WO2012065233A1 - Tracking location of mobile devices in a wireless network - Google Patents
Tracking location of mobile devices in a wireless network Download PDFInfo
- Publication number
- WO2012065233A1 WO2012065233A1 PCT/AU2012/000018 AU2012000018W WO2012065233A1 WO 2012065233 A1 WO2012065233 A1 WO 2012065233A1 AU 2012000018 W AU2012000018 W AU 2012000018W WO 2012065233 A1 WO2012065233 A1 WO 2012065233A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- tracking
- range
- location
- node
- model
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/0009—Transmission of position information to remote stations
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
- G01S5/0284—Relative positioning
- G01S5/0289—Relative positioning of multiple transceivers, e.g. in ad hoc networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
- G01S5/0294—Trajectory determination or predictive filtering, e.g. target tracking or Kalman filtering
Definitions
- the invention relates generally to wireless communications, and particularly to methods and apparatus for dynamically tracking the location of mobile devices in a wireless network.
- GPS Global Positioning System
- GPS Global Positioning System
- a solution in such cases is to deploy a local positioning system consisting of an infrastructure of anchor nodes at known locations and mobile nodes attached to the objects to track.
- the location of the mobile nodes can be determined in two or three dimensions through an exchange of radio signals between the mobile nodes and the anchor nodes.
- cooperative localisation signals are not only exchanged between anchors and mobile nodes, but they are also exchanged between mobile nodes, hence nodes cooperate to determine their location using the network.
- Cooperative localisation usually uses distance measurements between communicating node pairs, the distance being estimated f om the exchanged radio signals and typically be based on time of arrival or signal strength.
- a simple form of cooperative localisation is iterative multilateration. Multilateration is used to determine the locations of mobile nodes with sufficient distance
- a method for dynamically tracking the location of mobile nodes in a wireless network comprising:
- a wireless tracking system comprising:
- each said mobile device including a transmission circuit for the sending data as radio signals, and a range estimation module for dynamically measuring the range between said device and at least one neighbouring device;
- each said anchor device being at a known location and including a range estimation circuit for dynamically measuring the range between said anchor device and at least one neighbouring device; and a Bayesian tracking filter for each said mobile device for receiving said measured range as an input and exchanging data with tracking algorithms for neighbouring mobile nodes, said tracking filter utilising a statistical model of error in measured range and a statistical model of node motion to dynamically determine location.
- a mobile wireless device comprising: a transceiver circuit for sending and receiving data to neighbouring said mobile devices;
- a range estimation module for dynamically measuring the range between said device and at least one neighbouring device
- a Bayesian tracking filter that receives said measured range as an input and exchanges data with tracking algorithms for neighbouring mobile nodes, and utilises a statistical model of error in measured range and a statistical model of node motion to dynamically determine location.
- a mobile wireless device comprising:
- a range estimation module for dynamically measuring the range between said device and neighbouring devices
- a transmitter circuit for sending measured ranges to separate device having a tracking filter for said mobile device.
- the method and apparatus combine cooperative localisation with temporal tracking in a Bayesian framework that allows other information to be included in the tracking algorithm.
- the tracking method uses a statistical model of the mobile node motion to constrain the node track and uses a statistical model of the error in the measured range/distance between nodes.
- the use of information about the distribution of measured range distance error has a beneficial effect on the tracking accuracy, particularly in radio propagation environments (such as indoors and underground) where the range/distance error is biased.
- Most distance measurement systems will produce some outlier measurements with potentially large errors, and a simplified distance error likelihood function is used in the Bayesian framework that is robust to such outliers.
- This f amework also allows other information to be included such a data from an inertial motion unit or a map of the location.
- a tracking filter is used for each mobile node.
- the computations for this filter can be performed on each mobile node, which has the benefit of allowing the network to scale to arbitrary size, however the computations could be distributed in other ways or centralised.
- Each tracking filter exchanges information with filters for neighbouring nodes so that mobile node filters have information including the range to and location of neighbouring nodes.
- the Bayesian framework can make use of the information about the uncertainty in node locations, such as a distribution or co variance, when available. As communication bandwidth is limited when node location distributions are exchanged usually only a simplified model rather than a full two- or three-dimensional array will be exchanged.
- a 'locate and track' method uses a locate step to provide a location to the tracking filter. This allows a computationally simple filter to be implemented that is particularly suited to situations where the distance measurement error is
- a more complex 'direct range fusion' method allows tracking to continue where there are too few range measurements for the location step in the locate and track algorithm, and readily handles non-Gaussian noise. This approach is well suited to complex radio propagation environments such as within buildings and underground.
- embodiments of the present invention provide more accurate tracking through the greater use of information about and constraints on the mobile node. This also allows tracking to continue in situations where other methods would lose track, and also provides methods for determining when the track has been lost and for reinitialising the track of the mobile node.
- the estimation of kinematic parameters of the mobile nodes (such as velocity) that form part of the motion model state can also be estimated.
- Figure 1 is an example network in which an embodiment of the invention can be used.
- Figure 2 is a flowchart of the method of the invention in a broad form.
- Figure 3 is a flowchart of an embodiment tracking method.
- Figure 4 shows plots of measured range error distribution outdoors and Gaussian approximation.
- Figure 5 shows plots of measured range error distribution indoors and simplified likelihood functions.
- Figure 6 is a flowchart for node tracking, using a locate and track method.
- Figure 7 is an embodiment of the locate and track method for outdoor tracking.
- Figure 8 is a flowchart for node tracking using direct range fusion method.
- Figure 9 is an embodiment of the direct range fusion method for indoor tracking.
- Figure 10 is a schematic block diagram of an anchor node device.
- Figure 11 is a schematic block diagram of a mobile node device.
- Figure 12 is a schematic block diagram of another form of mobile node device.
- FIG. 1 presents a top-sectional view of a building 100.
- a network 102 within and around the building 100 includes nodes (shown as stars) 104, 106, 108 representing fire trucks which surround a building 100, with fire fighter nodes 110 regard (shown as dots) located inside the building 100.
- a "node” refers to a device that has wireless communications capability. ("Node” and “device” are used interchangeably.)
- the fainter lines represent node pairs between which range is measured. There is a need to measure the distance between the nodes.
- the range measurements between nodes are updated regularly and these measurements are used to track the locations of the nodes as they move. (Note that "range” and “distance” too are interchangeable terms.)
- the computations to update the location of the mobile nodes can be distributed or centralised, with the two most common situations being all locations computed on a single computational device within the network (e.g., attached to one of the fire trucks) or location being distributed and calculated on each of the mobile devices.
- the system will be described such that the tracking of each mobile node is distributed and performed within each mobile node, which has the advantage of scaling well with network size.
- the computations of range and the method for tracking could equally be distributed or centralised.
- Some nodes in the network 102 may have an independent means of determining their location. For example, for the fire trucks 104, 106, 108 outside the building 100 a GPS device could be used to determine their location. In other applications a conventional survey could be used to determine the location of fixed nodes. These nodes are known as the “anchor nodes", and are used to locate the other nodes in the network that are known as the “mobile nodes”. Mobile nodes in general do not have an independent means to locating themselves. They make available information about their location and possibly a measure of the reliability of the location. Anchor nodes are often fixed so no tracking is required.
- a mobile node may include other aids to localisation such as an inertial motion unit or odometry, and this information can be used in the tracking filter for each mobile node.
- Figure 2 shows a generalised method 200 for dynamically tracking the location of a mobile node in a wireless network.
- the method includes, in step 202, dynamically measuring the range between the mobile node and at least one other node. The measured range is provided as an input to a tracking algorithm.
- the tracking algorithm 204 is Bayesian in nature.
- the algorithm 204 exchanges data with tracking algorithms for neighbouring mobile nodes, and implements a statistical model of error in measured range 206, and a statistical model of node motion 208 to dynamically determine the mobile node's location.
- a Bayesian algorithm involves the calculation (usually an approximation) of the probability of the state given all data to date, and uses this to estimate the location.
- the terms "neighbour/ing node” includes any two nodes between which range can be measured.
- a process begins to track the location of the node.
- mobile nodes require cooperation with other nodes in the network to determine their location, and when turned on mobile nodes have no information about their current location.
- nodes measure the distance to their neighbours, obtain information about the state of their neighbours and send information about their own state to their neighbours. This information is used by the tracking method in each node to update their current state, which includes information about the location of the node and possibly other parameters such as velocity.
- a node tracking filter needs to be initialised, in step 308, with information such as an initial estimate of the respective node location. In general, there is no initial estimate of the state parameters at turn on, and the parameters become invalid if the location of the node is unknown. If the track is not initialsed, then an initialisation is performed in step 308, else if so, then node tracking begins in step 310. A test is then performed in step 312 of whether the track is valid, and if not, then in step 314 the track is terminated (to begin again).
- Information exchange step 304 returns information from all neighbours with which the current node has successfully communicated, which includes the possibility of information obtained from no neighbours or only a few. Is it also possible that a distance estimate to a node is obtained, but that node is not initialised so has no information about its location to send.
- the neighbour information must include information about the location of a neighbour.
- the specific information depends upon the details of the node tracking step 310, and such possibilities include coordinates for each node, or mean and covariance of node location estimate, or a distribution (two- dimensional or three-dimensional) for the node location estimate, or a compact representation of the distribution. It may be beneficial to exchange other information such as kinematic parameters for the mobile nodes.
- the information exchanged can also be about the last tracked location of the node, or a prediction of the node location at a future time.
- node tracking filters i.e., step 310
- the node tracking filters are distributed in a network due to limited communication bandwidth
- the information exchanged between nodes will be called "exchanged node information" hereinafter. This exchange is applicable to all nodes. Mobile nodes will exchange information from their tracking filters, and anchor nodes will provide their location and may also have an uncertainty in their location (e.g. if a GPS reference, the GDOP can also be sent).
- the node tracking step 310 can predict the current location of the node, however, in the absence of usable measurements, the uncertainty in the node location grows until the estimated location is useless and the track is regarded as lost.
- step 314 previous state information about the track is discarded and the track is reinitialised when suitable data becomes available.
- a counter is incremented every time the current information is set to the predicted state and the same counter is set to zero when the current information is set to the updated information. If the counter exceeds a predefined threshold, meaning that the node state information has been predicted for long enough, the track is terminated.
- the track quality is evaluated using the log-likelihood track score function. When the track score is lower than a predefined threshold the track is terminated.
- Initialisation of a mobile node tracking requires that an initial location (this could be coordinates or a distribution) for the mobile node is determined. There are often other parameters to be initialised, such as kinematic parameters or estimates of measurement and process noise, which are particular to the tracking method and motion model employed.
- a known simple technique is when range measurements to a sufficient number of neighbouring anchor nodes are available through the range and location information exchange with neighbours, in which case a multilateration method such as the well- known iterative least squares algorithm can be used. In the more difficult situation where cooperative localisation is required, such a technique preferably is not used.
- a known DV distance algorithm can be used.
- the multi-hop distance from each node to each anchor is determined using a distributed routing algorithm.
- the estimated multi-hop distance will be greater than the actual distance between the node and the anchors, and in a second phase a correction term is determined and propagated to all nodes.
- the mobile nodes can use a
- An optional refinement step to the DV distance position estimate can be applied, which iteratively updates the node location by exchanging information with neighbours.
- Node tracking can be viewed as inference on a dynamical system.
- the objective is to infer the state of the system, which consists of parameters that are not directly observable from noisy measurements.
- a dynamical system is often defined using two models: a model describing the change of the state with time (called a 'state transition model') and a model that describes the relationship between the state and the measurement (called a 'measurement model').
- state is traditionally used in filtering, and for the purposes of 'tracking', the 'state transition model' is a model of the motion of the node, parameterised by the values in the state (such as location and velocity).
- the 'state transition model' will be called the 'motion model' to emphasise its purpose, however it will be understood by those skilled in the art that a state transition model can include more than a motion model.
- the motion and measurement models are available in a probabilistic form to account for disturbances that occur in the real world. Treating the system as a dynamical system, and using temporal information and a motion model, allows more accurate tracking of the node and estimation of kinematic parameters if desired.
- v A _ is the process noise that accounts for the fact that a mobile node, in practice, cannot hold a constant velocity motion with certainty. It is commonly assumed to be Gaussian distributed with zero mean and covariance matrix Q k , given by where q is the process noise intensity.
- the nearly constant velocity model given in equation (3) is linear in the assumed state vector.
- the motion is not adequately modelled using a nearly constant velocity model and a constant acceleration model is used to approximate the state evolution.
- This model is described mathematically as
- ⁇ k [x k , x k , x k , y k , y k , y k ] T .
- the state transition matrix is given by
- the measurement model for tracking in a wireless network using range measurement between the nodes in the network in equation (2) will have the following form
- the characteristic of the (range) measurement error 3 ⁇ 4 is dependent on the environment in which the wireless network is operated.
- the measured range error distribution in an outdoor environment is shown in Figure 4, along with a Gaussian approximation. In this environment the range error distribution is unbiased and well approximated by a Gaussian except for a higher probability of outliers.
- Figure S shows (in the smooth curve) the measured range error distribution in an indoor office environment, which is clearly biased and non-Gaussian.
- Most current tracking systems assume a symmetric (and usually Gaussian) distribution for this error, however by using a better model a substantial improvement in the performance of the tracking algorithm can be obtained.
- the exact range error distribution is not usually known and an approximate distribution is required to be used in the tracking algorithm.
- the use of a simplified distribution reduces the computational cost for the tracking filter.
- a rectangular and a triangular approximation to the range error distribution are also shown in Figure 5. These show the bias in the measurement (range is likely to be overestimated due to the radio propagation environment).
- a further feature of these approximations is that they to not go to zero but to a small non-zero value outside the main region. This is an advantageous feature to make the tracking algorithm robust to outliers as measurements with very large errors do not have undue influence on the result (such as occurs with least squares techniques). These approximations are not true probability distributions as their integral is not bounded, however they can be considered as likelihoods of range error.
- the optimal Bayesian state estimation continuously updates the conditional posterior density of the state.
- the optimal Bayesian recursion for the estimation of the state of a node, given the range measurements up to and including sampling time A: is given by
- x k is the state of the node at time k
- z k is the set of ranges measurements obtained from all neighbours
- z lk _ t ) is the posterior density at time k - 1 , which is assumed to be available from the previous
- x t _, ) is the prior density and p(z k ⁇ x k ) is the likelihood, which can be calculated from the state and measurement models, respectively.
- z u _,) is the normalizing constant of the posterior
- the extended Kalman filter is a modification of the Kalman filter to handle nonlinear state-space equations.
- the nonlinear state and/or measurement equation is linearised using a first or second order Taylor series expansion.
- the EKF assumes a sufficient statistic of mean and covariance for the state, and propagate these through the linearised state-space model.
- the unscented Kalman filter approximates the state-space distribution using a Gaussian sufficient statistic. Unlike the EKF, the mean and covariance of the state are propagated through the nonlinear system as described by the state-space model, i.e., no linearization is performed. This is achieved by using a deterministic sampling approach in which the state is represented using a carefully chosen sample points.
- the particle filtering technique is a sequential Monte Carlo technique that approximates the optimal Bayesian recursion using a point mass representation of the posterior densities.
- This representation consists of a set of random samples and associated weights. Denote by ⁇ x ⁇ _, , ⁇ _, ⁇ , the random sample- weight pairs that represent the prior density at time& - 1.
- the particle filter algorithm approximates the posterior density at time k by a new set of random sample-weight pairs ( * , w k ' ⁇ , , through sampling and resampling.
- the measurement equation for range is non-linear, hence a non-linear filter is required, and where the noise is non-Gaussian the most suitable technique is the particle filter.
- location is computed for each set of measurements independently (cooperative localisation) then the measurement equation is linear and we can use a simple linear filter to track location. This is termed a "locate and track” method, and the more complex filter based on range is termed the "direct range fusion” method.
- the locate and track approach is computationally simpler and well suited to simpler applications particularly where the noise is Gaussian distributed.
- the direct range fusion filter performs better with non-Gaussian noise and allows tracking to continue when there are too few measurements for an independent locate step.
- An optional step preceding the node tracking step 310 is to validate the range measurements, and remove those that are inconsistent with the system state and presumably erroneous from consideration by the tracking algorithm. This is to prevent range measurement with large errors (outliers) from adversely affecting the tracking performance. In one embodiment, if the measured distance between the node and a neighbour exceeds by a threshold the expected distance (based on either previous or predicted locations) then that distance measurement is rejected as an outlier.
- FIG. 6 An example embodiment of the tracking method (step 310 in Figure 3) is shown in Figure 6, where the Bayesian filter uses the locate and track method.
- the input in step 602 holds range and exchanged information for each neighbour.
- This tracking method first predicts the track of the node in step 604 by applying the prediction step of the filtering algorithm it uses, based on the selected motion model.
- step 606 After predicting the state information of the node the locate and track method attempts, in step 606, to calculate a new node location using the most-recent set of measurements. This can fail to find a valid solution for a number of reasons including too few measurements, ill-conditioned data or inconsistent data.
- step 608 if a new location was not able to be calculated then the current state of the node becomes the predicted state in step 610. If a new location can be computed then, in step 612, the tracking algorithm uses the predicted state and calculated location to update the state information in step 614. The current state is output in step 616.
- the localisation step 606 of the algorithm There are a number of algorithms that can be used for the localisation step 606 of the algorithm. If the range error is Gaussian distributed then the least squares method is well suited to calculate the node location. If the range error is roughly Gaussian distributed but with a large number of outliers, such as measured outdoors and shown in Figure 4, then the robust least squared algorithm performs well. Where the range error is not approximately Gaussian distributed, such as shown for indoor measurements in Figure 5, and particularly where it is not symmetric then a Bayesian approach using an approximation of the range error distribution is better. Other localisation algorithms could also be used for this step.
- node location (a two-dimensional array for two-dimensional localisation) is exchanged between nodes, then the marginal location distribution of the mobile node can be found from the range error distribution and the location distribution of the neighbours. This is only practical in centralised tracking as the amount of information to be exchanged between neighbours is in general too large.
- a solution for distributed implementation is to exchange summarised information such as the mean and covariance of the node location.
- the range error can be approximated by a Gaussian distribution
- the weighted least squares algorithm can be used to localise the node in which the weight given to any particular measurement is determined by the covariance of the estimate of the location of the neighbour.
- the range error distribution can be adjusted according to the uncertainty in the neighbour location estimate.
- the width of the base rectangular shape is adjusted according to the uncertainty in the location of the neighbour. That is the possible region where the mobile node could lie is expanded according to the covariance of the location estimate of the neighbour.
- the state update step 612 uses the current location measurement to update the current node state.
- the advantage of the locate and track method is that the measurement equation given in (14) is linear in terms of the state x k .
- the alman filter is used for the state update step 612 and the track prediction step 604.
- a nonlinear filtering algorithm such as the extended Kalman filter or the unscented Kalman filter is used for the state update step 612 and track prediction step 604.
- a particle filtering algorithm is used for the state update filter step 612 and the track prediction step 604.
- a preferred embodiment of the locate and track approach that is suitable for cooperative tracking in environments with a range error distribution that can be approximated by a Gaussian distribution (possibly with heavy tails), such as typically occurs with line of sight measurements indoors or outdoors, is now described with several variants.
- the range error model is a Gaussian
- the motion model is the nearly constant velocity model
- the Kalman filter is used and the localise step uses a robust least squares algorithm so as to minimise the errors caused by outlier range measurements.
- the exchanged node information is only node location.
- the locate step uses the robust least squares algorithm.
- the exchanged node information is the mean and covariance of the node location, and a weighted version of the robust least squares algorithm is used to take into account the neighbour location reliability as provided by the covariance.
- a variant of this embodiment, shown in Figure 7, is a process 700 that uses the robust least squares algorithm for the localise step and the Interacting Multiple Model (IMM) estimator as the preferred tracking method to update the state of the nodes.
- the IMM estimator uses multiple Kalman filters, one for each motion model, and determines a probability that each model is active.
- the steps shown in Figure 6 are largely in common with those in Figure 7, with the processing in step 604 shown in more detail in step 704, and the processing in step 612 shown in more detail in step 716.
- the input data 702 is used to predict the location of the node in step 708 using each of the Kalman filters, and prior to running the Kalman filter prediction step the IMM interaction step 706 mixes the current states of the Kalman filters as is know in the art of using IMM estimators. Localisation then occurs in step 710 using the robust least squares technique. Step 712 tests whether the localisation was successful. If not, the predicted state, covariance, and model probability is set as the current state, covariance and model probability respectively in the IMM estimator in step 714. If so, then the track is updated using the IMM estimator in step 716 and assigned to the output in step 724. Each Kalman filter produces an updated state estimate in step 718, and the probability that each model is active is updated in step 720.
- Each node internally propagates a mean and covariance matched to each of the models, and a probability that specifies how likely a particular model is active.
- the states of the multiple Kalman filters are also combined in step 722 to provide a single estimate of the state of the node.
- the exchanged node information is the node location and the IMM estimator consists of two nearly-constant velocity models: one with low process noise intensity and the other with high process noise intensity.
- Another variant of this embodiment is for the exchanged node information to consist of the mean and covariance of the node location, and a weighted version of the robust least squares algorithm is used for the localise step to utilise the neighbour location covariance.
- step 310 of Figure 3 An embodiment of the tracking method in step 310 of Figure 3 is shown in the process flow 800 of Figure 8, where the Bayesian filter uses the direct range fusion method.
- the input 802 is range and exchanged information for each neighbour node.
- This tracking method first predicts, in step 804, the track of the node by applying the prediction step of the filtering algorithm it uses, based on the selected motion model. If, in step 806, there are no valid range measurements to a neighbour at a known location then, in step 808, the current state is the predicted state. If there are any valid measurements, then the state update filter uses the measurement to update the current state in step 810. The current state information is set to be the updated state information in step 812. An output is provided in step 814.
- the measurement equation for the filtering algorithm in the direct range fusion method is given by equations (11) and (12), since the range information is directly used to update the state.
- the measurement equation is nonlinear.
- the state equation can be assumed to be linear, a nonlinear filtering algorithm is required.
- the direct range fusion method uses the extended alman filter or the unscented Kalman filter.
- a particle filter is used in the direct range fusion method.
- the mean and covariance of the node location is exchanged between the nodes. If the state of the mobile node at time k is ⁇ k and that of its neighbour i is and if they are known exactly, then the measured range between these two nodes can be written as z k ⁇ H(x k -x' k ) ⁇ +cj k (16)
- This technique is suitable for range measurement errors that can be approximated by Gaussian distributions. If the measurement noise is non-Gaussian and for example, a rectangular distribution is used as the approximate measurement noise distribution, as in the locate and track method, the width of the base shape of the rectangular approximation can be varied according to the covariance of the neighbour location. Such incorporation of the non-Gaussian range error distribution would necessitate the use of a particle filter.
- a preferred embodiment of the direct range fusion method that is suitable for difficult radio propagation environments with biased and non-Gaussian measured range error distribution, such as occurs within buildings and underground, is now described with variants.
- the embodiment uses the nearly constant velocity motion model, particle filter and the triangular likelihood function shown in Figure 5 is assumed for measurement error model.
- the exchanged information consists of node location.
- the exchanged information consists of the node location mean and covariance and measurement Equation 19 is used. To do this one could convolve the assumed range error distribution with the Gaussian distribution of the node location error, however a simpler approximation is to stretch the assumed distribution (such as the triangular approximation) by the standard deviation of location error.
- a variation of this embodiment is to use multiple motion models with the multiple model particle filter (MMPF) as the preferred Bayesian tracking filter to update the state of the nodes.
- MMPF multiple model particle filter
- the MMPF uses two nearly constant velocity models one with a low process noise intensity and the other with a high process noise intensity.
- This embodiment can use only exchanged node location or exchanged node location mean and covariance with the modified measurement equation as described in the previous paragraph.
- the MMPF algorithm first samples the motion model based on a transition probability matrix (TPM).
- TPM transition probability matrix
- the TPM consists of various probabilities of the possible switch between the different motion models.
- the th element of the matrix is the probability that the model switch from i to j happening, that is,
- r k denotes the active model at time k .
- TPM the following TPM
- the motion model state values are sampled in step 906.
- the prior distribution is used to sample the state in step 908. That is, the i th sample is drawn according to (22)
- the preferred embodiment sets the sampled particles as the updated particles with the weights of these particles unmodified in step 912. Otherwise, the weights of the sampled particles are updated using the range error model in step 91 .
- the weight of thei th particle is updated according to
- step 916 Only a few particles will have non-zero weights after a few measurement updates, which is known as the sample degeneracy problem. This is avoided by resampling the updated particles in step 916; that is, replicating the particles that have higher weights and discarding the particles that have lower weights through a probabilistic selection process in step 918. In the preferred embodiment a known systematic resampling procedure is used, however many other resampling algorithms also be applied. An output is provided in step 920.
- All nodes are transceivers with processing hardware that have the ability to measure range to neighbouring nodes and form a data network. Range can be measured, for example, based on received signal strength (RSS) or time of arrival (TO A) using radio or acoustic signals.
- RSS received signal strength
- TO A time of arrival
- One example of a hardware arrangement measuring TOA is given in commonly-owned International Publication No. WO 2010/000036 Al, the entire contents of which are incorporated herein by reference.
- FIG 10 is a schematic block diagram of an anchor node device 1000.
- An antenna 1002 is coupled to a radio frequency transmitter (Tx) circuit 1004 that outputs a data stream from a computational processor 1006 containing information about the node's location and range to neighbouring nodes.
- Location in one embodiment, is known from a GPS device 1012 providing such data to the processor 1006.
- a range measurement module 1008 measures the range to neighbouring devices, and may accomplish this using the transmitter 1004.
- FIG 11 is a schematic block diagram of a mobile node device 1100.
- An antenna 1002 is coupled to a radio frequency transceiver (Tx/Rx) circuit 1104 that provides a received data stream containing information about the node's range to and location of neighbouring nodes to a tracking filter 1110 implemented within a computational processor 1106, and transmits a data stream containing information about itself for use by neighbouring nodes.
- FIG 12 is a schematic block diagram of another form of mobile node device 1200.
- the device 1200 has a transmitting antenna 1202, a transmission (Tx) circuit 1204 and a range estimation module 1206, implementing the methods described above.
- Tx transmission
- the device 1200 transmits range measurements to another computational device that executes the tracking filter for said mobile node.
- the device 1200 transmits range measurements to one of a plurality of access points (not shown), and the tracking filters (not shown) are accessed via said access points.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
- Navigation (AREA)
Abstract
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2012203094A AU2012203094B2 (en) | 2010-11-19 | 2012-01-11 | Tracking location of mobile devices in a wireless network |
US13/885,881 US9197996B2 (en) | 2010-11-19 | 2012-01-11 | Tracking location of mobile device in a wireless network |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2010905143A AU2010905143A0 (en) | 2010-11-19 | Tracking location of mobile devices in a wireless network | |
AU2010905143 | 2010-11-19 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012065233A1 true WO2012065233A1 (en) | 2012-05-24 |
Family
ID=46083433
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/AU2012/000018 WO2012065233A1 (en) | 2010-11-19 | 2012-01-11 | Tracking location of mobile devices in a wireless network |
Country Status (2)
Country | Link |
---|---|
AU (1) | AU2012203094B2 (en) |
WO (1) | WO2012065233A1 (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104936147A (en) * | 2015-05-08 | 2015-09-23 | 中国科学院上海微系统与信息技术研究所 | Positioning method based on building layout constraint under complex indoor environment |
EP2952925A1 (en) * | 2014-06-06 | 2015-12-09 | Technische Universität Graz | Method, device and system for indoor localization and tracking using ultra-wideband radio signals |
CN105491661A (en) * | 2015-12-10 | 2016-04-13 | 上海电机学院 | Improved Kalman filtering algorithm-based indoor positioning system and method |
CN105517150A (en) * | 2015-12-29 | 2016-04-20 | 江南大学 | Particle swarm positioning algorithm based on adaptive differential |
FR3058796A1 (en) * | 2016-11-17 | 2018-05-18 | Thales | METHOD AND SYSTEM FOR COLLABORATIVE GEOLOCATION |
US10989531B2 (en) | 2014-08-15 | 2021-04-27 | Commonwealth Scientific And Industrial Research Organisation | Method of setting-up a range-based tracking system utilizing a tracking coordinate system |
WO2021082571A1 (en) * | 2019-10-29 | 2021-05-06 | 苏宁云计算有限公司 | Robot tracking method, device and equipment and computer readable storage medium |
CN113709662A (en) * | 2021-08-05 | 2021-11-26 | 北京理工大学重庆创新中心 | Ultra-wideband-based autonomous three-dimensional inversion positioning method |
CN114537477A (en) * | 2022-03-01 | 2022-05-27 | 重庆交通大学 | Train positioning and tracking method based on TDOA |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7076259B2 (en) * | 2003-03-13 | 2006-07-11 | Meshnetworks, Inc. | Real-time system and method for improving the accuracy of the computed location of mobile subscribers in a wireless ad-hoc network using a low speed central processing unit |
WO2006073816A2 (en) * | 2004-12-30 | 2006-07-13 | Meshnetworks, Inc. | System and method for determining the mobility of nodes in a wireless communication network |
US7263349B2 (en) * | 2002-03-12 | 2007-08-28 | Qualcomm Incorporated | Velocity responsive time tracking |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2158497A4 (en) * | 2007-06-19 | 2012-03-21 | Magna Electronics Inc | Mobile control node system and method for vehicles |
-
2012
- 2012-01-11 WO PCT/AU2012/000018 patent/WO2012065233A1/en active Application Filing
- 2012-01-11 AU AU2012203094A patent/AU2012203094B2/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7263349B2 (en) * | 2002-03-12 | 2007-08-28 | Qualcomm Incorporated | Velocity responsive time tracking |
US7076259B2 (en) * | 2003-03-13 | 2006-07-11 | Meshnetworks, Inc. | Real-time system and method for improving the accuracy of the computed location of mobile subscribers in a wireless ad-hoc network using a low speed central processing unit |
WO2006073816A2 (en) * | 2004-12-30 | 2006-07-13 | Meshnetworks, Inc. | System and method for determining the mobility of nodes in a wireless communication network |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2952925A1 (en) * | 2014-06-06 | 2015-12-09 | Technische Universität Graz | Method, device and system for indoor localization and tracking using ultra-wideband radio signals |
WO2015185668A1 (en) * | 2014-06-06 | 2015-12-10 | Technische Universität Graz | Method, device and system for indoor localization and tracking using ultra-wideband radio signals |
US10989531B2 (en) | 2014-08-15 | 2021-04-27 | Commonwealth Scientific And Industrial Research Organisation | Method of setting-up a range-based tracking system utilizing a tracking coordinate system |
CN104936147A (en) * | 2015-05-08 | 2015-09-23 | 中国科学院上海微系统与信息技术研究所 | Positioning method based on building layout constraint under complex indoor environment |
CN105491661B (en) * | 2015-12-10 | 2019-01-25 | 上海电机学院 | Indoor locating system and method based on improved Kalman filter algorithm |
CN105491661A (en) * | 2015-12-10 | 2016-04-13 | 上海电机学院 | Improved Kalman filtering algorithm-based indoor positioning system and method |
CN105517150A (en) * | 2015-12-29 | 2016-04-20 | 江南大学 | Particle swarm positioning algorithm based on adaptive differential |
FR3058796A1 (en) * | 2016-11-17 | 2018-05-18 | Thales | METHOD AND SYSTEM FOR COLLABORATIVE GEOLOCATION |
WO2021082571A1 (en) * | 2019-10-29 | 2021-05-06 | 苏宁云计算有限公司 | Robot tracking method, device and equipment and computer readable storage medium |
CN113709662A (en) * | 2021-08-05 | 2021-11-26 | 北京理工大学重庆创新中心 | Ultra-wideband-based autonomous three-dimensional inversion positioning method |
CN113709662B (en) * | 2021-08-05 | 2023-12-01 | 北京理工大学重庆创新中心 | Autonomous three-dimensional inversion positioning method based on ultra-wideband |
CN114537477A (en) * | 2022-03-01 | 2022-05-27 | 重庆交通大学 | Train positioning and tracking method based on TDOA |
CN114537477B (en) * | 2022-03-01 | 2023-06-09 | 重庆交通大学 | Train positioning tracking method based on TDOA |
Also Published As
Publication number | Publication date |
---|---|
AU2012203094B2 (en) | 2016-08-04 |
AU2012203094A1 (en) | 2013-06-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9197996B2 (en) | Tracking location of mobile device in a wireless network | |
AU2012203094B2 (en) | Tracking location of mobile devices in a wireless network | |
Jourdan et al. | Monte Carlo localization in dense multipath environments using UWB ranging | |
Li et al. | New efficient indoor cooperative localization algorithm with empirical ranging error model | |
Sathyan et al. | Fast and accurate cooperative tracking in wireless networks | |
Wang et al. | Toward robust indoor localization based on Bayesian filter using chirp-spread-spectrum ranging | |
US9217788B2 (en) | Location and tracking system | |
Kulmer et al. | Anchorless cooperative tracking using multipath channel information | |
Zafari et al. | A novel Bayesian filtering based algorithm for RSSI-based indoor localization | |
JP2008128726A (en) | Positioning system, device and method using particle filter | |
Addesso et al. | Adaptive localization techniques in WiFi environments | |
Froehle et al. | Cooperative multipath-assisted indoor navigation and tracking (Co-MINT) using UWB signals | |
Monica et al. | Impact of the number of beacons in PSO-based auto-localization in UWB networks | |
chander Janapati et al. | Enhanced mechanism for localization in wireless sensor networks using PSO assisted extended Kalman filter algorithm (PSO-EKF) | |
Zdruba et al. | Monte Carlo sampling based in-home location tracking with minimal RF infrastructure requirements | |
Lee et al. | IEEE 802.15. 4a CSS-based mobile object locating system using sequential Monte Carlo method | |
Benkouider et al. | Divided difference Kalman filter for indoor mobile localization | |
Van de Velde et al. | Cooperative hybrid localization using Gaussian processes and belief propagation | |
Aggarwal et al. | Joint sensor localisation and target tracking in sensor networks | |
Heinemann et al. | RSSI-based real-time indoor positioning using zigbee technology for security applications | |
Vaghefi et al. | Improving mobile node tracking performance in NLOS environments using cooperation | |
Mankotia et al. | Error minimization in bluetooth based indoor localization of a mobile robot using Cuckoo search algorithm | |
Liu et al. | A robust method of fusing ultra-wideband range measurements with odometry for wheeled robot state estimation in indoor environment | |
Shao et al. | The research of Monte Carlo localization algorithm based on received signal strength | |
Barcelo et al. | Advances in indoor location |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 12721156 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2012203094 Country of ref document: AU Date of ref document: 20120111 Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13885881 Country of ref document: US |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 12721156 Country of ref document: EP Kind code of ref document: A1 |