MULTIPLE OBJECT LOCALISATION WITH A NETWORK OP RECEIVERS
Field of the invention
The present invention relates to the unique localisation of a plurality of objects using passive receivers and at least one transmitter.
Background
Passive receiver networks are regarded as an accurate and effective technique for the localisation of an RF signal source and/or signal reflecting object. The major advantage of passive receiver networks is that they operate without the need for knowledge of the incoming signal and they do not require transmission of a localisation signal.
The common technique for localising in passive receiver networks is multilateration, also known as hyperbolic positioning, which is based on the time-difference-of- arrival (TDOA) between the various receivers . As with triangulation, to localise uniquely a single object, a specific number of receivers is required, In passive receiver networks, it is required to maintain a larger number of receivers than objects (for 2-D localisation a minimum of four receivers is required for two or more objects). When this condition is not met, it results in intersections which provide false object positions, without any real substance, which are called "phantoms".
It is therefore imperative to remove these phantoms in order to estimate the true location of the objects.
There have been numerous techniques proposed for this
"spatial filtering", i.e. the removal of the phantoms in the localisation process using passive networks. However, the solutions are based on angular and elevation measurements, exhaustive processing, hardware alterations and tracking. None of these solutions can be applied to a scenario in which:
o There are no multiple acquisitions, which means that tracking cannot be used.
o The receivers might be stationary; therefore it is impossible to move them in order to identify the phantoms as the intersections that move to a physically impossible position.
o The available receivers do not have the capability of measuring anything else apart from time, so that angular and elevation information is not available.
o Each receiver operates stand-alone and transmits the received signal and/or other related information from the signal and the position of the receiver, to a data processing unit.
Summary of the Invention
According to the present invention, there is provided a method of determining the positions of a plurality of objects by processing signals emitted from at least one transmitter, reflected by the objects and received by a plurality of receivers, the method comprising: processing the signals from the receivers to generate a first set of time difference measurements comprising a measurement for each object and each group of receivers
defining the time difference of arrival between the signals received at each of the receivers in the group from the object; processing the first set of time difference measurements to calculate a respective position curve for each object and each group of receivers on which the position of the ob]ect must lie; determining intersections of the generated position curves to generate a plurality of possible positions for each individual object; processing the signals from the receivers to generate a second set of time difference measurements comprising a respective measurement for each receiver-object pair defining the time difference between the reception by the receiver of the direct signal from the transmitter and the reception by the receiver of the signal reflected by the object; associating the time difference measurements in the first set with the time difference measurements in the second sec; and determining the positions of the objects by relating the calculated intersections of the generated position curves to the time difference measurements in the second set.
The present invention also provides an apparatus for determining the positions of a plurality of objects by- processing signals emitted from at least one transmitter, reflected by the objects and received by a plurality of receivers, the apparatus comprising: means for processing the signals from the receivers to generate reflected signal time difference of arrival measurements, comprising time difference of arrival measurement for each object and each group of receivers defining the time difference of arrival between the reflected signal received at each of the receivers in the group from the object; processing the reflected signal time difference of arrival measurements to calculate a respective
position curve for each object and each group of receivers such that the position curve defines a curve upon which the position of the ob3ect must lie; determining intersections of the generated position curves to calculate a plurality of possible positions for each individual object; processing the signals from the receivers to generate direct-reflected signal time reception measurements comprising a respective time measurement for each receiver-object pair defining the time difference between the reception by the receiver of the signal direct from the transmitter and the reception by the receiver of the signal reflected by the object; and selecting from the plurality of calculated positions for each object a single position for the object by relating the plurality of calculated positions to the direct-reflected signal time reception measurements.
The present invention further provides a computer program product to program a programmable processing apparatus to become operable to perform a method as set out above or to become configured as an apparatus as set out above.
An embodiment of the present invention provides unique localisation, i.e. spatial filtering, of multiple objects in a passive receiver network, which:
o Allows the unique identification of an object's position and rejection of the "phantom" position based solely on time measurements using simple passive receivers.
o The minimum amount of receivers required to uniquely identify the position of the objects is three but the embodiment can be extended for more sensors .
o The embodiment is applicable to both 2 -D and 3-D scenarios .
o There is no additional signal requirements, since the whole procedure is based on the signal provided by the transmitter,
o No assumptions are made on the initial objects' position, and only the values obtained from the localisation are used.
o The embodiment can employ simple receivers as there is no need for hardware changes to the antenna.
o No additional signal acquisition is required because the embodiment can operate on a single snap-shot of the environment - there is no need to track the object
o There is no need for angular or elevation measurements which increase the complexity and cost of the receiver.
o There is no additional transmission requirement from the receiver's point of view.
Brief Description of the Drawings
Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, m which:
Figure 1 is a block diagram showing the functional units
in a signal processing apparatus of an embodiment.
Figure 2A. illustrates an example scenario and Figure 2B shows the scenario of Figure 2A in generic schematic form.
Figure 3 shows an example form of signal received at a receiver .
Figure 4, comprising Figures 4A and 4B, is a flow chart showing the processing operations performed by the apparatus of Figure 1 in an embodiment .
Figure 5 shows an example of an auto-correlation function for a transmitter signal.
Figure δ shows an example of an auto-correlation function of a signal received at a receiver.
Figure 7 shows an example of plotted hyperbolae for the scenario illustrated in Figure 2A.
Figure 8 shows the hyperbolae of Figure 7 after processing to remove values corresponding to the transmitter.
Figure 9 is a flow chart showing the processing operations performed at step S4-14 in Figure 4.
Figure 10 is a flow chart showing the processing operations performed at step S4-16 in Figure 4,
Figure 11 shows the identity of the true object positions calculated as a result of the processing in the
embodiment .
Embodiments
The principles of the processing operations performed in this embodiment will be described first. An illustrative example will follow.
As stated above, the localisation technique used in the scenarios investigated are based on time difference of arrival (TDOA) . It is not required to know the signal transmitted from the transmitter or the time it was transmitted. Therefore the only available information obtained from a time difference estimation technique, such as correlation, can geometrically correspond to the propogation path distance differences of the signal to the receivers. For convenience, correlation will be used as the standard technique in this embodiment, but alternative time delay estimation techniques could be considered for implementation, such as that described in
WO-A- 00/39643, for example.
The passive receivers transmit the received signal, which is a combination of the reflections from the objects whose positions are to be determined and the direct signal from the transmitter, or other appropriate information of that signal (such as described in US 7,170,820}, to a Centralised Data Processing (CDP) unit. When all the data are collected, standard localisation based on TDOA is performed. The TDOA measurements are obtained from the peaks of the cross- correlation function of a sensor pair i-j . TDOA localisation is based on hyperbolae, A time measurement corresponds to a peak obtained from the correlation R1-,
of receiver pair i-j , where the correlation function is calculated with respect to the signal received on receiver j . For each one of these time measurement a hyperbola is generated based on the TDOA between the examined receiver i and receiver j , For n number of receivers and using all possible sensor pair combinations, the maximum available number, hraax, of resulting hyperbolae will be: h =___z_L_ CD
"" 2(«-2)! for each individual object. Furthermore, for each individual object, the hyperbolic localisation will provide two valid intersections. A valid intersection is an intersection resulting from hmax hyperbolae crossing at that point (or region in case of noise) . Therefore, there will be 2*(v+l) valid intersections, where v is the number of objects. The reason v +1 is used, is based on the fact that the transmitter is also treated as an object at this stage. Nevertheless, since its position is known to the CDP unit and the receivers, it can be uniquely identified. The final stage of the localisation is to identify the pairs of intersections corresponding to each individual object, as the intersections which are generated from hmax hyperbolae. If an intersection comprises of less than hmax hyperbolae, it is disregarded.
Geometrically, both resulting valid intersections of a single object share the same geometrical properties compared to the respective receivers i , j and the resulting hyperbolae. Since the intersections are points of a hyperbola, the difference of the distances between any of these two intersections and foci points (receivers i and j) are the same. Therefore, this cannot be used to extract the true objects position.
The only other information which can be obtained without any extra hardware or additional received signals is the autocorrelation of R3j for any receiver j . The resulting feature will provide with a relative time measurement for a particular receiver. This cannot be used directly for localisation purposes as the time difference measured from R-,-, is not directly linked to the distance to the object. Nevertheless, it can provide a numerical value TA] for the time measurement based on the following equation:
which will only provide the relative difference between the transmitter signal direct path to the receiver j and the route reflected from the object A, DTj is the distance of the transmitter to receiver j , OTA IS the distance of the transmitter to object A, DA] the distance from object A to receiver j and c the speed of light. The auto-correlation feature will have a peak at relative time zero, which will correspond to the direct path from the transmitter since this is the one that will arrive first. The remaining peaks will correspond to the path from the reflected objects as given in equation (2) , Generating equation (2) for each object and from each receiver will provide the additional information to uniquely identify the position of the object. Assuming a second object B, auto-correlation from receiver i, R1^, will give a numerical value for TBi corresponding to the equation τ ^{DTB+DBl)-DTi (3) ' c
The notation follows the pattern described in the previous equation. Depending on the number of objects and
receivers in the scenario examined, it is possible to generate more equations similar to (2) and. (3) .
The problem now is that the peaks of the auto- correlation function of any receiver cannot be directly associated to a particular object. The cross-correlation function provides the TDOA measurements Tjj for an object which cannot be directly related to the time measurements provided by the autocorrelation function of the i or j receivers, Assume now that, from receiver pair i-], labelled as n, there is a TDOA measurement value τ vri which corresponds to object v. This value can be written as
_DΛJ-DΛI (4}
C Having numerically established all the values for the time measurement T
1^, where k is the k
th receiver, a vector is generated based on the combination of the values
Then, based on the valid intersections provided by the TDOA localisation, the following function is used to generate a vector:
T T T DTj-DT,+ (Ant, "Ant;) fς)
where Dinti is the distance of the valid intersection of object v selected from receiver i and Dint} is the distance of the same valid intersection of the same object to receiver j. The component Dinti-Dint:) is the same regardless of selecting the true or phantom location of the object (because of the properties of a hyperbola and its foci points) , and in the time domain corresponds to the TDOA measurement T vn obtained using cross-correlation on the nth pair of receivers. Therefore, by selecting any intersection of a particular object, it is possible to calculate Thyp_dif for each object. Then, by comparing the
values of Thyp_dif and the numerical values obtained from TVi-T11J, it is possible to associate the auto-correlation peaks with the TDOA measurements from cross-correlation from the receiver pairs. The technique described above shows that it is possible to associate the time measurements of auto-correlation to the TDOA measurements of cross -correlation but it is by no means the sole method to achieve this. For example, there are alternative techniques that can be used which are well known to those skilled in the art, such as circular error probable, rms based techniques, etc.
Having established this relationship, it is now possible to relate the valid intersections (true or phantom) with specific time measurements from auto-correlation. A testing function is generated as shown below (example assuming two objects, and three receivers) :
where T
B2 is the auto-correlation time measurement for object B from receiver 2, T
A1 is the auto-correlation time measurement for object A from receiver 1 and T
A3 is the autocorrelation time measurement for object A from receiver 3. Since these values are known, the — number c will be known as well, which will be called an objective number herein. Then, a list of all possible combinations of potential objects' positions is generated, which in this case will be four, namely Int Ai-Int B
1, Int Ai-Int B
2, Int A
2-Int B
1, and Int A
2~Int B
2, where Int corresponds to a valid potential intersection, the letter A or B corresponds to object A or B and the number to the potential position for that object. Then a function from the left-hand side component of equation 6 is formed,
which is called the hypothesis function H (posi,pos
2) ,
, (OT2-DTl-DT3 + 2^DT -Dj^+D 1 -D 2 + D 3) ff(POS1,POS1)- ™ —- ^ '—3 ^^— (7) c
The above example corresponds to equation 6, where posu. is the hypothesised position for object A, pos2 is the hypothesised position from object B, DTn is the distance of the transmitter to receiver n, DTpOBV is the distance of the transmitter to hypothesised position of object v and DpOSkn is the distance of the hypothesised position of object v from receiver n. By selecting mm[—H(posx,pos2)} (8) c the combination of the true position of object A and the true position of object B is revealed.
In the case of more objects, equivalent functions to equations (6) and (7) have to be generated. In the case where more receivers are available, they could be incorporated in the equations but that is not essential. Checking each component individually with the time measurement obtained from the auto-correlation functions of each receiver, it will introduce an error m the case of symmetrical scenarios. Therefore, by cross-checking measurement from different receiver and different objects, the probability of a completely symmetrical case is reduced. Furthermore, the above functions are only illustrative and more complex ones can be generated, Nevertheless, the main point to note here is to avoid containing only components such as DA]-Dftl because for both the true and phantom intersection of that object the result is the same. Therefore, the hypothesis and testing function must be generated by cross -combining time
measurements from different objects and different receivers. The major points to note with the processing described above are:
o Good time difference estimation techniques should be available and short pulses from the transmitter would be advantageous for this purpose to avoid interleaving with received pulses from reflections. Nevertheless the processing can operate in noisy environments without specially designed signals, as it is based on the provided intersections.
o No need for extra signals to be transmitted from the receivers to the CDP unit.
o No need for additional types of measurements {angle, elevation , etc) .
o No need for complex hardware additions.
o The hypothesis function shown earlier is linear and based on difference and summation of distances.
o Flexibility of combinations of receivers and objects for the hypothesis function and objective number.
o Solutions are unique as they are based on unique positions in space.
o Accuracy can be improved by designing dedicated hypothesis functions.
o The processing can be applied to 3-D problems, without significant changes to its structure as it
is based on ranges and time measurements.
o The processing is applicable to more than 2 objects scenario, where the testing function which will incorporate time measurements for the added objects.
o If more than three receivers are used for the localisation, or are available, it is possible to use them in the testing and hypothesis functions. Nevertheless, based on the function discussed above, only three receivers are necessary. Therefore, the optimum three receivers can be selected to minimise the probability for error.
o The processing can be applied to both centralised, distributed or any other hybrid architecture localisation networks.
Figure 1 shows a block diagram of an embodiment of a CDP unit 100 for performing processing operations based upon the principles described above.
Referring to Figure 1, a central controller 120 is operable to provide control and coordination of the other functional units. Working memory 130 is provided for use by the central controller 120 and the other functional units .
Input data interface 140 is operable to receive the signals 150 transmitted to the CDP unit 100 from the receivers, and is further operable to control the storage of data defining the signals in data store 160.
Object-receiver pair TDOA calculator 170 is operable to
process the .signals from the receivers to generate time difference of arrival measurements, each of these comprising a measurement for a respective object and a respective pair of receivers which defines the difference between the time of arrival of the signals from the object at each of the receivers in the pair.
Hyperbolae processor 180 is operable to perform hyperbolic localisation for the objects based upon the time difference of arrival measurements generated by ob] ect-receiver pair TDOA calculator 170. More particularly, hyperbolae processor 180 is operable to calculate a respective hyperbola for each object and each pair of receivers such that the positions of the receivers comprise the foci points of the hyperbola, and the hyperbola itself comprises a line upon which the position of the object must lie. Hyperbolae processor 180 is further operable to calculate the intersections of the hyperbolae and identify therefrom the intersections corresponding to possible positions of each individual obj ect
Direct-reflected signal time difference calculator 190 is operable to process the signals from the receivers to calculate a direct-reflected signal time reception measurement for each receiver-object pair which defines the time difference between the reception by the receiver of the direct signal from the transmitter and the reception by the receiver of the signal reflected by the object. Direct-reflected signal time difference calculator 190 is operable to perform processing in accordance with the principles described above with reference to equations (2) - (4)
Time measurement associator 200 is operable to perform processing to associate the time difference of arrival measurements calculated for the object and each pair of receivers by object -receiver pair TDOA calculator 170 with the direct-reflected signal time reception measurements calculated for the receiver-object pairs by direct-reflected signal time difference calculator 190. Time measurement associator 200 is operable to perform processing in accordance with the principles described above with reference to equation (5) .
Time measurement -posit ion associator 210 is operable to perform processing to associate the possible object positions determined by hyperbolae processor 180 with the direct -reflected signal time reception measurements calculated for the receiver-object pairs by direct- reflected signal time difference calculator 190. Time measurement -position associator 210 is further operable to determine the correct positions for the objects based upon this processing. Time measurement-position associator 210 is operable to perform processing in accordance with the principles described above with reference to equations (6) - (8) .
The functional units shown m Figure 1 may be implemented as hardware units and/or as units resulting from the programming of a programmable processing apparatus. In the case of a programmable processing apparatus, the apparatus may be programmed to operate in accordance with computer program instructions input, for example, as data stored on a data storage medium 110 (such as an optical CD ROM, semiconductor ROM, magnetic recording medium, etc.), and/or as a signal 112 (for example an electrical or optical signal) input to the programmable processing
apparatus, for example from a remote database, by transmission over a communication network such as the internet, or by transmission through the atmosphere.
An illustrative example will now be described for a better understanding of the processing m an embodiment. This particular example will depict a scenario with the minimum amount of objects and receivers (RF receivers) that the system may be faced with. It is assumed that pre-filtering and optimum techniques are used to provide accurate time measurements from both cross and autocorrelation functions without large ambiguities resulting from noise or the motion of the object. Nevertheless, the embodiment can operate in the case of noisy measurements since it operates in space but will be biased as the intersection based distance calculations will be impaired by che time measurements.
There are numerous applications where passive localisation is used. Due to the added restrictions concerning the spectrum usage, passive systems are considered as the "greenest" solution to the ever tighter legislation, since they re-use the existing frequency bands. In the example described below, the traffic flow on a motorway is monitored, and it is possible to use existing transmitters of other applications situated at the side of the road. The receivers are distributed along the motorway but do not necessarily follow the curvature of the road. The objective is to increase the Geometric Dilution of Precision (GDOP) , as it is known that receivers in straight lines may cause degradation to the localisation accuracy m certain object orientations.
The scenario is illustrated at Figure 2A Three receivers
2, 4, 6 acquire a signal directly from a transmitter 8. At the same time, two vehicles {objects} 10, 12 reflect the transmitter's signals back to the receivers 1, 4, 6. To minimise the cost and complexity of localisation, only the minimum number of receivers are installed. Additionally, it is possible that receivers could malfunction and could not be used Therefore, it is necessary to use processing as described in the previous section to remove the phantom intersections resulting from the hyperbolic localisation. The signals are then relayed from the receivers to the CDP unit 100 where the vehicle's position is calculated and the flow and volume of traffic is monitored without the need of optical means. Nevertheless, it is possible to combine the proposed RF system with existing optical systems to complement each other.
The scenario illustrated m Figure 2A is converted to a generic schematic form shown m Figure 2B, where the receivers 2, 4, 6 are presented as circles, the transmitter 8 as a clear diamond and the objects (cars) 10, 12 as solid diamonds There are three passive receivers 2, 4, 6 distributed in. the area under monitoring. For this particular scenario, the receivers are randomly distributed in space. The three receivers 2, 4, 6 are situated at [-90 0], [10 0] and [50 -45]. In the scenario used and illustrated, all the distances and coordinates quoted are m metres. These are passive omnidirectional receivers 2, 4, 6 which can only transmit the received signals or other necessary information of the received signals to the CDP unit 100 for the purpose of localisation {centralised architecture) . In a case of distributed sensor network, the processing will take place at each receiver 2, 4, 6 individually, so they will
require information from each other receiver. The positions of the receivers 2, 4, 6 are known. Furthermore, there is a transmitter 8 which transmits the signals that will be used for the localisation from the receiver network. The position of the transmitter 8 is known to the CDP unit 100. Finally, there are two objects 10, 12 inside the monitoring area positioned as shown in Figure 2B at [-20 65] and [10 -15] . The system is not aware of the position of the objects 10, 12.
In the example, the transmitter 8 transmits a short pulse, although it could be any signal suitable for localisation. Each of the receivers 2, 4, 6 will pick up a signal that will contain the pulse from the direct line of sight, the pulse reflected off object A and the pulse reflected off object B. If the duration of the pulse is shorter that the distance (converted in time) , the signal received at any receiver 2, 4, 6 will have the form shown in Figure 3. The first received pulse will always belong to the transmitter 8 and the subsequent ones to the objects 10, 12. Nevertheless, at this point, it is impossible to associate a pulse with a particular object. This signal is then transmitted to the CDP unit 100, since the present embodiment comprises a centralised architecture network.
Figure 4 is a flow chart showing the processing operations performed by CDP unit 100 in this embodiment.
At step S4-2, data is stored in data store 160 defining the location of the transmitter 8 and the respective location of each of the receivers 2, 4, 6.
At step S4-4, input data interface 140 receives the
signals 150 transmitted from the receivers 2, 4, 6, and stores data m data store 160 defining the received signals .
At steps S4-6 and S4-8, object -receiver pair TDOA calculator 170 and direct-reflected signal time difference calculator 190 perform processing to find the time measurements required.
The TDOA measurements t and the direct-reflected signal time measurements T can be found at steps S4-6 and Ξ4-8 using correlation-based techniques or any other process that can be used to determine the time difference between the arrivals of two or more signals, such as for example described in WO-A-00/39643.
In the present embodiment, at step 4-6, direct-reflected signal time difference calculator 190 performs an autocorrelation technique to generate a respective direct- reflected signal time reception measurement for each receiver-object pair. Each of these measurements defines the time difference between the reception by the receiver of the direct signal from the transmitter and the reception by the receiver of the signal reflected by the object. If the transmitter signal is a linear frequency modulated (LFM) pulse, its auto-correlation function will be similar to the one shown m Figure 5. Nevertheless, the signal observed at each receiver 2, 4, 6 contains all pulses from both direct and indirect paths Therefore, for the scenario of Figure 2B, the auto-correlation function of the signal received at receiver i based on the transmitted pulse from the transmitter 8 will be as shown in Figure 6. As noted previously, the autocorrelation at step S4-6 is performed m the CDP unit
100, but it can be done instead in the receiver. The auto-correlation can be done in various ways. First, the signal can be auto-correlated with its copy, using a conventional technique. However, to improve the quality of the auto-correlation function, in the present embodiment a signal Is generated synthetically that imitates the reception of the transmitter's signal at a particular receiver. This is possible because both the distance of the transmitter to the receiver is known and the signal is known or can be acquired. The synthetically generated signal for each receiver is then used to auto-correlate its corresponding receiver. The resulting peaks correspond to the time difference of reception of the direct path signal and the reflected signals, as given from equations {2) , For a receiver i, the auto-correlation function will provide the values TAi and TBi. Nevertheless, for the reasons explained above, this cannot be used for radial localisation, i.e. to provide direct relation to the distance of the object from the receiver. Accordingly, CDP unit 100 stores all the values from the autocorrelation of each individual receiver for further processing. For the scenario illustrated in this example, the result will be
where v is the vth object and n is the nth receiver.
At steps S4-8 to S4-12, object -receiver pair TDOA calculator 170 and hyperbolae processor 180 perform a standard procedure for hyperbolic localisation. It is
required to estimate the TDOA between the signals arriving at receiver pairs. To do that, object -receiver pair TDOA calculator 170 calculates the cross-correlation of the signal of receiver i based on the signal from receiver j . Each TDOA measurement is then used by hyperbolae processor 180 to generate a hyperbola using as foci points the two receivers used from the correlation function. This technique is well known to those skilled in the art (see, for example, "Wireless Sensor Networks, An Information Processing Approach", The Morgan Kaufmann series in networking by F. Zhao, L. Guibas, 2004, ISBN 1558609148). Accordingly, it will not be explained here, Nevertheless, it must be noted that all the possible receiver pairs are used in order to minimise the amount of potential positions (true and phantom intersections) for each object. In this case, the possible receiver pairs are Receiver 1 -Receiver 2, where Receiver 2 acts as MASTER, Receiver 3-Receiver 2 (2 as MASTER) , and Receiver 3 -Receiver 1 (1 as MASTER) . The MASTER receiver corresponds to the receiver used to provide the time reference for the calculation of the time difference. From the cross-correlation function at step S4-8, the TDOA time measurements τ will be obtained as:
Ή ' 21 ' 31 τ. vk τ\2 ττι τιi : io )
■ 13 T 2. 3 T- 33
where v is the vth object and k is the kth receiver pair.
It is now required to associate each column to an object.
But in order to minimise the iterations required, it is advisable to plot first the hyperbolae in order to find
the r values corresponding to the transmitter. This processing is performed by hyperbolae processor 180 at step S4-12. The hyperbolae plotted from the τ matrix can be seen in Figure 7. Each of these hyperbolae corresponds to an individual TDOA and can be easily identified. Therefore, it is possible to remove the values corresponding to the transmitter, as it is possible to find which three hyperbolae intersect at the transmitter's position, which is well known and does not require to be spatially filtered. Thus, we are left with the minimum number of hyperbolae and minimum number of intersections, which for this scenario are four, two for each object, as shown in Figure 8. At the same time, it is now possible to associate each object with its respective TDOA measurement from each receiver pair. The T can now be re-written as
where v is the v
th object, and k is k
th receiver pair.
At step S4-14, time measurement associator 200 performs processing to associate the TDOA time measurements τ with the time measurements T obtained from the autocorrelation. Figure 9 shows the processing operations performed by time measurement associator 200 to do this. These processing operations are based upon the principles described previously with reference to equation 5. It is only required to use one of the intersections of a particular object since, in equation 5, the subtraction term in brackets is equal for both the true and phantom position. Furthermore, from the matrix
in equation 11, each τ value has been associated with a hyperbola, and therefore with the particular intersections. Thus, it is possible to re-organise the T matrix in such a way as to associate the time measurements with the particular objects, ϊn the absence of any extra information, each possible combination is used, as the first object will be tested against all single receiver peaks, and associated to the closest value, then the second object with the remaining values, and so on so forth.
Having associated each T value with its corresponding τ value, time-measurement position associator 210 performs processing at step S4-16 to relate the hyperbolae intersections calculated at step S4-12 (defining the possible positions of the objects) to the time measurements T calculated at step S4-S, so as to determine the unique, true positions of the objects. This processing is shown in Figure 10, and is based on the principles described previously with reference to equations (6) - (8) .
Initially, all the potential object combinations are identified (step S10-2) . In this case, there are only four. Note here that the objects cannot be positioned on the same hyperbola twice. Each hyperbola corresponds to the potential positions of a single object. From the numerical values, the testing function in equation 6 is formed (step SlO -4) and for each combination of positions the hypothesis function in equation 7 is calculated (step S10-6) . The correct object placement will correspond to the combination that minimises function 8, and is selected at step S10-8. Using the above procedure it is possible to uniquely identify the true positions of the
objects, as shown in Figure 11.
Modifications and Variations
Many modifications and variations can be made to the embodiment described above.
For example, in the embodiment described above, a single transmitter 8 is employed. Nevertheless the processing in the embodiment above can be applied when multiple transmitters are available. One way to do that is to select the most preferable transmitter based on the power of the signal, temporal and spectral properties and/or transmitter's availability with the aim to improve the overall accuracy of the object localisation. Furthermore, an alternative approach assumes simultaneous use of all the available transmitters by reconfiguring the testing function so it is cross-combining not only the time measurements from the objects and the receivers but from different transmitters as well. In this case, the time- measurement associator 200 should be upgraded in order to facilitate the processing of multiple signals from various transmitters and allow it (the associator 200) to distinguish between the various signals, following the principle describe in the embodiment above.
The above technique can be developed further to incorporate more objects or more receivers. It should be noted that, in the case of having more than three receivers available for the localisation, it is possible to only use the optimum three following the equation discussed above. However, the extra receivers can still be used by re-arranging the equations accordingly.
Additionally, the processing can be applied to a 3-D scenario. Following the equations explained in the processing analysis, it is clear that only time and distance measurements are used which are not associated directly to whether the scenario examined is two- or three-dimensional. Therefore, applying the processing to a 3-D case can be achieved by incorporating the three- dimensional valid intersections xn the processing and obtaining the distance measurements from three- and not two-dimensional coordinates.
It is possible to change the hypothesis and testing function to a more complicated form. This can be achieved by mathematically designing a test function, linear or non-lmear, that will reflect the characteristics that are required to remove potential problems from symmetries or maximise the separation between the values obtained for the various combinations of positions.
The processing described above was described m a passive localisation scheme which produces ambiguous objects' locations based on time measurements. Nevertheless, without derailing from the main description of the processing, it is possible to adapt it so that it can be applied to other schemes of passive, active or hybrid localisation schemes. As mentioned earlier, it can be applied to both 2-D and 3-D without changes to the core of the processing.