as in a partially compiled form, or in any other desirable form.

[0057] The invention also relates to an information medium, irremovable, or partially or totally removable, readable by a computer, and comprising instructions of a computer program as mentioned above.

[0058] The information carrier may be any entity or device capable of storing the program. For example, the carrier may include a storage medium, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or a magnetic recording medium, such as a hard disk, or a USB flash drive (“ USB flash drive » in English).

[0059] On the other hand, the information carrier may be a transmissible carrier such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The computer program according to the invention may in particular be downloaded from a network such as the Internet.

[0060] Alternatively, the information carrier may be an integrated circuit in which the program is incorporated, the circuit being adapted to perform or to be used in the performance of any of the channel estimation methods according to the invention.

Claims

1. Computer-implemented radio communication channel estimation method comprising, for at least one time gap observed between two sub-sequences of radio communication channel measurements or of radio communication channel estimates, consisting of complex vectors or scalars, the following steps: - a first extrapolation is performed on the basis of radio communication channel measurements or radio communication channel estimates of the sub-sequence that precedes said time gap, going forward in time, - a second extrapolation is performed on the basis of radio communication channel measurements or radio communication channel estimates of the sub-sequence that follows the time gap, going backward in time, and - a weighted mean of said radio communication channel measurements or estimates extrapolated forward in time and of said radio communication channel measurements or estimates extrapolated backward in time is calculated so as to obtain radio communication channel measurements or radio communication channel estimates that are regularly spaced out over the time gap.

2. Channel estimation method according to Claim 1, characterized in that said first extrapolation and said second extrapolation are performed, and said weighted mean is taken, by means of Kalman smoothing.

3. Channel estimation method according to Claim 1 or Claim 2, characterized in that it is implemented during a communication between a network entity and a mobile communicating object comprising at least one Predictor Antenna and at least one main antenna.

4. Network entity for channel estimation, comprising, for a given time gap between two sub-sequences of radio communication channel measurements or of radio communication channel estimates, consisting of complex vectors or scalars, means for: - performing a first extrapolation on the basis of radio communication channel measurements or radio communication channel estimates of the sub-sequence that precedes said time gap, going forward in time, - performing a second extrapolation on the basis of radio communication channel measurements or radio communication channel estimates of the sub-sequence that follows the time gap, going backward in time, and - calculating a weighted mean of said radio communication measurements or estimates extrapolated forward in time and of said measurements or estimates extrapolated backward in time so as to obtain radio communication channel measurements or radio communication channel estimates that are regularly spaced out over the time gap.

5. Network entity according to Claim 4, characterized in that it further comprises means for performing said first extrapolation and said second extrapolation, and for taking said weighted mean, by means of Kalman smoothing.

6. Network entity according to Claim 4 or Claim 5, characterized in that it is accommodated in a base station of a cellular communication network.

7. Network entity according to any one of Claims 4 to 6, characterized in that performance of the first and / or the second channel estimation and calculation of said weighted mean are implemented during a communication between, on the one hand, said network entity, and on the other hand, a mobile communicating object comprising at least one Predictor Antenna and at least one main antenna.

8. Communication network, characterized in that it comprises: - at least one network entity according to any one of Claims 4 to 6, and - at least one mobile communicating object provided with at least one Predictor Antenna and at least one main antenna.

9. Irremovable, or partially or fully removable data storage means that includes computer program code instructions for carrying out the steps of a radio communication channel estimation method according to any one of Claims 1 to 3.

10. Computer program that can be downloaded from a communication network and / or is stored on a computer-readable medium and / or is able to be executed by a microprocessor, characterized in that it comprises instructions for carrying out the steps of a radio communication channel estimation method according to any one of Claims 1 to 3 when it is executed on a computer.