WO2010001376A1

WO2010001376A1 - Signal strength prediction in mobile devices

Info

Publication number: WO2010001376A1
Application number: PCT/IE2009/000042
Authority: WO
Inventors: Eamonn O'nuallain; Gregor Gaertner
Original assignee: Provost Fellows And Scholars Of The College Of The Holy And Undivided Trinity Of Queen Elizabeth Near Dublin
Priority date: 2008-07-01
Filing date: 2009-07-01
Publication date: 2010-01-07

Abstract

A method of predicting electromagnetic signal strength into the future between a mobile device and a base station is described. A processor of the mobile device monitors the signal strength and executes a signal strength predictor on the basis of a shadowing signal of the electromagnetic signal and on the basis that the shadowing signal is self-correlated. The signal strength predictor is executed without using topological data. A dynamic decision is made between optimal and sub-optimal signal strength predictors, the sub-optimal predictors being faster to execute. Factors such as predictive accuracy requirements and whether the network is infrastructure-based or is infrastructureless are taken into account in the decision. The processor executes the signal strength predictor using as inputs current signal strength, speed of at least one device, and autocovariance data of the shadowing signal. The autocovariance data of the shadowing signal may comprises an autocorrelation function governing self-correlation of the shadowing signal or de-correlation time of the shadowing signal.

Description

"Signal Strength Prediction in Mobile Devices"

INTRODUCTION

Field of the Invention

The invention relates to handover of call signalling between mobile devices and base transceiver stations. More specifically, the invention relates to predicting with a high degree of accuracy the strength of an electromagnetic signal into the future for moving devices, such as a pedestrian walking with a mobile phone or somebody in a vehicle working on a computer connected to a wireless network.

Prior Art Discussion

The value of such fore-knowledge allows choice of a vacant channel during link setup or jumping 'in session' to a channel which has better signal stability characteristics, thereby improving communications quality. The reason for this is that better communications quality and data rates are obtained on a more reliable link channel since signal strength is a primary determinant of these. This is however likely to change 'in session' due to change in location where channel occupancy and/or propagation conditions vary.

Hitherto such predictive techniques and variations thereof have largely been based either explicitly or implicitly on the 'path loss'. The path loss is an estimate of the degradation of the signal with distance from the transmitter. In its simplest form it is a function that takes the form

n

where r is the distance from the transmitter and n is either a deterministic or empirically obtained constant.

There are, however, more sophisticated variations of the path loss function but all are based on this central form. A problem with using path loss as a basis for link quality prediction is that the electromagnetic signal does not degrade smoothly with distance from the transmitter and in fact exhibits significant variation largely due to shadowing effects by buildings, hills, trees and the like. The resulting undulating signal is referred to here as the 'Shadowing Signal' for the purposes of this description. It is otherwise known as the 'Large-Scale Fading Signal'. Foreknowledge of the shadowing signal would greatly improve communications quality in all types of wireless communications networks and is used in the planning stages of 2G and 3G cellular networks (i.e. where to best locate base-stations). For in-session link quality prediction for the operation of 2G and 3 G infrastructure-based wireless networks the shadowing signal has been regarded as something that can only be determined to any reasonable degree of accuracy using memory- consuming topological data and is thus computationally expensive to compute. Since link session setup or change must take place in real-time or near real-time, computing the shadowing signal for the purposes given above is not an option for link quality prediction. As a result there are excessive dropped calls, outages, noisy links, and poor data rates.

In the advent of research into ad-hoc or infrastructureless networks, link quality prediction has become a topic of more research interest because it is has become regarded as an important facet of session routing. Most approaches base, either implicitly or explicitly, their predictions on the path loss function. Again, the shadowing signal has been regarded as something that can only be determined to any reasonable degree of accuracy using memory-consuming topological data and is computationally expensive to compute. However path loss-based predictors are insufficient for reliable communications using these networks because of shadowing effects. Augmenting such predictors with stochastic realizations of the shadowing effect serve only to realize error margins and in themselves do not improve communications quality.

In cognitive radio technology, currently in its nascent stage, link quality prediction on the basis of signal strength seems to have received no attention at all. Since in these networks there are likely to be many channels to choose from, reliable link quality prediction would be very valuable in these networks. It is also noteworthy that it is envisaged that cognitive radio technology will incorporate ad-hoc networking.

The invention is therefore directed towards providing improved signal strength prediction in networks. SUMMARY OF THE INVENTION

According to the invention there is provided a method of predicting electromagnetic signal strength into the future between devices having wireless communications interfaces and which are relatively moving, the method being implemented by at least one of said devices, said device having a processor, the method comprising the steps of:

monitoring the signal strength at one or both devices, and

executing a signal strength predictor on the basis of a shadowing signal of the electromagnetic signal and on the basis that the shadowing signal is self-correlated.

In one embodiment, the signal strength predictor is executed without using topological data.

In one embodiment, a dynamic decision is made between optimal and sub-optimal signal strength predictors, the sub-optimal predictors being faster to execute.

In one embodiment, the decision is made on the basis of predictive accuracy requirements.

hi one embodiment, the decision is made on the basis of whether the network is infrastructure- based or is infrastructureless.

In one embodiment, the decision is made on the basis of computational capability of the processor.

hi one embodiment, the decision is made on the basis of bandwidth limitations on communication between the devices,

hi one embodiment, the decision is made on the basis of memory capacity of the processing device.

hi one embodiment, the devices are a mobile device and a base station transceiver. In one embodiment, the processor executes the signal strength predictor using as inputs current signal strength, speed of at least one device, and autocovariance data of the shadowing signal.

hi one embodiment, the autocovariance data of the shadowing signal comprises an autocorrelation function governing self-correlation of the shadowing signal.

hi one embodiment, the autocovariance data comprises de-correlation time of the shadowing signal.

In another embodiment, the autocovariance data comprises de-correlation distance of the shadowing signal.

In one embodiment, the de-correlation distance is of the order of the size of shadowing objects or clusters of such objects.

In one embodiment, the de-correlation distance comprises empirical data obtained from a device or from a locally-hosted database.

In one embodiment, for a sub-optimal predictor, the processor assumes constant speed of the device.

In one embodiment, for a sub-optimal predictor, the processor performs estimations to determine the autocovariance data.

In one embodiment, the processor performs fast matrix inversion techniques to determine autocovariance data.

hi one embodiment, the processor simplifies an autocovariance matrix of the shadowing signal, and uses sparse matrix processing techniques to process said matrix.

In a further embodiment, the processor approximates the autocovariance matrix to an identity matrix, providing for scalar processing instead of matrix processing. In one embodiment, the processor uses the output of the signal strength predictor to estimate the quality of a channel before deciding to use it.

hi one embodiment, the processor uses the output of the signal strength predictor to perform preemptive channel hopping on anticipation of a degradation in link quality.

hi one embodiment, the processor uses the output of the signal strength predictor to perform handover.

hi one embodiment, the processor uses the output of the signal strength predictor to perform preemptive routing in an infrastructureless network.

In one embodiment, the processor uses the predictions to build a space domain radio emissions map.

hi another aspect, the invention provides a communication circuit comprising a processor adapted to perform any method defined above. In a further aspect, the invention provides a mobile device comprising such a communication circuit.

hi a further aspect, the invention provides a computer readable medium having software code embodied thereon, said code being adapted to implement any method defined above when executed by a digital processor of a mobile device.

DETAILED DESCRIPTION OF THE INVENTION

Brief Description of the Drawings

The invention will be more clearly understood from the following description of some embodiments thereof, given by way of example only with reference to the accompanying drawings in which:-

Fig. 1 is a flow diagram illustrating a signal strength prediction method of the invention.

Description of the Embodiments - O -

Referring to Fig. 1 a signal strength-based predictor operates in the time domain where there is relative movement between communicating nodes, in one embodiment a mobile device and a base station. The processor can reliably predict signal strength a few seconds into the future - more than enough time to estimate the quality of the link.

Foreknowledge of signal strength conditions can be used by the mobile device: a) to estimate the quality of a channel before deciding to use it, b) to perform pre-emptive channel hopping on anticipation of a degradation in link quality, c) as a basis for powerful handoff/handover algorithms in wireless networks, and d) to perform optimal (on terms of link quality) or pre-emptive routing in infrastructureless networks.

A "mobile device" is intended in this specification to mean any portable device with the ability to communicate and process data, for example a mobile phone or a laptop computer.

The mobile device can use the predictions to build a space domain radio emissions map. It does not require topological data on which to base its predictions. The predictive method uses the sequence of signal-power measurements that are available (as obtained by continuous sampling and storing by the mobile device) at the time a prediction is made on which to base its prediction. The predictive ability stems from: i) Identifying that the shadowing signal and not the path-loss of the (fading) signal is the primary determinant of accurate link quality prediction, ii) Knowing that the shadowing signal is self-correlated (has an autocorrelation function) over a period of a few seconds, iii) Obtaining a statistically optimal signal strength predictor for moving nodes without using topological data, iv) Deriving a sub-optimal predictor that can be implemented very quickly or in real-time based on iii) even by a simple device. In our experiments with 802.11 systems, results given by the fastest sub-optimal predictor are very close to those given by the optimal predictor.

There are three components to an electromagnetic signal that can be used to perform predictions. These are: the path-loss, the shadowing signal (or large-scale fading signal), and the small scale- fading signal. Like the shadowing signal, the small-scale fading signal is also self-correlated (has an autocorrelation function). Its predictive properties are however very poor. The path-loss function has predictive value only in the vicinity of the transmitting node- in which case signal strength prediction is not a problem for which a solution is needed because the signal strength will be high. Outside of this vicinity the shadowing signal has the greater impact on link quality and greater predictive ability due to the nature of its autocorrelation function. This is why the algorithms below are far superior to the commonly used path-loss based algorithms.

The benefits of the fact that the shadowing signal is self-correlated are that there is a relationship between the shadowing signal strength at one point in time or space and that at another point in time or space. This relationship is a 'stochastic' one and it is by virtue of exploiting this relationship that the time-domain signal strength predictor described above is realized i.e. that the signal strength can be predicted to within a certain degree of accuracy into the future for a moving body. Since time and space are related by velocity, the same set of algorithms can be employed in the space domain — in other words the signal strength can be predicted to within a certain degree of accuracy at a location other than that of the moving node.

The algorithms presented below allow a moving mobile device to estimate (give a numerical value for) the signal strength into the future. There are two types of algorithms:

Optimal algorithm - which gives the most accurate predictions possible, Sub-optimal (but much faster) algorithms based on the optimal algorithms.

There are many possible permutations in the manner in which these algorithms can be implemented. The appropriate choice of algorithm and the manner in which it is implemented depends on (but is not limited to) the following factors: a) Predictive accuracy requirements b) Whether or not the network is infrastructure-based (i.e. has base-stations) or is infrastructureless (e.g. ad-hoc or mesh) c) Computational capability of the mobile device (depending on factors such as processor speed etc.) d) Bandwidth limitations (in communicating data to and/or from the base station(s). e) Memory (which often is not a problem) Both types of algorithm require knowledge of the speed of the mobile device, which can be obtained using an inexpensive accelerometer chip incorporated into the device hardware or obtained from change in GPS position.

The signal strength is sampled by the mobile device on a continuous basis. Depending on the capability of the mobile device and bandwidth limitations this information is either stored locally or relayed to the base station.

Where the optimal predictor is being used Equations (1) to (4) below are executed. The major data inputs are signal strength readings, node velocity and autocovariance data of the shadowing signal. This latter information may be given/obtained in a number of forms such as the 'autocorrelation function', the 'decorrelation time' or the 'decorrelation distance' of the shadowing signal. The 'decorrelation distance' is of the order of the size of shadowing objects or clusters of such objects. Where this can be done locally (i.e. in real-time or near-real-time) the necessary (location dependent) empirical data is either obtained directly from the base-station or obtained from a locally hosted database.

For the sub-optimal predictors the following are the main data inputs: signal strength readings, node velocity and the autocovariance data of the shadowing signal. This latter information may be given/obtained in a number of forms such as the 'autocorrelation function', the 'decorrelation time' or the 'decorrelation distance' of the shadowing signal. The 'decorrelation distance' is of the order of the size of shadowing objects or clusters of such objects. Where this can be done locally (i.e. in real-time or near-real-time) the necessary (location dependent) empirical data is either obtained directly from the base-station or obtained from a locally hosted database.

Where the computation is beyond the ability (in the sense of fast execution-time) of the mobile device some or all of these equations or just the time-consuming portion namely: the inversion of matrix A(d) of Equation (2) can be implemented by the base-station either on request or in an anticipative fashion and the result(s) relayed to the mobile device.

In an infrastructureless network (ad-hoc or mesh) where it is desired to use the optimal predictor, the empirical data may be hosted locally and the mobile device should have sufficient computational ability. Otherwise one of the suboptimal estimators (given current processing speeds) most probably ought to be used. - -

Whether the mobile device is operating in an infrastructure-based or infrastructureless network, one of the sub-optimal predictors can be used with the second of these (given below) being executable in real-time in all cases.

Algorithms:

The following algorithms are formulated in the time domain. Using the simple velocity relationship between time and space the following algorithms can be realised in the. space domain.

(i) Optimal Predictor P denotes a sequence of averaged received signal-power samples as a vector of size N_c xl :

N_N-I where Rr - T NT L Rr-i

H., i=0 and_i?₍is the instantaneous signal strength sample as measured at time 't' by the mobile device. The appropriate sampling rate and number of samples, N₁₁, is obtained as described by G.

Gaertner and E. O. Nuallain in "Using the Rice and Nakagami Distributions to Model Wideband Small-Scale Fading in Urban Microcells," IEEE Transactions on Vehicular Technology VoI 56, Issue 2, Nov 2007.

C_T = [1 1 ... I]AOS)-¹ ^•[1 1 ... I]A(S)-¹ -P (2)

where A{δ) is the autocovariance matrix of the shadowing signal: - -

where σ_s is the variance of the shadow fading stochastic process. p_ref = p^ where p_D is the normalized covariance of the shadowing signal. D is the distance separating the two nodes and δτ_t__τ Λ is the additional displacement of mobile nodes z andy in the time interval

to their relative motion.

Having obtained C₁. , S_T is then:

(3)

Then the optimum prediction for the signal strength made at time 'T' for r seconds into the future is:

where δ_T^_T is estimated from the node velocity using an accelerometer and/or GPS data.

(ii) Sub-Optimal Predictors

1) If, for example, a constant velocity is assumed and/or fast matrix inversion techniques are used, such as sparse-matrix techniques, the autocovariance matrix of the shadowing signal, A(δ) , is simplified allowing for quicker inversion of this matrix in Equation (2). In this case Equations (1), (2), (3) and (4) are executed. This is the first means of sub- optimal prediction.

2) If it is assumed that the autocovariance matrix, A{δ) , approximates the Identity Matrix, I, then: N_c

— the computational complexity of the algorithm is then reduced to a scalar. In this case Equations (1), (5), (3) and (4) are executed. This is the second means of sub- optimal prediction.

It will be appreciated that the invention provides for accurate and fast signal strength prediction, allowing any of a wide variety of actions to be performed based on the predictions. This has been achieved without need for more powerful processing hardware on the mobile device.

The invention is not limited to the embodiments described but may be varied in construction and detail.

Claims

1. A method of predicting electromagnetic signal strength into the future between devices having wireless communications interfaces and which are relatively moving, the method being implemented by at least one of said devices, said device having a processor, the method comprising the steps of:

monitoring the signal strength at one or both devices, and

2. A method as claimed' in claim 1, wherein the signal strength predictor is executed without using topological data.

3. A method as claimed in claims 1 or 2, wherein a dynamic decision is made between optimal and sub-optimal signal strength predictors, the sub-optimal predictors being faster to execute.

4. A method as claimed in claim 3, wherein the decision is made on the basis of predictive accuracy requirements.

5. A method as claimed in either of claims 3 or 4, wherein the decision is made on the basis of whether the network is infrastructure-based or is infrastructureless.

6. A method as claimed in any of claims 3 to 5, wherein the decision is made on the basis of computational capability of the processor.

7. A method as claimed in any of claims 3 to 6, wherein the decision is made on the basis of bandwidth limitations on communication between the devices,

8. A method as claimed in any of claims 3 to 7, wherein the decision is made on the basis of memory capacity of the processing device.

9. A method as claimed in any preceding claim, wherein the devices are a mobile device and a base station transceiver.

10. A method as claimed in any preceding claim, wherein the processor executes the signal strength predictor using as inputs current signal strength, speed of at least one device, and autocovariance data of the shadowing signal.

11. A method as claimed in claim 10, wherein the autocovariance data of the shadowing signal comprises an autocorrelation function governing self-correlation of the shadowing signal.

12. A method as claimed in either of claims 10 or 11, wherein, the autocovariance data comprises de-correlation time of the shadowing signal.

13. A method as claimed in any of claims 10 to 12, wherein the autocovariance data comprises de-correlation distance of the shadowing signal.

14. A method as claimed in claim 13, wherein the de-correlation distance is of the order of the size of shadowing objects or clusters of such objects.

15. A method as claimed in claim 14, wherein the de-correlation distance comprises empirical data obtained from a device or from a locally-hosted database.

16. A method as claimed in any of claims 10 to 15, wherein for a sub-optimal predictor, the processor assumes constant speed of the device.

17. A method as claimed in any of claims 10 to 16, wherein for a sub-optimal predictor, the processor performs estimations to determine the autocovariance data.

18. A method as claimed in claim 17, wherein the processor performs fast matrix inversion techniques to determine autocovariance data.

19. A method as claimed in claim 18, wherein tihe processor simplifies an autocovariance matrix of the shadowing signal, and uses sparse matrix processing techniques to process said matrix.

20. A method as claimed in claim 19, wherein the processor approximates the autocovariance matrix to an identity matrix, providing for scalar processing instead of matrix processing.

21. A method as claimed in any preceding claim, wherein the processor uses the output of the signal strength predictor to estimate the quality of a channel before deciding to use it.

22. A method as claimed in any preceding claim, wherein the processor uses the output of the signal strength predictor to perform pre-emptive channel hopping on anticipation of a degradation in link quality.

23. A method as claimed in any preceding claim, wherein the processor uses the output of the signal strength predictor to perform handover.

24. A method as claimed in any preceding claim, wherein the processor uses the output of the signal strength predictor to perform pre-emptive routing in an infrastructureless network.

25. A method as claimed in any preceding claim, wherein the processor uses the predictions to build a space domain radio emissions map.

26. A communication circuit comprising a processor adapted to perform a method of any preceding claim.

27. A mobile device comprising a communication circuit as claimed in claim 26.

28. A computer readable medium having software code embodied thereon, said code being adapted to implement a method of any of claims 1 to 24 when executed by a digital processor of a mobile device.