Methods for characterizing vehicles equipped with radio transmitters
By measuring radio signal characteristics using a passive detection system and employing mobility models and Bayesian inference, the applicability of existing radar detection technologies in identifying airborne radio transmitters is addressed, enabling accurate quantitative identification of vehicle types.
Patent Information
- Application Number
- CN202480090470.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-04-11
- Filing Date
- 2024-11-07
- Publication Date
- 2026-11-17
AI Technical Summary
Existing radar detection technologies are not applicable to detecting airborne radio transmitters, especially when the vehicle does not generate sufficient radiation reflection, making it difficult to effectively identify them. Furthermore, active devices have complex structures.
A passive detection system is used to determine the vehicle type by measuring the characteristics of radio signals emitted by the vehicle, calculating matching scores using mobility and relational models, and combining Bayesian inference and data-driven methods to provide quantified vehicle type information.
It enables accurate identification of airborne radio transmitters in various environments, reduces errors caused by single-vehicle type characteristics, and provides quantitative indications for multiple vehicle types.
Smart Images

Figure CN123125637A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for characterizing a vehicle equipped with a radio transmitter, and also to a detection system suitable for performing this method. Priority is claimed to European Patent Application No. EP24290008.2, filed April 11, 2024, the contents of which are incorporated herein by reference. Background Technology
[0002] Radar is an active detection device capable of detecting aircraft or ships and providing information about their trajectories. However, radar detection is not suitable for all vehicles to be detected and for all environmental conditions, especially when the vehicle does not generate sufficient radiation reflection. Furthermore, active devices are complex in structure because they involve radiation sources and systems for directing radiation emission towards the vehicle to be detected or for generating scans of the geographic area to be monitored.
[0003] Today, most road vehicles and mobile vehicles such as aircraft, drones (especially unmanned aerial systems also known as UAS), and road vehicles are equipped with onboard radio transmitters to send radio signals to external receivers, regardless of the type of the external receiver. For example, cars may be equipped with radio transmitters to communicate with roadside units, cellular base stations, control servers, etc. Radio signals propagate in free space and can therefore be detected by receivers specifically designed for monitoring, such as for statistical traffic analysis or security purposes, especially in areas surrounding sensitive locations.
[0004] Passive detection devices are designed to detect radio signals emitted from airborne vehicles equipped with radio transmitters that generate these radio signals. A passive detection system is then needed that can characterize the detected vehicle as an example of air traffic, specifically indicating whether the detected vehicle might be a jet aircraft, helicopter, unmanned aerial system, etc. A passive detection system is also needed that is configured to provide quantitative indications associated with each of the various vehicle types. In fact, this quantitative information reduces errors that might arise from characteristics provided individually for each vehicle type.
[0005] This invention aims to meet this need. Summary of the Invention
[0006] To this end, a first aspect of the present invention provides a method for a vehicle equipped with a transmitter, wherein the transmitter transmits radio signals, and at least one receiver is capable of measuring at least one characteristic of the radio signals. The method of the present invention includes the following steps: - Obtain a set of multiple mobility models, each mobility model determining at least one state value or at least one state change value that a vehicle equipped with the transmitter may have; - Obtain a relational model that connects the state values that the vehicle equipped with the transmitter may have to the values of at least one feature of the radio signal; - Collect at least one measurement result of at least one characteristic of the radio signal from the at least one receiver; - For each mobility model in a subset containing at least two mobility models from the mobility models, calculate a matching score associated with at least one collected measurement result to match the mobility model using the relationship model; and - Identify a mobility model, referred to as the best-fit mobility model, from the mobility models in the subset, which corresponds to the highest matching score among the matching scores.
[0007] Therefore, the method of the present invention combines information derived from measurements with an initially available mobility model. The method provides multiple matching scores associated with multiple mobility models, thereby providing the operator with quantitative information about a transmitter-equipped vehicle that satisfies each of the multiple mobility models.
[0008] Furthermore, the method of the present invention can be implemented using a passive detection system.
[0009] If each mobility model in the subset is associated with a vehicle type such that at least one state value or at least one state change value of the mobility model is compatible with a device of the vehicle type equipped with a transmitter, the method further includes assigning the vehicle type of the best-fit mobility model to the vehicle equipped with the transmitter.
[0010] Therefore, the response to the vehicle type provided to the operator by the method of the present invention is also represented as multiple quantized values in order to reach a maximum value, but without being distorted as a response composed of a single vehicle type.
[0011] The vehicle type associated with at least one of the mobility models may belong to a list including: automobiles, trucks, aircraft (especially passenger planes, light passenger planes or jet fighters), helicopters, unmanned aerial systems (especially quadcopters), orbital satellites, bicycles, unicycles, mobile robots, walking manned vehicles, ships, and submarines.
[0012] In various implementations of the method of the present invention, the measured characteristics of the radio signal include at least one of the following: If radio power is received, then the relationship model may include a path loss model; The direction of receiving the radio signal, and then the relationship model may include at least one detection angle range, such as azimuth and / or elevation. At least one radio signal detection time, and then the relational model may include at least one detection delay range; and The Doppler frequency shift value, and therefore the relationship model may include a Doppler frequency shift model.
[0013] Obviously, the method of the present invention can be combined with other types of radio signal measurements.
[0014] In a first type of implementation of the method of the present invention, Bayesian inference is used to calculate a matching score for each mobility model in the subset. Here, the matching score is the probability value that a vehicle equipped with a transmitter satisfies the mobility model.
[0015] In other implementations of the method of the present invention, a matching score is calculated for each mobility model in the subset using a metric suitable for quantifying the difference between at least one collected measurement result and at least one value of radio signal characteristics inferred from the mobility model through the relational model.
[0016] Generally, for this invention, each mobility model can be obtained from a database.
[0017] The method of the present invention may further include: comparing each of the calculated matching scores with a mismatch threshold; and generating a warning signal if all the calculated matching scores are below the mismatch threshold.
[0018] Similarly, the method of the present invention may further include comparing a matching score calculated for the best-fit mobility model with a positive identification threshold. Then, if the matching score calculated for the best-fit mobility model is higher than the positive identification threshold, the method of the present invention may further include supplementing the best-fit mobility model with at least one additional data point corresponding to or inferred from at least one measurement of the collected radio signals. This supplementation of the best-fit mobility model can be readily implemented when the best-fit mobility model is data-driven. However, this may also be applicable to other mobility model types.
[0019] Similarly, generally, for the present invention, each mobility model can be applied to determine the variation of at least one characteristic of a radio signal over time using a relational model. In this case, the step of collecting at least one measurement result may include collecting multiple consecutive measurements associated with various time values.
[0020] Specifically, each mobility model can be at least one of the following types: Data-driven, in which the mobility model includes a series of state values that appear at various time points; Physics-based equations that connect state values to various time values; An empirical equation, specifically determined from collected measurements using a neural network prior to implementing the method, connects state values to individual time values; and A parameterized type where the state value changes over time depending on one or more model parameters.
[0021] If the vehicle is a moving vehicle, and the transmitter transmits radio signals during the vehicle's continuous movement, and the value of the radio signal characteristics depends on the vehicle's position or speed, then at least one state value may include at least one of the following values: The possible position values of the vehicle at each time value. The possible speed values of the vehicle at various time points, and At each time value, the possible value of at least one angle between the direction tangent to the trajectory of the vehicle and the reference direction.
[0022] Obviously, any state parameter type can be used to implement this invention.
[0023] In this scenario involving moving vehicles and continuous measurements, the method does not distort the results by assuming that a single mobility model corresponds to the vehicle, whereas such distortion would occur if a trajectory were first inferred from the measurement results and then compared to the mobility model.
[0024] When collecting continuous measurement results from at least one receiver while the vehicle is moving, the method of the present invention may include: for each mobility model in a subset: -Infer information about the state of the pixel vehicle at a time value corresponding to or later than at least one newly collected measurement from the mobility model; and - An updated matching score related to the vehicle is calculated using the information and the at least one newly collected measurement result to match the mobility model.
[0025] Possibly, the updated matching score for each mobility model in the subset can be calculated by combining the information, at least one probability value associated with the at least one newly collected measurement result, and the matching score for the vehicle, to match the same mobility model available prior to the at least one newly collected measurement result.
[0026] While the vehicle is still moving and continuous measurements are being taken, the steps of collecting at least one further measurement result, inferring further information about the state of the vehicle at a time value corresponding to or later than the further measurement result from each mobility model in the subset, and calculating a further updated matching score associated with the vehicle using the further information and the further measurement result to match each mobility model in the subset are performed as a loop repeated while the vehicle is moving.
[0027] To simplify the calculation of the updated matching score, it can be assumed that the measurement noise effective for each of the continuously collected measurements is independent of each other.
[0028] Generally, in this invention, radio signals transmitted from the vehicle can be detected by multiple time-synchronized receivers, each of which performs a measurement of at least one characteristic of these radio signals. A matching score can then be calculated by a computing node shared by the multiple time-synchronized receivers. Using multiple receivers in this way provides more measurement information, thereby allowing for a more accurate characterization of the vehicle provided to the operator.
[0029] Similarly, generally, for the present invention, if at least one of the mobility models in the subset is of a parameterized type, the corresponding matching score can be calculated separately for multiple test values of each model parameter, and one of the test values of each model parameter can be selected based on the highest value among the matching scores.
[0030] Similarly, generally, for this invention, matching scores can be calculated using prior knowledge values associated with mobility models in the subset, respectively. These prior knowledge values may originate from knowledge acquired prior to executing the method of this invention. In this way, faster convergence is achieved for the matching scores that are repeatedly updated for each mobility model in the subset, because the initial calculation entries already contain more information when available.
[0031] Finally, the method of the present invention can be implemented for many applications, including monitoring geographic areas, particularly statistically analyzing traffic currently occurring in geographic areas, or characterizing mobile vehicles equipped with transmitters that are currently intruding into geographic areas.
[0032] Another object of the present invention is to provide a detection system comprising: - At least one receiver, said at least one receiver being capable of measuring at least one characteristic of a radio signal; - At least one computing node; and - A storage unit, accessible by the at least one computing node, and configured to store mobility models, each of which determines at least one state value or state change value.
[0033] The detection system is configured to perform the method according to the first aspect of the invention, including its improvements and advantageous optional features.
[0034] Advantageously, for many applications, the detection system may also include a housing that accommodates the computing node and the storage unit, and the at least one receiver is connected to the computing node through the housing. Thus, a mobile system is obtained, enabling on-site installation and easy use of this detection system in any environment.
[0035] These and other features of the invention will now be described with reference to the accompanying drawings, which relate to preferred but non-limiting embodiments of the invention. Attached Figure Description
[0036] Figure 1 The operation of the detection system according to the present invention is illustrated schematically; and
[0037] Figure 2 The steps of the method according to the present invention are shown. Detailed Implementation
[0038] like Figure 1As shown, a detection system suitable for implementing the present invention includes one or more radio receivers, such as three radio receivers labeled 1a, 1b, and 1c. In the case of multiple radio receivers, the multiple radio receivers may, but are not necessarily, time-synchronized. The detection system also includes a computing node 2, denoted COMP., and a storage unit 3, denoted DATABASE. The computing node 2 and storage unit 3 may be enclosed within a housing 4, and the connection to the receivers 1a-1c extends through a dedicated channel in the housing 4. Each receiver 1a-1c is configured to perform measurements on radio signals detected by that receiver. For example, each receiver may measure the radio power, detection direction, detection time, and / or Doppler shift value of each detected radio signal. Thus, for a radio signal originating from a single wireless power source but detected by the system's receivers 1a-1c, as a result of each measurement, each of these receivers assigns a corresponding detection power value, a corresponding azimuth and elevation value, a corresponding detection time value, and / or a corresponding Doppler shift value to that radio signal. These measurement results are transmitted to the computing node 2 for processing according to the present invention. For this purpose, computing node 2 can access the database recorded in storage unit 3. This detection system can be dedicated to continuously monitoring traffic currently existing within a geographic area. For example, the detection system can be dedicated to monitoring road traffic, air traffic, or maritime activities in a specific area, such as for road traffic statistics purposes, or to identifying intrusions in airspace near sensitive locations such as industrial sites.
[0039] exist Figure 1 In the example, an unmanned aerial system, typically referred to as a flying drone 10, has penetrated the monitored geographic area along trajectory T. This intrusion must be characterized, and it can be compared with reported traffic to determine whether it should be considered a potential threat to a sensitive location. Assume that drone 10 carries a radio transmitter 11, which transmits radio signals S to, for example, a radio station (not shown) during drone flight to allow flight control. As long as drone 10 is within the detection range of these receivers, the radio signals S can be repeatedly detected and measured by each of the radio receivers 1a-1c. Specifically for the statistical and intrusion identification purposes described above, characterizing the vehicle type of the radio transmitter 11 from the transmitted radio signals S is an issue. In the example shown, the vehicle type is a quadcopter drone, but the vehicle type could be any vehicle carrying a radio transmitter.
[0040] To implement this invention, the database in storage unit 3 contains multiple mobility models, each corresponding to an identified vehicle type. Providing such a mobility model library corresponds to... Figure 2Step S1 in the specification. In the overview section of this document, this content of the database is referred to as a mobility model set. For example, one or more mobility models corresponding to the vehicle type of a civil airliner may be obtained; other mobility models may correspond to fighter jets, some to helicopters, and possibly others to various types of unmanned aerial vehicles. Each mobility model may be one of several predefined types, including physics-based types, data-driven types, and empirical equation types, and may be parameterized in each case. It is important to note that each mobility model may not be associated with an identified vehicle type, but simply with an identified utility model.
[0041] Physics-based mobility models rely on equations that determine the movement of vehicles of the type of vehicle associated with the mobility model. For example, the mass value of the vehicle can constitute a parameter of the mobility model. Different mass values may correspond to different numbers of passengers on the vehicle. The general formula for such a physics-based mobility model is:
[0042] in, The vehicle's state is represented by a hidden variable containing all the characteristics describing its behavior. u represents inputs such as forces and actions applied to the vehicle, y represents outputs—measurable quantities that provide partial or complete information about the vehicle's state—and t represents time. All these quantities change over time t. f is a mathematical function that determines the time-varying nature of the vehicle's state x based on physical rules. This function f actually constitutes the mobility model described in the overview section of this specification. h is another mathematical function that determines the measured quantities based on the current vehicle state x and the input u. The h function and the y output are possible implementations of the relational model and radio signal characteristics, respectively, as described in the overview section of this specification. Most commonly, but not limited to, the vehicle's state consists of its instantaneous position and velocity in three-dimensional space. In the case of complex vehicle structures, other vehicle state parameters, such as the vehicle's angular orientation (if relevant for any reason), may also be added. In the context of this invention, y may include the radio power of the radio signal S measured by each radio receiver, the Doppler shift of these signals S detected by each radio receiver, and the detection angle and / or detection time at each of the radio receivers 1a-1c. In the case of measured radio power, the h function includes the physical laws governing radio signal propagation, including propagation delay and reduction in radio signal power, as a function of the separation distance between the vehicle and each radio receiver. In the case of measured Doppler shift, the h function includes the physical laws governing the Doppler effect related to radial velocity. When the mobility model is parameterized, at least one model parameter is involved in the f function, and at least one model parameter may also be involved in the h function.
[0043] In practical calculations, discretization is applied to time, so the above formula for the physics-based mobility model becomes:
[0044] in, and Two consecutive times and The corresponding value for the vehicle's status at that location. exist Afterwards. Perhaps, but not necessarily, radio receivers 1a-1c in time... and New measurements have been taken between the two. Therefore, k is an integer index indicating the continuous updates of the probability values, and f k and h k These are functions corresponding to the previous f-function and h-function, respectively, but the changes in the functions in the time discretization formula reflect the time dependence of the f-function and h-function. When constant, the time interval... This can be referred to as the sampling time. k Indicates time and The results of the measurements performed between, and This indicates input that is valid during the same time interval.
[0045] Another well-known formula for physics-based mobility models combines the transfer function with the Laplace domain.
[0046] The data-driven mobility model consists of a list of values for vehicle state x, each value corresponding to a time value t. n : x n (t n (This is related to) a well-known algorithm that allows scanning such a list from variable initial states and interpolating between the listed state values.
[0047] Similarly, the mobility model could consist of empirical equations determined based on measurements taken beforehand on the vehicle. These empirical equations connect state values to their corresponding time values. This can be determined using a neural network based on the state measurements and their corresponding times.
[0048] Simple mobility models relate to the constant velocity or constant acceleration motion of a vehicle. More complex mobility models may specifically involve bicycles, unicycles, or quadcopter drones, when considering not only position and velocity coordinates but also the vehicle's azimuth and / or internal degrees of freedom.
[0049] Figure 2Step S10 involves providing prior knowledge that may be applicable to the mobility models in step S1. This prior knowledge can be represented as initial corresponding probability values for the movement of vehicle 10, matching each of the mobility models in the entire set stored in the database. For example, this prior knowledge can come from statistical knowledge about traffic occurring in the monitored geographic area, where proportions are associated with each vehicle type or several mobility models. Prior knowledge may also come from an initial analysis of some measurements, for example, excluding vehicle types such as cars when the initial measurements demonstrate the presence of a vertical velocity component. When the prior knowledge is limited to assuming the movement of vehicle 10 matches one of N identified mobility models without further information, where N is an integer greater than 1, the initial probability value assigned to each of these identified mobility models can be equal to 1 / N. The mobility model associated with a non-zero initial probability value in step S10 is denoted as M. i , where i is an integer index, and forms a subset of all the content in database 3. This is related to the mobility model M. i The associated initial probability value is represented as p(M) i ).
[0050] Then, the method of the present invention can be performed as a loop S2, which is repeatedly executed during the movement of vehicle 10, with each execution of the loop passing an updated value of the probability of movement of vehicle 10 to match each mobility model M. i For example, loop S2 can be executed at constant time intervals, or each time radio receivers 1a-1c provide a new measurement of the defined number of radio signals S. Therefore, the integer index k also identifies the number of iterations of loop S2. (Referring to...) Figure 2 In the described implementation example, loop S2 includes sub-steps S2-1 to S2-4, and optional sub-steps S2-5 to S2-8. The result of each execution of loop S2 is more accurately determined at each given time t1, ..., t2. k The obtained measurement results z1, ..., z k In this case, the updated value of the probability of vehicle 10 moving is used to match each mobility model M. i Each of these probabilities is denoted as . .
[0051] probability It can be calculated in the following way:
[0052] in, It is the sequence of measurement results z1, ..., z k During measurement time The probability distribution, Given a mobility model M i The state sequence x1(t1), ..., x k (t k The probability of ), and corresponding to the so-called information introduced in the overview section of this specification; and It is a given sequence of states And obtain the measurement result sequence The probability is based on the state measurement mapping, and also takes into account the measurement noise caused by sensor defects and radio propagation from transmitter 11 to each of receivers 1a-1c.
[0053] The first transformation in the above derivation involves the Bayesian identity for conditional probability, the second transformation involves the independence between the mobility model and measurement time, and the third transformation involves... The fourth transformation, relative to the marginalization of state values, involves a Bayesian identity for the assemblage probability, and the fifth transformation is derived from measurements independent of the mobility model and measurement time when the state is known. Because With mobility model M i It is irrelevant and therefore can be ignored, such that:
[0054] Then, two simplifying assumptions can be made. The first simplifying assumption is the following Markov assumption, denoted as: When applied to each mobility model M i This will lead to:
[0055] When the fundamental formulas of a physics-based model do not obey the Markov assumptions, a new formula for the model can be given through some transformations, where the new states satisfy the Markov assumptions. This is called the state distribution, and it is the study of correlation distributions. The state values... Connected to The function that allows information to be propagated from one state value to the next is called the state transfer function. The state transfer function is generally fully described by the mobility model. Then, when the relevant noise distribution is known and easily transformed, the state distribution can be calculated analytically. However, when the noise distribution is difficult to calculate analytically, numerical methods provide a more tractable approximate representation of the state distribution.
[0056] The second assumption relates to the uncorrelatedness of continuous measurements. This can be expressed as:
[0057] in, This assumes the vehicle's state is x. k Obtaining measurement results z under the following circumstances k The probability of this is usually referred to as the observed distribution.
[0058] For example, the measurement may involve the Received Signal Strength Indicator, commonly known as RSSI. The measurement result z is then expressed using the following formula. k Position p connected to vehicle 10 k :
[0059] in, This is the transmission power of transmitter 11. It is possible that it has a standard deviation σ w Shaded noise defined by a zero-centered Gaussian variable: , and It is a path loss parameter. This is the location selected from radio receivers 1a-1c. Assume... A, B and Given the position of vehicle 10 Measurement results It has a probability distribution:
[0060] The same principle applies to any state-observation mapping that contains some randomness. If the probability distribution of the random components contributing to the measurement is known, the observation distribution can be computed using a function transformation (e.g., additive Gaussian noise) where possible and simple. When the noise probability distribution is unknown or difficult to model, numerical methods again allow for the computation of a representation of the observation distribution by taking samples from the noise distribution and transforming them in the domain of the measurement through the state-observation mapping.
[0061] As another example, this measurement might be related to the Doppler frequency shift present in the received signal. The measurement result z is then expressed using the following formula. k The velocity vector v connected to vehicle 10 k :
[0062] in, It is the measurement time Doppler frequency shift at that location It is the transmitter at the time of transmission. The frequency of the signal sent at that location, It's the speed of light. It is the transmitter at the time of transmission. speed, It is based on the standard deviation. : The Doppler frequency shift noise is defined by a zero-center Gaussian variable. Therefore, the relational model for this Doppler frequency shift measurement can be:
[0063] Assuming the Markov assumption and the independent measurement assumption, given k collected measurements, vehicle movement can be obtained to match the mobility model M. i The probability of updating:
[0064] The last transformation described above is achieved by separating the factors corresponding to j=k and identifying all other factors j=1, ..., k-1 as To obtain.
[0065] This indicates that, under the above assumptions, the output generated by the previous execution of loop S2 can be used. To recursively calculate the probability value As a reminder, It is to form a mobility model M i The state distribution of information, This refers to the observed distribution of the system of interest. Therefore, the updated probability value... It is obtained as the result of the kth execution of loop S2, and the probability value It is obtained as the result of the previous execution of loop S2.
[0066] Sub-step S2-1 in loop S2 involves acquiring a new measurement result z from radio receivers 1a-1c, following the measurement results already used in previous executions of loop S2. k .
[0067] Substep S2-2 in loop S2 is to obtain the state value x. k Observed distribution of the function .
[0068] For each mobility model M in the subset i Execute sub-steps S2-3 and S3-4 in loop S2.
[0069] Sub-step S2-3 involves obtaining the state value x. k-1 and x k State distribution of the functions of both .
[0070] Sub-step S2-4 involves comparing the probability distributions obtained in sub-steps S2-2 and S2-3 with the probability values obtained in previous executions of loop S2. Combined to calculate probability values .therefore, This makes the movement of vehicle 10 match the mobility model M. i The probability value of the k-th update. The following three sub-steps can be optionally executed: Sub-step S2-5: If the mobility model M i At least one of them is parameterized, and if a value for each model parameter is calculated in substeps S2-4... The value is significantly higher than the values obtained for other model parameter values: Choosing to make The model parameter values are maximized, and subsequent execution of loop S2 is performed using only the selected model parameter values for the relevant mobility model. The model probability values can be calculated as follows:
[0071] Where θ is the model parameter, p(θ) is the relevant probability value, and This is the set of possible values for the model parameter θ. Preliminary steps are performed to identify the optimal model parameter values for fitting the measurements based on the mobility model under consideration, and then probability values are calculated as follows. This can simplify the process:
[0072] in, .question The solution can be determined using heuristic methods such as genetic algorithms to search for the optimal or near-optimal solution that matches the measurements for the corresponding mobility model. Sub-step S2-6: If in sub-step S2-4 all mobility models M of the subset i Calculated If the value is lower than the mismatch threshold, a warning signal is generated to indicate that vehicle 10 does not correspond to any mobility model used and may be considered an intruder. For example, the mismatch threshold can be equal to 0.1; Sub-step S2-7: The values obtained in sub-steps S2-4 and S2-5 correspond to the values obtained in sub-steps S2-4 and S2-5. The highest value in the mobility model M i The associated vehicle type, called the best-fit mobility model, is the most likely type for vehicle 10. Several mobility models M in the subset... i For the same vehicle type, for example, when combined with different motion types in each of these mobility models, the relevant mobility model probability value will be added. To obtain the updated probability value associated with that vehicle type; and Sub-step S2-8: If the values obtained in sub-steps S2-4 and S2-5 for the best-fit mobility model If the values are above the positive identification threshold, at least some of the measurements collected to supplement the best-fit mobility model in the database, as recorded in storage unit 3, are used. This mobility model supplementation is easier to implement when the best-fit mobility model is data-driven. For example, the positive identification threshold could be equal to 0.9.
[0073] Sub-steps S2-2 to S2-8 are executed by compute node 2.
[0074] While the above detailed description addresses the Bayesian implementation of the present invention, alternative implementations are also possible. For example, the well-known maximum likelihood estimation (MLE) method can be used for this purpose. Other frequency estimation methods are also suitable for the implementation of the present invention, such as confidence interval estimation, confidence region estimation, and error margin estimation. A matching score can be calculated according to the methods described above, which can be a probability, an information element (i.e., the logarithm of the log-likelihood function), a confidence level, a standard deviation, or a standard error.
[0075] Although the invention has been described and illustrated with the specific example of an airborne vehicle within a monitored geographic area, it can be applied to any radio-connected vehicle within any geographic area. In addition to air traffic, the invention may also be useful in applications to road and maritime traffic.
Claims
1. A method for characterizing a vehicle (10) equipped with a transmitter (11), wherein, The transmitter transmits a radio signal (S), and at least one receiver (1a-1c) is capable of measuring at least one characteristic of the radio signal. The method includes the following steps: - Obtain a set of multiple mobility models, each mobility model determining at least one state value or at least one state change value that a vehicle (10) equipped with the transmitter (11) may have; - Obtain a relational model that connects the state values that the vehicle (10) equipped with the transmitter (11) may have to the values of at least one feature of the radio signal (S); - At least one measurement result of at least one characteristic of the radio signal (S) collected from the at least one receiver (1a-1c); - For each mobility model in a subset containing at least two mobility models from the mobility models, calculate a matching score associated with at least one collected measurement result to match the mobility model using the relationship model; and - Identify a mobility model, referred to as the best-fit mobility model, from the mobility models in the subset, which corresponds to the highest matching score among the matching scores.
2. The method according to claim 1, wherein, Each mobility model in the subset is associated with a vehicle type such that at least one state value or at least one state change value of the mobility model is compatible with a device of the vehicle type equipped with a transmitter, and the method further includes assigning the vehicle type of the best-fit mobility model to the vehicle (10) equipped with the transmitter (11).
3. The method according to claim 1 or 2, wherein, The measured characteristics of the radio signal (S) include at least one of the following: Received radio power, and the relationship model includes a path loss model; The receiving direction of the radio signal (S), and the relationship model includes at least one detection angle range; At least one radio signal detection time, and the relational model includes at least one detection delay range; and The Doppler frequency shift value, and the relationship model includes a Doppler frequency shift model.
4. The method according to any one of the preceding claims, wherein, The matching score is calculated for each mobility model in the subset using Bayesian inference. The matching score is a probability value that the vehicle (10) equipped with the transmitter (11) satisfies the mobility model. Alternatively, the matching score can be calculated for each mobility model in the subset using a metric suitable for quantifying the difference between at least one collected measurement result and at least one value for the radio signal feature inferred from the mobility model through the relation model.
5. The method according to any one of the preceding claims, further comprising one of the following steps: Each of the calculated matching scores is compared to a mismatch threshold, and a warning signal is generated if all calculated matching scores are below the mismatch threshold; and The matching score calculated for the best-fit mobility model is compared with a positive identification threshold, and if the matching score calculated for the best-fit mobility model is higher than the positive identification threshold, the best-fit mobility model is supplemented with at least one additional data, which corresponds to or is inferred from at least one measurement of the collected radio signal (S).
6. The method according to any one of the preceding claims, wherein, Each mobility model is suitable for using the relationship model to determine the change of at least one characteristic of the radio signal (S) over time. Furthermore, the step of collecting at least one measurement result includes collecting multiple consecutive measurement results associated with each time value.
7. The method according to claim 6, wherein, Each mobility model is at least one of the following types: Data-driven, in which the mobility model includes a series of state values that appear at various time points; Physics-based equations that connect state values to various time values; An empirical equation, specifically determined from collected measurements using a neural network prior to implementing the method, connects state values to individual time values; and A parameterized type where the state value changes over time depending on one or more model parameters.
8. The method according to claim 6 or 7, wherein, The vehicle (10) is a mobile vehicle, and the transmitter (11) transmits the radio signal (S) during the continuous movement of the vehicle, wherein the value of the radio signal characteristic depends on the position or speed of the vehicle, and the at least one state value includes at least one of the following values: The possible position values of the vehicle (10) at various time values, The possible speed values of the vehicle (10) at various time values, and At each time value, at least one possible angle between the direction tangent to the trajectory of the vehicle (10) and the reference direction.
9. The method according to claim 8, wherein, While the vehicle (10) is moving, continuous measurement results are collected from the at least one receiver (1a-1c), and wherein the method includes: for each mobility model in the subset, -Information inferred from the mobility model regarding the state of the vehicle (10) at a time value corresponding to or later than at least one newly collected measurement result; and - An updated matching score related to the vehicle is calculated using the information and the at least one newly collected measurement result to match the mobility model.
10. The method according to claim 9, wherein, The updated matching score for each mobility model in the subset is calculated by combining the information, at least one probability value associated with the at least one newly collected measurement result, and a matching score for the vehicle, to match the mobility models available prior to the at least one newly collected measurement result.
11. The method according to claim 9 or 10, wherein, The steps of collecting at least one further measurement result, inferring further information about the state of the vehicle (10) at a time value corresponding to or later than the further measurement result from each mobility model in the subset, and calculating a further updated matching score associated with the vehicle using the further information and the further measurement result to match each mobility model in the subset are performed as a loop repeated while the vehicle is moving.
12. The method according to any one of the preceding claims, wherein, The matching score is calculated using prior knowledge values associated with the mobility models in the subsets, the prior knowledge values being derived from knowledge acquired prior to the execution of the method.
13. The method according to any one of the preceding claims, wherein the method is implemented for monitoring a geographic area, particularly for statistical analysis of traffic currently occurring in the geographic area, or for characterizing mobile vehicles equipped with transmitters currently intruding into the geographic area (10).
14. A detection system, the detection system comprising: - At least one receiver (1a-1c), said at least one receiver (1a-1c) being capable of measuring at least one characteristic of a radio signal (S); -At least one computing node (2); and - Storage unit (3), which can be accessed by the at least one computing node (2) and is configured to store mobility models, each of which determines at least one state value or state change value. The detection system is configured to perform the method according to any one of the preceding claims.
15. The detection system according to claim 14, the detection system further comprising a housing (4) accommodating the computing node (2) and the storage unit (3), and the at least one receiver (1a-1c) being connected to the computing node through the housing.
