Method and system for determining mean and variance of measurement data

Abstract

A computer-implemented method for determining a mean and variance of measurement data comprises the following steps performed by a computer hardware component: acquiring measurement data for a current time step using a sensor; updating the plurality of intermediate data for the previous time step based on the measurement data for the current time and based on the forgetting factor to obtain a plurality of intermediate data for the current time step; and determining an average value and a variance for the current time step based on the plurality of intermediate data for the current time step.

Description

Technical Field

[0001] This disclosure relates to methods and systems for determining the mean and variance of measurement data. Background Technology

[0002] Calculating the sample mean and sample variance of large datasets for statistical hypothesis testing can be used in a wide variety of fields and many different applications. Typically, applications can perform these calculations offline once all data has been collected. However, in online applications, such as those using radar to track / filter road users, detection results can be collected over time, and more and more information can be obtained over time.

[0003] However, calculating the mean and variance of a large dataset can be computationally expensive and require a lot of memory.

[0004] Therefore, there is a need to provide enhanced methods for determining the mean and variance of a dataset. Summary of the Invention

[0005] This disclosure provides a computer-implemented method, computer system, vehicle, and non-transitory computer-readable medium according to the independent claims. Embodiments are given in the dependent claims, the specification, and the drawings.

[0006] In one aspect, this disclosure relates to a computer-implemented method for determining the mean and variance of measurement data, the method comprising the following steps performed (in other words, carried out) by computer hardware components: acquiring measurement data for a current time step using a sensor; updating a plurality of intermediate data for a previous time step based on the measurement data for the current time step and based on a forgetting factor to obtain a plurality of intermediate data for the current time step; and determining the mean and variance for the current time step based on the plurality of intermediate data for the current time step. The previous time step may be a time step immediately preceding the current time step. The time step may be provided at predetermined fixed intervals, for example, based on the frequency of availability of new measurement data.

[0007] According to various embodiments, the method can be executed iteratively, such that after the current time step, the method is executed again for subsequent time steps (where the current time step becomes the previous time step).

[0008] According to various embodiments, the sensor may be a radar sensor or a lidar sensor.

[0009] According to various embodiments, the intermediate data includes three numbers. For example, intermediate data for the current time step may include three numbers for the current time step; intermediate data for the previous time step may include three numbers for the previous time step.

[0010] According to various embodiments, these three numbers are initialized to values ​​of 0 (e.g., for a first time step; e.g., when the method is executed for the first time (in other words: the first iteration); e.g., as intermediate data for a previous time step). In other words, each of the three numbers for the first time step is set to zero. The first time step can then be the previous time step in the first iteration of the method.

[0011] According to various embodiments, these three numbers include a first number, a second number, and a third number. According to various embodiments, updating the first number includes multiplying the first number by a forgetting factor and adding the number of measurements included in the measurement data. According to various embodiments, updating the second number includes multiplying the second number by a forgetting factor and adding the measurements included in the measurement data. According to various embodiments, updating the third number includes multiplying the third number by a forgetting factor and adding the square of the measurements included in the measurement data. The updates will be described in the equation in equation (1) below.

[0012] According to various embodiments, determining the average value includes determining the ratio between the second number and the first number. The determination of the average value will be described in the equation (2) below.

[0013] According to various embodiments, determining the variance includes determining the ratio between the difference between the third number and the square of the mean and the difference between the first number and 1. The determination of the variance will be described in the equation in equation (3) below.

[0014] According to various embodiments, the forgetting factor can be a number between 0 and 1. The lower the forgetting factor, the less influence older measurements have on the mean and variance.

[0015] According to various embodiments, the measurement data includes altitude information.

[0016] For movable objects, the assumption of a constant position / height of the detected object may not apply. According to various embodiments, incorporating a forgetting factor into the determination of the mean and variance can provide enhanced results. Assuming that the accuracy of the radar elevation angle is approximately the same regardless of the object distance, the height variance may be larger at greater distances, and according to various embodiments, this can also be accounted for by incorporating a forgetting factor.

[0017] According to various embodiments, the computer-implemented method further includes the following steps performed by computer hardware components: determining underdrivability and / or overdrivability based on mean and variance.

[0018] According to various embodiments, the drivability below and / or drivability above are determined based on a Student's T distribution, which is based on the mean and variance.

[0019] According to various embodiments, the measurement data includes distance rate information.

[0020] According to various embodiments, the computer-implemented method further includes the step of determining outliers in the association method during object tracking, performed by computer hardware components. Outliers may be measurements associated with an object but not actually belonging to that object (but, for example, belonging to another object, or, for example, noise measurements).

[0021] According to various embodiments, outliers are determined based on a threshold, wherein the threshold is determined based on the mean and variance.

[0022] According to various embodiments, the method can provide iterative and adaptive mean and variance filters, for example, with (or for) hypothesis testing.

[0023] In another aspect, this disclosure relates to a computer system comprising a plurality of computer hardware components configured to perform some or all of the steps of the computer-implemented methods described herein.

[0024] A computer system may include multiple computer hardware components (e.g., a processor (e.g., a processing unit or processing network), at least one memory (e.g., a memory cell or memory network), and at least one non-transitory data storage device). It will be understood that further computer hardware components may be provided and used to perform the steps of the computer-implemented method within the computer system. The non-transitory data storage device and / or memory cell may include a computer program for instructing the computer to perform several or all of the steps or aspects of the computer-implemented method described herein, for example, using the processing unit and at least one memory cell.

[0025] In another respect, this disclosure pertains to vehicles, including the computer systems and sensors described herein.

[0026] In another aspect, this disclosure relates to non-transitory computer-readable media, which include instructions for performing several or all of the steps or aspects of the computer-implemented methods described herein (in other words, instructions that cause a computer system to execute when performed by a computer system). Computer-readable media can be configured as: optical media, such as optical discs (CDs) or digital versatile optical discs (DVDs); magnetic media, such as hard disk drives (HDDs); solid-state drives (SSDs); read-only memory (ROMs), such as flash memory, etc. Furthermore, computer-readable media can be configured as data storage devices accessible via a data connection (e.g., an internet connection). Computer-readable media can be, for example, online data repositories or cloud storage devices.

[0027] This disclosure also relates to a computer program for instructing a computer to perform several or all of the steps or aspects of the computer-implemented method described herein. Attached Figure Description

[0028] This document describes exemplary embodiments and functions of the present disclosure in conjunction with the following accompanying drawings, which schematically illustrate:

[0029] Figure 1 This is a flowchart illustrating methods for determining the mean and variance of measurement data according to various embodiments; and

[0030] Figure 2 It is a computer system having multiple computer hardware components configured to perform steps of a computer-implemented method for determining the mean and variance of measurement data according to various embodiments. Detailed Implementation

[0031] Large amounts of data can be acquired for a variety of purposes. For example, sensors (e.g., radar sensors and / or lidar sensors and / or ranging sensors) can acquire large amounts of data for (at least partially) autonomous vehicles. Large datasets can include multiple measurements, each of which can be referred to as a sample. Using each of the samples individually may be inappropriate, for example, due to the computational workload and / or due to noise and / or due to the presence of outliers. Therefore, figures obtained over the entire sample set or a subset of samples (e.g., over a predetermined number of recent samples) can be used. These figures can, for example, represent one or more statistical properties of the sample. Examples of such figures could be the sample mean and the sample variance.

[0032] Calculating the sample mean and sample variance of large datasets for statistical hypothesis testing can be used in a wide variety of fields and many different applications. For example, when all data has been collected, the calculation can be performed offline. However, in online applications, such as using radar to track / filter road users, detection results can be collected over time, and more and more information can be obtained over time.

[0033] According to various embodiments, methods and apparatus are provided for determining the mean and variance during an iterative process, which enables their use in online applications even if the mean and variance change over time. Iterative determination of the mean and variance allows this determination to be applied to signals (in other words, samples) that may have mean and variance that vary over time.

[0034] According to various embodiments, the signal e at the current time n = N can be estimated. n The mean and variance (already acquired at different time steps, such as at time steps n = 1, 2, ..., N, where N is an integer) even if the signal mean and / or variance change over time.

[0035] At each moment, signal y n,i There may be multiple noise measurements, where i∈{1,2,…,k} n}, and k n It is the number of measurements obtained at time n. Signal y n,i It can also be called measurement data. For example, y n,i This can be referred to as the measurement data of the previous time step; and y n+1,i This can be referred to as the measurement data at the current time step. Therefore, it can be understood that for each time step, the measurement data may include multiple measurements.

[0036] According to various embodiments, aggregated measurement data can be used to determine the average value. and variance As shown below:

[0037] as well as

[0038]

[0039] In the equation above, the variable λ is an adjustment parameter corresponding to the forgetting factor, which gives newer measurements a greater weight than older measurements. λ can be a fixed number between zero and one: 0 ≤ λ ≤ 1. A value of λ of zero is equivalent to considering only the most recent measurement (i.e., measurement y). N,i A value of λ of one is equivalent to not forgetting older measurements, but rather considering all past measurements with the same weight.

[0040] One approach to aggregating enough information to calculate the desired mean and variance is to store all measurements over a period of time. However, this is impractical for online, real-time applications due to memory constraints.

[0041] According to various embodiments, the state vector s n (This can be called intermediate data) can be defined as follows:

[0042]

[0043] According to various embodiments, the state vector can be updated iteratively online, for example, each time a new measurement is obtained. The state update can be performed according to the following formula:

[0044]

[0045] As can be seen from the above formula, according to various embodiments, intermediate data can be updated using only new measurement information, and it may not be necessary to store all older measurements.

[0046] From the state vector, the average value at the current time n = N and variance It can be calculated as

[0047]

[0048] It is understandable that equation (3) only applies to s greater than 1. 1,N Perform the calculation.

[0049] The forgetting factor may lead to a decrease in confidence in older measurements / data points compared to newer detection points, and may prevent the mean and variance filters from converging to a fixed number.

[0050] According to various embodiments, the determination of the mean and variance (which may be referred to as a filter) can be used in conjunction with commonly used statistical testing methods (e.g., the Student's T-hypothesis test).

[0051] According to various embodiments, drivability below / drivability above can be detected for moving objects tracked using only radar detection, as described below.

[0052] Underdrivability / overdrivability detection can be understood as determining whether a vehicle is likely to travel under or over another object (e.g., because there is a bridge), or whether a vehicle is unlikely to travel under or over another object (e.g., because there is no bridge, but a regular road intersection or crossroads). Underdrivability / overdrivability detection classification can refer to determining one of several categories (e.g., indicating the degree of certainty that an object is underdrivable or overdrivable).

[0053] Radar detection can be obtained on moving objects overhead, such as vehicles crossing the main vehicle path on a bridge. If the height of these objects is not classified / estimated, they can cause problems, such as when using these detections in radar tracker applications (e.g., these types of objects might be mistaken for intersecting objects on the ground, and thus could lead to interventions, such as issuing a warning in the main vehicle, or even causing the main vehicle to brake).

[0054] For example, when this vehicle (in other words, the primary vehicle) approaches an intersection with a side road from the left or right, and another vehicle is also approaching the intersection from its side road, a collision may occur if neither vehicle brakes, and the drivers should be aware of this. In such a situation, there is no below-driving capability or above-driving capability.

[0055] In another example, when the primary vehicle approaches a bridge that crosses the road it is traveling on, and another vehicle is crossing the bridge, a collision is impossible, and therefore the driver should not be notified of any potential danger. In this case, there is a situation of drivability below / drivability above.

[0056] According to various embodiments, for each estimated (in other words: tracked) radar object, a mean and variance filter can be applied to the elevation (in other words: height) measurement associated with the radar detection. The mean and variance can be determined iteratively, and therefore, information can accumulate over time (although since the filter is iterative, it may not be necessary to store individual detections over time, but only the three single states of the mean and variance filter can be stored). In each time frame, a Student's T-hypothesis test can be applied to the estimated object height mean and variance to test whether the object height could be below a predetermined threshold, which can be an adjustment parameter.

[0057] Using a forgetting factor when determining the mean and variance makes the determination applicable to objects whose height can change over time. It can be understood that "height" can refer to the vertical distance relative to the host vehicle. For example, an object might be driving from a ramp of a highway to an overpass spanning the highway over which the host vehicle is traveling. This can be a significant difference compared to a stationary environment where the object remains at the same ground level.

[0058] According to various embodiments, estimating the variance of object height, in addition to the average height, can yield better results than obtaining variance information directly from radar specifications. For example, the variance can vary depending on factors such as the type of object (e.g., truck, car, or pedestrian), the object's location within the radar field of view, etc. Therefore, using information from radar specifications may not always represent accurate data.

[0059] Furthermore, in cases where the method needs to be equally effective for different types of radars and configurations with varying specifications, by estimating the variance, it may be possible to use a single filter with the same adjustment capabilities to adapt to all different configurations, without requiring different adjustments for different configurations and with varying performance when measuring target elevation / altitude. This can make tracker code maintenance and integration much easier. The method, according to various embodiments, can be used, for example, for fifth-generation radars with elevation accuracy of ±2 degrees based on sensor specifications and seventh-generation HD radars with elevation accuracy of ±0.1 degrees based on radar specifications (and variations with accuracy between these two values). Therefore, there could be a difference of approximately 20 times between the worst and best results.

[0060] According to various embodiments, the following drivable classification enumeration value can be assigned to each of the tracked objects classified as drivable. For example, possible enumeration values ​​to choose from include:

[0061] • CAN_NOT_PASS_UNDER (e.g., indicating that the primary vehicle cannot pass under the tracked object);

[0062] • IS_LIKELY_TO_PASS_UNDER (e.g., indicating that the primary vehicle is likely to pass under the tracked object);

[0063] • CAN_PASS_UNDER (e.g., indicating that the master vehicle may pass under the tracked object);

[0064] • NOT_TO_CONSIDER / UNKNOWN (e.g., indicating that it is not possible to determine whether the master vehicle can pass under the tracked object, or indicating that no further processing of the tracked object is required, such as making a driving decision).

[0065] According to various embodiments, the following drivable classification method may include the following steps:

[0066] 1. State vector s n It can be initialized with zero: s n =[s 1,n s 2,n s 3,n ] T =[0 0 0] T , where n = 0.

[0067] For each time and each movable object, the following steps can be performed:

[0068] 2. (Radar) detection can be associated with an object. This step can be the same association performed when running a Kalman filter to track an object's 2D (two-dimensional) position (without altitude), velocity, and / or heading, and various methods can be used to associate radar detection with an object.

[0069] 3. For each associated detection, the detection height on the ground (assuming a flat surface) can be determined, for example, by utilizing the measurement range and elevation angle, as well as the sensor's installation height. If the determined height is negative, the absolute value of the height can be used; this may be due to ground bounce, which could cause the calculated detection height to have an incorrect sign.

[0070] 4. The state vector can be updated using new height information from all associated detections according to the above equation (1) (wherein, the sample (or measurement data) on which the updated state vector is based can be new height information from all associated detections). Assuming the detection measurements are normally distributed, the test variable is based on s 1,N The student is distributed using a T-distribution with -1 degrees of freedom.

[0071] 5. Average detection height and variance It can be determined according to equations (2) and (3). Only when s 1,N This step and the remaining steps can only be performed if the total number of tests (corresponding to the sum of the tests) is greater than 1 (to avoid division by zero, and because variance is not defined for a single measurement). If s 1,N If the value is less than or equal to 1, then the drivable enumeration value NOT_TO_CONSIDER / UNKNOWN can be assigned to the object to indicate that there are not enough data points to complete the classification.

[0072] 6. Test variables can be determined. Where h can be an adjustment variable defining a height threshold used for classifying drivability below. Test variable t N It can be based on a degree of freedom of s1,N The students are given a T-distribution with a mean of -1, assuming that the height measurements follow a normal distribution with a mean of h.

[0073] 7. The student T p-value can be determined. This can be given by the cumulative student T-distribution function. Approximate methods for determining this probability value can be used, and several methods are known and commonly used for this determination. The p-value can correspond to "the probability of obtaining an outcome equal to or more extreme than the actual observed outcome, assuming no effect or no difference (null hypothesis)". As used in this paper, the null hypothesis can be that the object is "above ground", where "above ground" can be defined as above h meters, where h can be the adjustment parameter referred to above. Therefore, what is being tested is the likelihood of obtaining an estimated mean and variance given the null hypothesis. The p-value can be this probability, and if the p-value is small, the null hypothesis is unlikely to be correct (i.e., the object is likely on the ground), but if the p-value is large, the null hypothesis is likely correct and the object is above ground.

[0074] 8. The obtained p-value can be compared with a threshold, such as a lower threshold (lower_threshold) and an upper threshold (high_threshold). The threshold can be an adjustment parameter.

[0075] If the p-value is below the lower threshold, a classification value can be assigned that indicates the primary vehicle cannot pass underneath the object:

[0076] p_value<lower_threshold=> CAN_NOT_PASS_UNDER

[0077] If the p-value is between the lower and upper thresholds, a classification value can be assigned that indicates the primary vehicle is most likely to pass under the object:

[0078] lower_threshold≤p_value≤high_threshold=>IS_LIKELY_TO_PASS_UNDER

[0079] If the p-value is higher than the upper threshold, a classification value can be assigned that indicates the primary vehicle can pass underneath the object:

[0080] p_value>high_threshold=>CAN_PASS_UNDER

[0081] 9. In addition to enumeration values, or as a substitute for enumeration values, probability values ​​can be provided as output.

[0082] It is understandable that although steps 1 through 9 are described as examples for the lower drivability category, similar steps can be performed for the upper drivability category.

[0083] According to various embodiments, the rejection of distance rate outlier measurements can be performed before performing Kalman filter measurement updates, as described below.

[0084] When using radar trackers to track multiple extended objects (e.g., an unknown number of extended objects), the data association problem (in other words, determining which of the obtained radar detection results belongs to which tracked object) can be a critical issue that needs to be addressed. An "extended object" can be an object with spatial extension (as opposed to an object that is merely a point).

[0085] The data association problem can be addressed by comparing radar detection attributes (e.g., detection location and detection range rate) with attributes of the tracked object (e.g., position and velocity). If the detected attributes are "sufficiently similar" to the object's attributes, the detection can be associated with the object (which could mean that the detection can be used to update and improve estimates of the object's attributes). This is likely how the data association problem is typically solved in the context of Kalman filtering.

[0086] The degree of similarity required for "sufficient similarity" can be termed "detection within the object's association gate." The association gate size must be large enough to account for various sources of error, such as the accuracy of the measured detection attribute and the accuracy of the estimated object attribute. For example, the state of the tracked object may have significant uncertainty, and since the performance of the detection association with the object depends on the accuracy of the tracking, a large gate may be desirable. Therefore, it is possible for detections not actually originating from the object itself to be associated with the object (thus associating with bad detections / outliers). Examples of this could be, for instance, when a slowly moving object approaches a stationary object (e.g., a guardrail), when two cars traveling at similar speeds approach each other in adjacent lanes, or when radar picks up a detection from the spinning wheel of a moving object (where the measured distance rate may be slightly inaccurate because the wheel's speed is slightly different from the object's). Due to the nature of the fundamental assumptions of the Kalman filter (which may include the assumption that the signal follows a normal distribution and measurements with short tails of the normal distribution), the Kalman filter can be sensitive to outliers. Therefore, using outlier measurements in Kalman filter updates can easily lead to corrupt / inaccurate (state) estimates. Therefore, it may be desirable to find and remove outliers before updating the filter. This can typically be done in a post-association step. In this step (in common methods for finding and rejecting outliers by distance rate), the attributes (e.g., distance rates) of all associated detections can be compared to each other (note that in the association step, the attributes of the detections are compared to the attributes of the object, while in the find and reject step, the attributes of the detections are compared to the attributes of the detections). For example, the measured distance rates of all associated detections can be compared to find detections whose distance rates differ significantly from all other detections. However, in some scenarios, this may not be robust enough to perform this operation or comparison. This might be the case when the number of detections is low, such as if the object is small (e.g., a pedestrian) or if the object is far away (understandably, the closer the main vehicle is to the object, the more detections are expected from the object). For example, if there is only a single associated detection within a time frame, there may be no other detections to compare with its distance rate. For example, if there are two associated detections with different distance rates, it may be difficult to distinguish which one is a normal value and which one is not an outlier.

[0087] Common approaches might assume that the predicted range rate error (e.g., the difference between the estimated object state assumption / predicted range rate and the actual range rate measured by the radar) should be close to zero. However, this may only hold true if tracking is successful and the object is in a steady state (and not performing any strenuous maneuvers). Therefore, this assumption may not always be valid.

[0088] According to various embodiments, methods for enhancing association results are provided.

[0089] According to various embodiments, instead of performing an instantaneous comparison between all currently associated detections in each individual time frame, a mean and variance filter can be applied to the predicted range rate error of all associated detections. The predicted range rate error can correspond to the difference between the expected range rate for a given estimated object attribute and the actual radar measurement of the detection. Since the filter according to various embodiments is iterative, it is not necessary to store individual detections over time; storing only the three single states of the mean and variance filter is sufficient. In each time frame, given the estimated filtered mean and variance of the predicted range rate error, a Student's T-hypothesis test can be performed on the range rates of all newly associated detections to determine whether the newly associated detection is likely a normal point or an outlier that should not be used in the Kalman filter update. Because the filter according to various embodiments uses not only current detections but also information from historical detections, outlier detection can also be provided in a more robust manner, even when the number of detections in each time frame is small.

[0090] According to various embodiments, outlier rejection based on detection distance rate can be performed after association with the tracked object.

[0091] According to various embodiments, starting from the set of radar detections already associated with the tracked radar object, range rate outliers can be found. In other words, it is possible to find associated detections whose range rates do not quite correspond to the range rates of other associated detections.

[0092] According to various embodiments, the method may include the following steps:

[0093] 1. State vector s n It can be initialized to zero: s n =[s 1,n s 2,n s 3,n ] T =[0 0 0] T where n = 0

[0094] For each time and each movable object, the following steps can be performed:

[0095] 2. (Radar) detection can be associated with an object. This step can be the same association performed when running a Kalman filter to track the 2D (two-dimensional) position, velocity, and / or heading of an object, and various methods can be used to associate radar detection with an object.

[0096] 3. For each associated detection, the predicted distance rate error (i.e., the difference between the expected distance rate and the actual measured distance rate for a given tracked object state) can be determined.

[0097] 4. The state vector can be updated using the new predicted distance rate error information from all associated detections according to the above equation (1) (wherein, the sample (or measurement data) on which the updated state vector is based can be the new predicted distance rate error information).

[0098] 5. The test threshold variable can be determined. The lower limit threshold t can be determined. low and upper limit threshold t high For example, it can be determined and Here, k can be an adjustment variable.

[0099] 7. For each association detection, the predicted distance rate error can be compared to two thresholds. If the predicted distance rate error falls between the lower and upper thresholds, the detection is considered normal; otherwise, it is identified as an outlier.

[0100] t low <predicted_range_rate_error<t high => If the value is normal, it is considered an outlier.

[0101] Figure 1 A flowchart 100 is shown, illustrating methods for determining the mean and variance of measurement data according to various embodiments. In 102, measurement data for the current time step can be acquired using a sensor. In 104, multiple intermediate data for the previous time step can be updated based on the measurement data for the current time step and based on a forgetting factor to obtain multiple intermediate data for the current time step. In 106, the mean and variance for the current time step can be determined based on the multiple intermediate data for the current time step.

[0102] According to various embodiments, intermediate data (for each time step) may include or may include three numbers.

[0103] According to various implementations, these three numbers can be initialized with the value 0.

[0104] According to various embodiments, these three numbers (for each time step) may include a first number (for each time step), a second number (for each time step), and a third number (for each time step). Updating the first number may include, or may be, multiplying the first number by a forgetting factor and adding the number of measurements included in the measurement data. Updating the second number may include, or may be, multiplying the second number by a forgetting factor and adding the measurements included in the measurement data. Updating the third number may include, or may be, multiplying the third number by a forgetting factor and adding the square of the measurements included in the measurement data.

[0105] According to various embodiments, determining the average may include or may include determining the ratio between the second number and the first number. According to various embodiments, determining the variance may include or may include determining the ratio between the difference between the third number and the square of the average and the difference between the first number and 1.

[0106] According to various embodiments, the forgetting factor can be a number between 0 and 1.

[0107] According to various embodiments, the measurement data may include or may be height information.

[0108] According to various embodiments, the method may also include determining lower drivability and / or upper drivability based on mean and variance.

[0109] According to various embodiments, the lower drivability and / or upper drivability can be determined based on a Student T-distribution, which is based on the mean and variance.

[0110] According to various embodiments, the measurement data may include or may be distance rate information.

[0111] According to various embodiments, the method may further include: determining outliers in the association method during object tracking.

[0112] According to various embodiments, outliers can be determined based on a threshold, wherein the threshold is determined based on the mean and variance.

[0113] Each of the steps 102, 104, 106 and the further steps described above can be performed by computer hardware components.

[0114] Figure 2 A computer system 200 with multiple computer hardware components is illustrated, the computer hardware components being configured to perform steps of a computer-implemented method for determining the mean and variance of measurement data according to various embodiments. The computer system 200 may include a processor 202, a memory 204, and a non-transitory data storage device 206. A sensor 208 may be provided as part of the computer system 200 (e.g., Figure 2 (as shown), or it can be set outside the computer system 200.

[0115] Processor 202 can execute instructions provided in memory 204. Non-transitory data storage device 206 can store computer programs, including instructions that can be transferred to memory 204 and then executed by processor 202. Sensor 208 can be used to acquire measurement data as described above.

[0116] Processor 202, memory 204, and non-transitory data storage device 206 may be coupled to each other, for example, via electrical connection 210 (e.g., cable or computer bus) or via any other suitable electrical connection to exchange electrical signals. Sensor 208 may be coupled to computer system 200, for example, via an external interface, or may be provided as part of computer system (in other words: internal to computer system, for example, coupled via electrical connection 210).

[0117] The terms “coupled” or “connected” are intended to include direct “coupling” (e.g., via a physical link) or direct “connection” as well as indirect “coupling” or indirect “connection” (e.g., via a logical link).

[0118] It is understood that the content described in one of the above methods can be similarly applied to computer system 200.

[0119] List of reference numerals

[0120] Flowchart 100 illustrates methods for determining the mean and variance of measurement data according to various embodiments.

[0121] 102 Steps for acquiring measurement data for the current time step using sensors

[0122] 104 Step to obtain multiple intermediate data for the current time step based on the measurement data for the current time step and based on the forgetting factor. 106 Step to determine the mean and variance for the current time step based on the multiple intermediate data for the current time step.

[0123] 200 Computer systems according to various embodiments

[0124] 202 processor

[0125] 204 memory

[0126] 206 Non-transitory data storage device

[0127] 208 sensor

[0128] 210 connection

Claims

1. A computer-implemented method for determining the mean and variance of measurement data. The method includes the following steps performed by computer hardware components: - Use sensors to acquire measurement data for the current time step; -Based on the measurement data for the current time and based on a forgetting factor, multiple intermediate data for the previous time step are updated to obtain multiple intermediate data for the current time step; and - Determine the mean and variance for the current time step based on the plurality of intermediate data for the current time step.

2. The computer-implemented method according to claim 1, in, The intermediate data includes three numbers.

3. The computer-implemented method according to at least one of claims 1 to 2, in, The three numbers are initialized with the value 0.

4. The computer-implemented method according to claim 2, in, The three numbers include the first number, the second number, and the third number; Updating the first number includes multiplying the first number by the forgetting factor and adding the number of measurements included in the measurement data; Updating the second number includes multiplying the second number by the forgetting factor and adding the measurement included in the measurement data; and Updating the third number involves multiplying the third number by the forgetting factor and adding the square of the measurement included in the measurement data.

5. The computer-implemented method according to claim 4, in, Determining the average includes determining the ratio between the second number and the first number; and Determining the variance includes determining the ratio between the square of the difference between the third number and the average value and the difference between the first number and 1.

6. The computer-implemented method according to at least one of claims 1 to 2, in, The forgetting factor is a number between 0 and 1.

7. The computer-implemented method according to at least one of claims 1 to 2, in, The measurement data includes altitude information.

8. The computer-implemented method of claim 7, further comprising the following steps performed by the computer hardware components: The lower drivability and / or upper drivability are determined based on the mean and the variance.

9. The computer-implemented method according to claim 8, in, The lower drivability and / or the upper drivability are determined based on a student T-distribution, which is based on the mean and the variance.

10. The computer-implemented method according to at least one of claims 1 to 2, in, The measurement data includes distance rate information.

11. The computer-implemented method according to claim 10, It also includes the following steps performed by the computer hardware components: Identify outliers in associated methods during object tracing.

12. The computer-implemented method according to claim 11, in, The outlier values ​​are determined based on a threshold, wherein the threshold is determined based on the mean and the variance.

13. A computer system configured to perform a computer-implemented method according to at least one of claims 1 to 12.

14. A vehicle comprising the computer system of claim 13 and sensors.

15. A non-transitory computer-readable medium comprising instructions for performing a computer-implemented method according to at least one of claims 1 to 12.

Images