Method and system for estimating loaded and empty shipping containers

Abstract

Embodiments of the present disclosure relate to methods and systems for estimating whether a shipping container is loaded or unloaded. A method at a computing device includes: obtaining sensor data for a vehicle, the sensor data providing vibration frequency and amplitude; calculating energy for each of a low-frequency passband and a high-frequency passband of a bandpass filter pair; determining an energy ratio based on the energy for the low-frequency passband and the energy for the high-frequency passband; applying a weighting constant to each of the following terms to calculate a decision variable: the energy of the low-frequency passband, the energy of the high-frequency passband, and the energy ratio; and determining that the vehicle is unloaded if the decision variable is below a threshold, and determining that the vehicle is loaded if the decision variable is above the threshold.

Description

Technical Field

[0001] The present disclosure relates to freight transportation, and in particular to a system for detecting the presence of a load on a vehicle or container on a trailer chassis. Background Art

[0002] During the transport of cargo, determining whether a trailer is loaded is an important aspect of trailer asset management. Where a tractor trailer is used to transport shipping containers, it is important to know whether the trailer (referred to herein as a chassis) is loaded.

[0003] Information about vehicle loading can be beneficial to transportation companies. Specifically, transportation companies that manage fleets of vehicles need to know which vehicles are loaded and which vehicles are empty.

[0004] Typically, confirming that a vehicle is loaded is done through manual inspection, which is a tedious process that can be slow and inaccurate. Summary of the Invention

[0005] The present disclosure provides a method at a computing device, the method comprising: obtaining sensor data for a vehicle, the sensor data providing a vibration frequency and amplitude; calculating energy for each of a low-frequency passband and a high-frequency passband of a bandpass filter pair; determining an energy ratio based on the energy for the low-frequency passband and the energy for the high-frequency passband; applying a weighting constant to each of the following terms to calculate a decision variable: the energy of the low-frequency passband, the energy of the high-frequency passband, and the energy ratio; and determining that the vehicle is unloaded if the decision variable is below a threshold, and determining that the vehicle is loaded if the decision variable is above the threshold.

[0006] The present disclosure also provides a computing device comprising: a processor; and a communication subsystem, wherein the computing device is configured to: obtain sensor data for a vehicle, the sensor data providing vibration frequency and amplitude; calculate energy for each of a low-frequency passband and a high-frequency passband of a bandpass filter pair; determine an energy ratio based on the energy for the low-frequency passband and the energy for the high-frequency passband; apply a weighting constant to each of the following items to calculate a decision variable: energy of the low-frequency passband, energy of the high-frequency passband, and energy ratio; and determine that the vehicle is unloaded if the decision variable is below a threshold, and determine that the vehicle is loaded if the decision variable is above the threshold.

[0007] The present disclosure also provides a computer-readable medium for storing instruction codes, which, when executed by a processor on a computing device, causes the computing device to: obtain sensor data for a vehicle, the sensor data providing a vibration frequency and amplitude; calculate the energy for each of a low-frequency passband and a high-frequency passband of a bandpass filter pair; determine an energy ratio based on the energy for the low-frequency passband and the energy for the high-frequency passband; apply a weighting constant to each of the following items to calculate a decision variable: the energy of the low-frequency passband, the energy of the high-frequency passband, and the energy ratio; and determine that the vehicle is unloaded if the decision variable is below a threshold, and determine that the vehicle is loaded if the decision variable is above the threshold.

[0008] According to the embodiments described below, a cargo load detection system utilizing vertical accelerometers and / or strain gauges is described.

[0009] In the embodiment described below, load detection is performed on a trailer chassis connected to a tractor. However, in other cases, measurements can be performed on other shipping containers, including but not limited to rail cars, trucks, cars, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The present disclosure may be better understood with reference to the accompanying drawings, in which:

[0011] Figure 1 is a side view of a trailer chassis suitable for receiving a shipping container, showing example placement of a sensor device;

[0012] Figure 2 is a block diagram of an example sensor device that can be used with embodiments of the present disclosure;

[0013] Figure 3 It shows the Figure 2 A block diagram of an example architecture of a sensor device;

[0014] Figure 4 is a block diagram showing the container and chassis acting on the springs and dampers;

[0015] Figure 5 is a graph showing the variation of vibration amplitude according to frequency for a loaded chassis and an unloaded chassis;

[0016] Figure 6 is a graph showing the distribution of eRatio with respect to the original data using two bandpass filters;

[0017] Figure 7 It shows that according to Figure 6 Plot of the dependent variable shift;

[0018] Figure 8is a diagram on a two-dimensional plane showing the relationship between eRatio and the energy of the lower bandpass filter;

[0019] Figure 9 is a process diagram illustrating a process of training and then acquiring a loaded / unloaded decision for a vehicle;

[0020] Figure 10 The figure shows the server calculating eRatio, E L and E H , and use them to calculate the data flow graph for the training process;

[0021] Figure 11 is a diagram showing the sensor device calculating E L and E H , and pass them to the server to calculate the data flow graph for the training process of eRatio and weight constants;

[0022] Figure 12 Figure 2 shows the process in which the sensor device calculates eRatio, E L and E H , and use them to calculate the process diagram for the training process;

[0023] Figure 13 The figure shows the server calculating eRatio, E L and E H , and a data flow diagram of a process for making a loaded decision / no-load decision using these constants and weighted constants to determine a decision variable;

[0024] Figure 14 Figure 2 shows the process in which the sensor device calculates eRatio, E L and E H , and a data flow diagram of a process for making a loaded decision / no-load decision using these constants and weighted constants to determine a decision variable;

[0025] Figure 15A is a diagram showing a histogram of the distribution of decision variables based on eRatio;

[0026] Figure 15B is a diagram showing a histogram of the decision variable distribution based on the minimum mean square error (MSE) solution;

[0027] Figure 16A is a diagram showing a histogram of the distribution of decision variables based on eRatio;

[0028] Figure 16B is a diagram showing a histogram of the distribution of decision variables based on a support vector machine (SVM) solution;

[0029] Figure 17 is a diagram on a two-dimensional plane showing the relationship between eRatio and the energy of the lower bandpass filter, and having an MSE decision line and an SVM decision line;

[0030] Figure 18A is a diagram showing a histogram of the distribution of decision variables based on eRatio;

[0031] Figure 18B is a diagram showing a histogram of decision variable distribution based on SVM using a plurality of bandpass filter pairs; and

[0032] Figure 19 is a block diagram of an example computing device or server that can be used with embodiments of the present disclosure. DETAILED DESCRIPTION

[0033] To perform load detection, a sensor device may be attached to a container, trailer, or other similar asset. Such a sensor device may, for example, be mounted inside the chassis of a flatbed trailer configured to receive a shipping container. Figure 1 .

[0034] exist Figure 1 In the embodiment of FIG, an example truck trailer 110 is shown. In one embodiment, the computing device can be mounted within the chassis of the trailer. For example, in one embodiment, the computing device can be mounted above the rear wheels of the truck trailer 110. For example, this Figure 1 The embodiment shown has a sensor device 112.

[0035] However, in other cases, it may be beneficial to have different locations for the sensor device.Further, in some embodiments, it may be useful to have multiple such sensor devices within the trailer 110.

[0036] In some embodiments, sensor devices within trailer 110 may be used individually, or may be combined into sets of two or more sensor devices and / or external sensors for use in load determination calculations.

[0037] exist Figure 1 In the embodiment of the present invention, the trailer 110 is adapted to secure and carry thereon the shipping container 120. Information about the presence or absence of the shipping container 120 may be useful to a transportation company.

[0038] Device

[0039] about Figure 2 A sensor arrangement is shown for use with a vehicle, chassis, trailer, container, or other transport asset. However, Figure 2 The sensor arrangement is merely an example, and other sensing devices may equally be used according to embodiments of the present disclosure.

[0040] Now, reference Figure 2 , which shows an example sensor device 210. Sensor device 210 can be any computing device or network node. Such a sensor device or network node can include any type of electronic device, including but not limited to mobile devices such as smartphones or cellular phones. Examples can also include fixed or mobile devices such as Internet of Things (IoT) devices, endpoints, home automation devices, medical devices in hospital or home environments, inventory tracking devices, environmental monitoring devices, energy management devices, infrastructure management devices, vehicles or vehicle devices, fixed electronic devices, etc.

[0041] Sensor device 210 includes a processor 220 and at least one communication subsystem 230, wherein processor 220 and communication subsystem 230 cooperate to perform the methods of the embodiments described herein. In some embodiments, communication subsystem 230 may include multiple subsystems, for example, for different radio technologies.

[0042] The communication subsystem 230 allows the sensor device 210 to communicate with other devices or network elements. The communication subsystem 230 may use one or more of a variety of communication types, including but not limited to cellular, satellite, Bluetooth, TM ,Bluetooth TM Low energy, Wi-Fi, wireless local area network (WLAN), super-gigahertz radio, near field communication (NFC), IEEE 802.15, wired connections such as Ethernet or fiber optic, and other options.

[0043] As such, the communication subsystem 230 for wireless communication typically has one or more receivers and transmitters and associated components such as one or more antenna elements, a local oscillator (LO), and may include a processing module such as a digital signal processor (DSP) or a system on a chip (SOC). It will be apparent to those skilled in the art of communications that the specific design of the communication subsystem 230 will depend on the communication network or communication technology with which the sensor device is intended to operate.

[0044] Processor 220 generally controls the overall operation of sensor device 210 and is configured to execute programmable logic, which may be stored along with data using memory 240. Memory 240 may be any tangible, non-transitory computer-readable storage medium, including DRAM, flash memory, optical media (e.g., CDs, DVDs, etc.), magnetic media (e.g., tape), a flash drive, a hard drive, or other memory known in the art.

[0045] Alternatively or in addition to memory 240 , sensor device 210 may access data or programmable logic from an external storage medium (not shown), such as through communication subsystem 230 .

[0046] exist Figure 2 In some embodiments, sensor device 210 may utilize multiple sensors, which may be part of sensor device 210 in some embodiments or in other embodiments may be in communication with sensor device 210. For internal sensors, processor 220 may receive input from sensor subsystem 250.

[0047] exist Figure 2 Examples of sensors in the embodiment of include positioning sensor 251, vibration sensor 252, temperature sensor 253, one or more image sensors / cameras 254, accelerometer 255, light sensor 256, gyroscope sensor 257, door sensor 258, strain gauge 259, and other sensors 260. The other sensors can be any sensor that can read or obtain data useful to sensor device 210. However, in Figure 2 The sensors shown in the embodiment of FIG. 1 are examples only, and in other embodiments, the sensors may be used in Figure 2 The different sensors or subsets of sensors shown in . For example, in some cases, the only sensor may be an accelerometer or a strain gauge.

[0048] Furthermore, the accelerometer 255 typically provides acceleration sensors in three dimensions. Thus, the accelerometer 255 typically includes three separate accelerometers. The readings from each of the three separate accelerometers can be isolated.

[0049] In one embodiment, communication between the various elements of sensor device 210 may occur via internal bus 265. However, other forms of communication are also possible.

[0050] exist Figure 2In some embodiments, the strain gauge 259 may be external to the sensor device (rather than an internal strain gauge sensor) and may be controlled by the sensor device 210. The strain gauge sensor 270 may, for example, be mounted together with the sensor device 210 or may form part of the sensor device 210. A single or multiple strain gauge sensors may be mounted on the longitudinal frame of the chassis to detect displacement when the container is loaded onto the chassis. In some embodiments, the strain gauge sensor may consist of a microprocessor, a strain gauge, a battery, and a short-range technology radio transmitter (Bluetooth, IEEE 802.15.4, or Wi-Fi). The sensor 270 may be configured to measure vertical displacement or both vertical and horizontal displacement relative to the chassis.

[0051] The sensor device 210 can communicate with the strain gauge sensor 270 or the strain gauge 259 to query the strain gauge reading under a trigger condition. Examples of trigger conditions can be: on a regular wake-up schedule negotiated between the sensor(s) and the sensor device unit; when the trailer is in motion; and / or when the trailer is stopped, etc.

[0052] As the container is loaded, the strain gauges are typically displaced and can be calibrated to estimate the relative or absolute weight of the container.

[0053] The sensor device 210 can be attached to any fixed or portable platform. For example, in one embodiment, the sensor device 210 can be attached to a shipping container or a truck trailer. Figure 1 As shown, in other embodiments, the sensor device 210 can be attached to the chassis of a trailer. In other cases, the sensor device 210 can be attached to any transportation asset requiring load detection, including motor vehicles (e.g., automobiles, cars, trucks, buses, etc.), rail vehicles (e.g., trains and trams, etc.), and other types of vehicles, including any combination of any of the above, whether currently existing or emerging in the future, etc.

[0054] In other cases, the sensor device 210 may be a container that can be carried on or within a vehicle (e.g., from a Figure 1 According to the present disclosure, the term container may include any type of cargo or item transportation, such as a vehicle, an intermodal container, a shipping container, a lock box, and other similar containers.

[0055] Such sensor device 210 can be a power-constrained device. For example, in some embodiments, sensor device 210 can be a battery-operated device that can be attached to a shipping container or trailer. Other limited power sources can include any limited power source, such as a small generator or dynamo, fuel cell, solar energy, energy harvesting, and other options.

[0056] In other embodiments, the sensor device 210 can utilize external power, such as from a battery or power system of a towing vehicle pulling the cart, such as via a wiring harness connected to a 7-pin plug, a land power source such as on a plug-in recreational vehicle, or a building power source, among other options. Thus, the sensor device 210 can also be connected to a power cord that receives its power from a power source.

[0057] The external power may also allow the battery to be charged, allowing the sensor device 210 to subsequently operate in a power limited mode again. Charging methods may also include other power sources such as, but not limited to, solar, electromagnetic, acoustic, or vibration charging.

[0058] From Figure 2 The sensor device can be used in a variety of environments. Figure 3 , shows an example environment in which the sensor device can be used.

[0059] refer to Figure 3 , three sensor devices are provided, namely, sensor device 310 , sensor device 312 , and sensor device 314 .

[0060] exist Figure 3 In the example of FIG, sensor device 310 can communicate through cellular base station 320 or through access point 322. Access point 322 can be any wireless communication access point.

[0061] Further, in some embodiments, sensor device 310 may communicate via a wired access point such as Ethernet or fiber optic, among other options.

[0062] Communication may then proceed through a wide area network, such as the Internet 330 , and to server 340 or 342 .

[0063] Likewise, sensor device 312 and sensor device 314 may communicate with server 340 or server 342 through one or both of base station 320 or access point 322, among other options for such communication.

[0064] In other embodiments, any of sensor devices 310, 312, or 314 may communicate via satellite communication technology. This may be useful, for example, if the sensor device is traveling to an area outside of cellular coverage or access point coverage.

[0065] In other embodiments, sensor device 312 may be out of range of access point 322 and may communicate with sensor device 310 to allow sensor device 310 to act as a communication relay.

[0066] Communication between sensor device 310 and server 340 can be one-way or two-way. Thus, in one embodiment, sensor device 310 may provide information to server 340, but server 340 may not respond. In other cases, server 340 may issue commands to sensor device 310, but the data may be stored internally on sensor device 310 until the sensor device arrives at a specific location (possibly within a specific time window). In other cases, there may be two-way communication between sensor device 310 and server 340.

[0067] Server, central server, processing service, endpoint, uniform resource identifier (URI), uniform resource locator (URL), backend and / or processing system may be used interchangeably in the description herein. The server functionality generally represents data processing / reporting that is not closely related to the location of the sensor devices 310, 312, 314, etc. For example, the server can be located essentially anywhere as long as it has network access to communicate with the sensor devices 310, 312, 314, etc.

[0068] Server 340 may be, for example, a centralized monitoring station for fleet management. In this case, server 340 may receive information from sensor devices associated with various trailers or cargo containers, providing information such as the location of such cargo containers, the temperature within such cargo containers, any abnormal events including sudden deceleration, temperature alerts for excessively high or low temperatures, the cargo loading status of the trailers, the quality of the trailers, and other data. Server 340 may compile this information and store it for future reference.

[0069] Other functional examples for server 340 are possible.

[0070] exist Figure 3 In the embodiment of the present invention, servers 340 and 342 can also access third-party information or information from other servers within the network. For example, data service provider 350 can provide information to server 340. Similarly, data repository or database 360 ​​can also provide information to server 340.

[0071] For example, data service provider 350 may be a subscription-based service used by server 340 to obtain current road and weather conditions, or in some cases, data service provider 350 may be an inventory control system.

[0072] The data repository or database 360 ​​may, for example, provide information such as image data associated with a particular location, aerial maps, detailed street maps, or other such information.

[0073] The types of information provided by data service provider 350 or data repository or database 360 ​​are not limited to the examples described above, and the provided information may be any data useful to server 340 .

[0074] In some embodiments, information from data service provider 350 or data repository from database 360 ​​may be provided to one or more of sensor devices 310 , 312 , or 314 for processing at those sensor devices.

[0075] Such as the above Figure 2 and Figure 3 The sensor arrangement described in can be used to detect trailer loading of a container or a trailer.

[0076] Calculating trailer loading

[0077] According to some embodiments of the present disclosure, the rear section of a transport vehicle or trailer can be modeled as a simple spring with mass "m" representing the mass of the container and the mass of the chassis, and a spring constant "k", where k is the stiffness of the structure and the suspension. Further, the damper of the suspension, which may depend on the chassis, can be represented using a damping constant "c". For example, now, referring to Figure 4 .

[0078] exist Figure 4 In the embodiment, the mass of the container 410 is "m c ”. The mass of the chassis 412 is “m1”. In this disclosure, the mass “m” is the mass m1 and m c combination.

[0079] Spring 420 is shown having a spring constant "k" which represents the stiffness of the structure and suspension.

[0080] Further, a damping constant "c" is shown using a representation 422 of the suspension.

[0081] The displacement "h" shows the amount of chassis movement from its rest position towards the ground, for example when a bump is encountered by the chassis while the vehicle is operating.

[0082] As provided below in the description of Equation 1, Equation 1 can express a chassis moving in one dimension using a single spring-damper:

[0083]

[0084] In Equation 1, m is the mass of the system, h is the displacement of the chassis from the rest position, t is time, k is the spring constant, and c is the damping constant.

[0085] Solving Equation 1 for the vibration frequency yields Equation 2:

[0086]

[0087] The vibration frequency is inversely proportional to the mass applied to the trailer based on Equation 2. When the trailer is subjected to the same moving conditions, the vibration frequency of the trailer will shift downward when the chassis is loaded with shipping containers.

[0088] If a truck, trailer or other similar asset is loaded within safe operating limits, the springs on the axle should not experience any significant deformations that could affect the vibration frequency.

[0089] The amplitude of the oscillations depends on any impacts between the wheels in the ground, but such impacts do not affect the vibration frequency.

[0090] For example, now refer to Figure 5 , which shows example graphs of vibration frequencies seen when the chassis is loaded with shipping containers or unloaded with shipping containers, respectively. Figure 5 It can be seen that when the chassis is loaded, the peak frequency of the vibration shifts.

[0091] based on Figure 5 By measuring the vibration frequency in the vertical direction as observed by accelerometers and / or as observed by strain gauges, it can be detected whether the chassis or trailer is loaded.

[0092] One approach to using different frequencies for loaded and unloaded chassis is to calculate an energy ratio (eRatio) for comparison between low frequencies and a combination of low and high frequencies. Specifically, a pair of bandpass filters can be used, one in the lower frequency band and one in the higher frequency band. These bandpass filters can be positioned around typical loaded and unloaded frequencies to capture the average energy around these frequencies.

[0093] After n tests, n log files of raw data can be collected from the sensor. For example, one log file can contain 1024 samples. However, this number of samples is provided for illustration only and is not a limitation.

[0094] The raw data from these files are then processed through a detector, where each of the low-band bandpass filters (LBPF) and high-band bandpass filters (HBPF) generates n values ​​of output average energy corresponding to n log files. For example, these output values ​​can be expressed as E L(i) and E H (i) represents, where the subscripts "L" and "H" indicate LBPF and HBPF respectively, and i = 1, 2, ..., n. In the detector, the energy ratio (eRatio) is also calculated as follows:

[0095]

[0096] Using Equation 3 above, when the cutoff frequencies of both filters are correctly set, when the loading state (LS) is "ON", ideally E H (i) should be zero, resulting in R(i) being 1. Conversely, when LS is “OFF”, ideally E L (i) should be zero, and R(i) should also be 0.

[0097] Under practical conditions, the energy ratio determined using Equation 3 above is usually between zero and one.

[0098] For example, now, refer to Figure 6 , which shows a graph 610 of eRatio values ​​in a real-world test obtained by analyzing raw data, where the bandpass of the LBPF is set between 2 Hz and 5.5 Hz, and the bandpass of the HBPF is set between 5.5 Hz and 9 Hz. Figure 6 As shown, when the loading state is off, the eRatio is between about 0.1 and 0.5, and when the loading state is on, the eRatio is between 0.6 and 1.

[0099] In some cases, it may be easier to handle a loading state of "off" corresponding to -1 and a loading state of "on" corresponding to 1. In this case, the dependent variable of eRatioR'(i) may be defined as follows:

[0100]

[0101] therefore, Figure 6 can be converted to R′(i) Figure 7 , so that the cutoff between the on and off loading states is "0". In particular, in this case, the decision rule may be as follows:

[0102]

[0103] The above ratio is based solely on the phenomenon that the chassis spring's vibration frequency (VF) changes depending on the load mass. Specifically, a heavier load results in a lower VF. Load status is detected by comparing the energy outputs of the LBPF and HBPF using eRatio. However, sometimes when R'(i) approaches 0 (i.e., when R(i) approaches 0.5), detection becomes unreliable, potentially leading to erroneous detection.

[0104] Vibration amplitude

[0105] To improve the reliability of loading status detection, the variation in the loading state's vibration amplitude (VM) can also be considered. Specifically, when the chassis is empty and therefore light, the springs attached to the chassis will typically vibrate more freely, resulting in a higher VM. Conversely, when the chassis is loaded with heavy containers, the spring vibration amplitude is suppressed. In other words, the VM also carries certain information about the loading state.

[0106] Therefore, according to the embodiment described below, the loading state detection can be based on the energy ratio R(i) and the energy ratio R(i) represented by E(i). L (i) and E H The value of (i) represents the vibration amplitude of both identically.

[0107] Now, reference Figure 8 . Figure 8 The embodiment shows a graph 810 showing the relationship between eRatio and E L An example relationship between . Figure 8 As can be seen from the examples, when using eRatio and E L In particular, referring to the reading where the eRatio is 0.5 (as shown at point 820), no conclusion can be drawn using eRatio alone as to whether the trailer is loaded or unloaded. However, by using E L , which is easier to distinguish as no load.

[0108] To facilitate this separation, we can adopt a method widely used in machine learning (ML) and define a decision variable S(i) for the i-th test data. For example, if we want to use eRatio, E L and E H The value of , then the following equation 6 can be defined as follows:

[0109] S(i)=w0+w1R(i)+w2E L (i)+w3E H (i) (6)

[0110] Equation 6 simply means that S(i) is R(i), E L (i) and E HA weighted linear combination of (i) with appropriate weighting coefficients (WC) w j , j = 0, 1, 2 and 3. These WCs will be optimized.

[0111] If you only want to use eRatio and E L , then in Equation 6 above, w3 can be set to zero.

[0112] In some cases, the detector can be trained using the collected test results with known loading states to achieve optimization. Assuming that enough raw data has been collected from n tests with known loading states, for R(i), E L (i) and E H The raw data of (i) can be derived. Then, using these variables and the known LS, the optimal WC can be obtained through training as described below.

[0113] However, in some cases, training may be performed for one type of chassis, and then the value for WC may be propagated to sensor devices in similar chassis. In this case, training may only need to occur on a subset of chassis.

[0114] After training is complete or once the WC value is received at the sensor device, the above equation 6 can then be used to calculate the value of S(i). Once the value of S(i) is known, a loading status decision can be made according to equation 7 below.

[0115]

[0116] In the above equation 7, the value "0" is an arbitrarily assigned value determined during training. Those skilled in the art will appreciate that when using equation 6, the value of S(i) can be determined around any midpoint by changing WC. Therefore, equation 7 is merely an example of a decision equation.

[0117] train

[0118] Optimizing the weighting coefficients in Equation 6 above through training is a major topic in machine learning. Many well-known ML techniques exist, including but not limited to logistic regression (LR) and neural networks (NN). In these methods, starting with a set of randomly selected initial values ​​for WC, a certain number of iterative calculations can be performed. WC is updated at each iteration, eventually converging to a final value for WC.

[0119] In this document, two solutions are described as examples of methods for obtaining an optimal WC variable for an application involving cargo load detection. The first solution uses a minimum mean square error (MSE) method. The second solution uses a support vector machine (SVM) method. However, these solutions are provided for illustrative purposes only; in other cases, different optimization algorithms may be used to determine the value of the WC variable.

[0120] Without loss of generality, assume that we use L and E H The value of the decision is to decide the solution of the above equation 6. In particular, raw data has been collected from n tests with known loading conditions, and from this raw data, R(i), E and L (i) and E H (i), where i = 1, 2, ..., n. In machine learning, these variables are called features, and the known loading state is called the target.

[0121] Since there are n samples, each of which has three features associated with a target, for convenience, the features can be expressed as L (i) and E H (i) x i1 , x i2 and x i3 The target can be expressed as y i In one example, if LS(i) is known to be ON, then the target y i =1, and if LS(i) is known to be OFF, then the target y i =-1.

[0122] MSE Solutions

[0123] Using the minimum mean square error method, since y i is the goal of each test, so ideally the decision variable S(i)=y i Therefore, according to Equation 6 above, the equation for each value of i in the training set is as follows:

[0124]

[0125] The purpose of training is to determine i and x ij The solution to the four unknown WCs in (8) above is obtained with the known values ​​of . This can be done by using matrix and vector notation as described below.

[0126] In particular, the X matrix can be defined as follows:

[0127]

[0128] Furthermore, the y matrix can be defined as follows:

[0129] y=[y1 y2...y n ] T (10)

[0130] And, the w matrix can be defined as follows:

[0131] w=[w0 w1 w2 w3] T (11)

[0132] In (10) and (11) above, the superscript T indicates transpose. Using the matrices of (9), (10), and (11) above, the formula from (8) can be written in a concise form as follows:

[0133] y=Xw (12)

[0134] Mathematically, when the number of samples is greater than the number of features plus 1 (in this case, when n>4), then the equation in (8) has no solution. However, WC has a unique least squares solution:

[0135] w=X\y=(X T X) -1 X T y (13)

[0136] In the above equation 13, X\y is read as 'y divided left by X'. Further, if all columns of X are independent, then equation 13 is valid. Equation 13 results in the minimum mean square error between y and S=Xw.

[0137] Generally speaking, in order to accommodate m features, m>1, the matrix X of (9) consists of (m+1) columns, where the first column consists entirely of "1", and w consists entirely of (m+1) WCs.

[0138] SVM Solution

[0139] Support vector machine is a supervised binary classification algorithm of ML that can be applied to embodiments of the present disclosure. SVM is a robust process that can solve both simple machine learning problems with a large number of features and a small number of samples and highly complex machine learning problems.

[0140] After enough training iterations, the SVM determines a linear decision surface that can separate the two classes with the largest margin. In other words, it can determine a decision line in a two-dimensional model or a decision hyperplane in a multidimensional model.

[0141] Mathematically, SVM gives a solution for the weighting coefficient w that has minimum magnitude ||w|| subject to the following constraint (14):

[0142] YXw≥1 (14)

[0143] In (14) above, 1 is a full column vector, and Y is a diagonal matrix whose diagonal elements are y j , in order, j = 1, 2, ..., n. Fast algorithms have been developed to implement SVMs. Two non-limiting examples of such algorithms include John C. Platt, "Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines," Technical Report MSR-TR-98-14, April 21, 1998; and Stanford University, "The Simplified SMO Algorithm," CS229.

[0144] Multiple filter pairs

[0145] In the above embodiment, a single pair of bandpass filters is used to filter the raw data in order to make a loaded / unloaded decision. This assumes that the passbands represented by the cutoff frequencies of the two bandpass filters have been correctly selected.

[0146] In practice, in some embodiments, it is not known exactly what frequency band or bands should be selected as passbands for each of the two bandpass filters. In this case, according to an alternative embodiment of the present disclosure, multiple bandpass filter pairs can be utilized.

[0147] Specifically, the detector contains K LBPF and HBPF pairs, where K is greater than or equal to 1. Each filter pair is set with a different passband. The raw data from the sensor can then be applied to the K filter pairs simultaneously.

[0148] For example, Table 1 below illustrates a case where K=3, and shows the passband of each of the respective pairs.

[0149]

[0150] Table 1: Example of 3 BPF pairs and passbands

[0151] Therefore, each BPF pair has a different cutoff frequency as shown in Table 1. For the i-th test, according to the k-th BPF pair, k equals 1, 2, and 3, we can derive three variables: Rk (i) E k，L (i) E k，H (i), a total of nine variables / features are available.

[0152] When there are three bandpass filter pairs, the above equation 6 can be rewritten as follows:

[0153] S(i)=w0+w1R1(i)+w2E 1,L (i)+w3E 1,H (i)+w4R2(i)+w5E 2,L (i)+w6E 2,H (i)+w7R3(i)+w8E 3,L (i)+w9E 3,H (i) (15)

[0155] Equation 15 above can be solved by various techniques, including using MSC or SVM as described above. Further, in some cases, if E is ignored, k，H (i), then Equation 15 can be rewritten as follows:

[0156] S(i)=w0+w1R1(i)+w2E 1，L (i)+w3R2(i)+w4E 2,L (i)+w5R3(i)+w6E 3,L (i) (16)

[0157] Again, Equation 16 can be solved by various techniques including using MSE or SVM as described above.

[0158] Although the above example provides three bandpass filter pairs, this is for illustration only, and more or fewer bandpass filter pairs can be used. In particular, if only one bandpass filter pair is used, the above solution can be used. However, in some cases, two bandpass filter pairs or four or more bandpass filter pairs can be used.

[0159] Determining the loaded / unloaded state

[0160] In practice, to determine the loaded / unloaded state of a vehicle, a sensor device associated with the vehicle can measure vibration frequency and amplitude. In some cases, this raw data can be processed at the sensor device or other computing device associated with the vehicle. In other cases, the data can be provided to a server for processing.

[0161] For example, in Figure 9 A high-level overview of the process used to make a loaded / unloaded decision is provided in . Figure 9 In an embodiment, the process begins at block 910 and proceeds to block 912 where a determination is made as to whether training is required at a particular vehicle. For example, if training has been previously performed and various weighting coefficients have been previously determined, training may not be required. These weighting coefficients may be stored at a sensor device or computing device associated with the vehicle, or may be stored at a server. Thus, in other examples, the determination at block 912 may include, among other examples, determining whether the coefficients are stored at the sensor device / computing device or whether a flag is present at the sensor device / computing device to indicate that the coefficients are stored at the server.

[0162] In other cases, if the chassis type is similar to other chassis types for which training has already been performed, training may not be required. In this case, the server may pass the weighting coefficients to the sensor device, or may indicate to the sensor device that the weighting coefficients exist at the server.

[0163] In other cases, the server may issue a command to the sensor device to initiate a loaded / unloaded determination. In this case, the command may include a weighting factor or an indication that the weighting factor exists at the server. If this information is missing from the command, the sensor device or other computing device associated with the vehicle may conclude that training is required.

[0164] There may also be other examples that may not require training.

[0165] If training is required, the process proceeds from block 912 to block 920 where training is performed. Figure 10 、 Figure 11 and Figure 12 The execution of the training is described in more detail.

[0166] Once training has been performed, the process proceeds to block 930 where weighting coefficients are determined. Determining the weighting coefficients results in the weighting coefficients being stored at the server or sensor device.

[0167] After training is completed, or if training is not required at box 912, in some cases, the sensor device and the server can use the weighting coefficients previously determined or assigned to the vehicle to make a loaded / unloaded decision as shown in box 940.

[0168] The process then proceeds to block 950 and ends.

[0169] based on Figure 9Raw data generated by sensors associated with the vehicle is collected by a device mounted on the chassis. Training can be performed at the sensor device or via a remote server. The device or server can then make a load / no-load decision on new data.

[0170] If the decision and / or processing is performed by a server, a reliable communication link typically exists between the sensor device and the server.

[0171] With respect to training, training may occur entirely at a server or other similar computing device, e.g. Figure 10 occurs at both the sensor device and the server or computing device, e.g. Figure 11 as shown; or entirely on a sensor device or computing device on the vehicle, e.g. Figure 12 Other options or other distribution models are possible.

[0172] Now, reference Figure 10 , which shows the main aspects of training that occurs at server 1012. In particular, in some cases, sensor device 1010 can receive a command 1020 from server 1012 to begin the training process. However, command 1020 is optional, and in other cases, training can occur based on other factors (such as input from the vehicle driver, a check within sensor device 1010 that weighting coefficients are unknown, and other options).

[0173] Once training begins, either the sensor device or the server needs to have the loading status of the trailer or chassis input. This input can be based on various factors. For example, in some cases, the operator of the vehicle can input the loading status into the vehicle or sensor device. This can occur based on a user interface on the sensor device, a user interface on the vehicle itself, or a communication interface accessible to the operator of the vehicle. For example, in one case, the operator of the vehicle can be asked to enter the loading status into a text message that is sent to the server or sensor device during the training cycle. In other cases, the operator of the vehicle can be asked to press a button or toggle a switch to indicate whether the trailer is loaded or unloaded. Other options for providing information about the loading status of the vehicle to the sensor device or server are also possible.

[0174] In other cases, during training, the server may know whether a trailer is loaded or unloaded, for example based on a shipping schedule, and use this information to correlate with the sensor data.

[0175] In other cases, external input can be used to receive information about the loading status of a vehicle. For example, cameras in a shipping yard can provide a visual image of the vehicle as it leaves the yard, identifying whether the trailer is loaded or unloaded. This image can verify whether the container is loaded or unloaded, or, in some cases, display additional options such as wheels closer to the chassis on the closed container if the trailer is loaded.

[0176] In other cases, the sensor device may have secondary mechanisms for identifying whether the vehicle is unloaded or loaded, such as cameras, cargo loading detectors (such as light beam or laser projectors, and detectors for sensing whether an enclosed trailer includes cargo), among other options.

[0177] Therefore, if Figure 10 As shown, one or both of the sensor device 1010 or the server 1012 may receive the loading status based on one or more of the above-mentioned techniques. Figure 10 In the embodiment of FIG, if the sensor device 1010 is receiving the loading status, this is shown at block 1030. If the server 1012 is receiving the loading status, this is shown at block 1032.

[0178] Then, the sensor device 1010 may collect raw data regarding the frequency and amplitude of the vibration, as shown in block 1040. This may be done, for example, based on accelerometers or strain gauges, among other options.

[0179] Once the raw data is collected, the raw data may be sent to the server 1012, as indicated by message 1042. If a loading status was received at block 1030, the message 1042 may also include the loading status.

[0180] In some cases, message 1042 includes a complete set of log information to allow server 1012 to determine the weighting factors. In other cases, messages 1042 with partial information (e.g., a single log entry) may be sent periodically. In this case, the steps at blocks 1020, 1030, 1032, 1040, and 1042 may be repeated multiple times to allow server 1012 to compile enough data to determine the weighting factors.

[0181] Once the complete set of training logs is received at server 1012, the server may then calculate R as shown in block 1050. k (i) E k，L (i) and E k，H (i). For example, E k，L (i) and E k，H (i) can be calculated by applying one or more of the bandpass filter pairs. These values ​​can then be used to calculate Rk (i).

[0182] Once R is calculated k (i) E k，L (i) and E k，H (i), the process proceeds to block 1052 where weighting coefficients may be calculated. The weighting coefficients may be calculated using other options such as the MSE or SVM methods described above.

[0183] Once the weighting coefficients are calculated, the process proceeds to block 1054 where the weighting coefficients may be stored at the server, provided to the sensor device, or both.

[0184] At this point, training is complete and the weighting coefficients can then be used for loaded / unloaded determinations.

[0185] In some cases, further processing may be performed at the sensor device or similar computing device associated with the vehicle. Figure 11 .

[0186] exist Figure 11 In the embodiment of FIG. 1 , the sensor device 1110 communicates with a server 1112 .

[0187] as Figure 10 Same as the example, in Figure 11 In an embodiment of the present invention, server 1112 may signal sensor device 1110 to begin training, as shown at message 1120. However, message 1120 is optional.

[0188] Further, with Figure 10 Similar to the embodiment of , the loading status may be received at one or both of the sensor device 1110 or the server 1112 , for example, as shown at blocks 1130 and 1132 .

[0189] Then, as shown in block 1140, the sensor device 1110 collects raw data. Once a complete training log is compiled, Figure 11 In the embodiment of the present invention, the process then proceeds to block 1142 where the sensor device or a computing device associated with the vehicle calculates E based on one or more bandpass filters. k，L (i) and E k，H (i).

[0190] Once E is calculated k，L (i) and E k，H(i), the sensor device 1110 can provide E to the server 1112 in message 1144 k，L (i) and E k，H (i) Further, if a loading status is received at block 1130 , the loading status may also be provided to the server 1112 in a message 1144 .

[0191] Once all log information has been received in one or more messages 1144, the server 1112 may calculate R as shown in block 1150. k (i).

[0192] The process then proceeds to block 1152 where weighting coefficients may be calculated (eg, as provided above).

[0193] The process then proceeds to block 1154 where the server 1112 may store the weighting coefficients, send the weighting coefficients to the sensor device 1110 , or store and send the weighting coefficients to the sensor device 1110 .

[0194] Thereafter, the weighting coefficients may be used to calculate the loaded / unloaded status of the vehicle.

[0195] In yet another embodiment, the process can be performed entirely at the sensor device. Figure 12 .

[0196] exist Figure 12 In an embodiment, the process begins at block 1210. The process can begin, for example, based on a trigger to begin the training process. Such a trigger can be a command from a server, an indication that no weighting coefficients are present on the trailer, a manual indication that training is to occur, and other options.

[0197] The process then proceeds to block 1220 where the vehicle's loading status is input or received at the sensor device. The loading status can be provided to the sensor device based on, for example, manual input from a vehicle operator, from sensors on the vehicle or in the area surrounding the vehicle, based on commands or information provided by a server or other computing device, or other options.

[0198] The process then proceeds to block 1222 where raw data is collected. Such raw data may include readings from accelerometers and / or strain gauges, among other options.

[0199] From block 1222, the process proceeds to block 1230 where R is calculated based on one or more bandpass filter pairs. k (i) E k，L (i) and E k，H (i).

[0200] The process then proceeds to block 1240 where weighting coefficients are calculated based on the techniques described above or based on similar techniques.

[0201] The process then proceeds to block 1242 where the weighting coefficients are stored at the sensor device, sent to the server, or stored at the sensor device and sent to the server.

[0202] The process proceeds from block 1242 to block 1250 and ends.

[0203] Thereafter, the weighting coefficients calculated at block 1240 may be used to make a loaded / unloaded decision.

[0204] Once the training is complete, the weighting coefficients can be used by one or both of the sensor device and the server to calculate the loaded / unloaded state. Figure 13 and Figure 14 .

[0205] exist Figure 13 In an embodiment, whether the vehicle is loaded or unloaded is determined at a server. Specifically, sensor device 1310 communicates with server 1312. In some cases, server 1312 may provide command 1320 to sensor device 1310 to determine whether the vehicle is loaded or unloaded. In other cases, sensor device 1310 may periodically determine the loaded or unloaded state, or some other trigger at sensor device 1310 may initiate a process to determine whether the trailer is loaded or unloaded.

[0206] The process then proceeds to block 1330 where raw data including vibration frequency and vibration amplitude is acquired.

[0207] The raw data may then be provided to server 1312, as indicated by message 1332. In some cases, if the weighting coefficients are stored at sensor device 1310, message 1332 may also include the vehicle's weighting coefficients.

[0208] Based on receipt of message 1332, server 1312 may calculate R as provided above. k (i) E k，L (i) and E k，H (i). For example, this Figure 13 In the embodiment of FIG. 1 , block 1340 is completed.

[0209] The process then proceeds to block 1350, where the weighting coefficients stored at server 1312 or received in message 1332 can be used to determine the loaded / unloaded state of the vehicle. For example, Equations 6, 15, or 16 can be used to determine the decision variable S(i). Once the decision variable S(i) is determined, the decision variable can be input into Equation 7 to determine whether the vehicle is loaded or unloaded. For example, Equation 12 can be solved by placing the variables into the matrix X of Equation 9 above and then using w and X of Equation 12.

[0210] In other cases, the process can be completed entirely at the sensor device. Figure 14 .

[0211] Figure 14 The process begins at block 1410 and proceeds to block 1420 where a check is made to determine if a trigger condition has been met to perform a loaded / unloaded status check. As indicated above, such a trigger may include a message from a server, a timer expiration, an external factor such as entering or exiting a geographic location, manual input, and other options.

[0212] Once the trigger is met, the process proceeds to block 1422 where raw data from the vehicle is collected. This may include sensor readings from accelerometers and / or strain gauges, among other options.

[0213] From block 1422, the process proceeds to block 1430 where R is calculated as described above. k (i) E k，L (i) and E k，H (i).

[0214] The process then proceeds to block 1450, where the weighting coefficients stored at the sensor device (or received from the server) can be used to determine the loaded / unloaded state of the vehicle. For example, Equations 6, 15, or 16 can be used to determine the decision variable S(i). Once determined, the decision variable can then be input into Equation 7 to determine whether the vehicle is loaded or unloaded. For example, this can be solved by placing the variables into the matrix X of Equation 9 above and then using w and X from Equation 12.

[0215] In some cases, once the loaded / unloaded status is determined, the sensor device may report the status to a server (not shown).

[0216] The process then proceeds to block 1450 and ends.

[0217] therefore, Figures 9 to 14An embodiment utilizes the frequency of vibration and the amplitude of vibration to train and then determines the loaded / unloaded state.

[0218] Practical Example 1

[0219] The above embodiments were tested in a real-world environment. In particular, as shown in Table 2 in Appendix A below, relevant data with known loading states are provided. In Table 2, the third column lists the known loading states, where 1 indicates a load and -1 indicates no load. This loading state constitutes the vector y in Equation 12 above. The fourth, fifth, and sixth columns of Table 2 list the R k (i) E k，L (i) and E k，H (i). These columns form the last three columns of the matrix X in equation 9 above.

[0220] As a simple example, consider using two features (i.e., R(i) and E L (i)) Loading state detector. Use equation 13 for MSE to calculate and determine the weighting coefficient w MSE =[-1.4742, 3.1522, -0.1449] T On the other hand, for SVM, the weighting coefficient is w SVM =[-2.9758, 6.3533, -0.3371] T .

[0221] In order to see the improvement compared to the detection scheme based only on R'(i), the calculation is performed according to the following equation 17:

[0222] S=Xw (17)

[0223] The resulting S(i) is listed in columns 2 to 4 of Table 4 in Appendix C.

[0224] Figure 15A and Figure 15B The histogram distribution of the decision variables obtained from MSE relative to the decision variables obtained based only on eRatio is shown. In particular, Figure 15A A histogram distribution 1510 based only on eRatio is shown. Figure 15B A histogram distribution 1520 based on MSE is shown.

[0225] Figure 16A and Figure 16B The histogram distribution of the decision variables obtained from SVM relative to the decision variables obtained based only on eRatio is shown. In particular, Figure 16A A histogram distribution 1610 based only on eRatio is shown. Figure 16B A histogram distribution 1620 based on SVM is shown.

[0226] exist Figure 15A 、 Figure 15B 、 Figure 16A and Figure 16B In the example, it can be seen that the detection reliability using SVM is better than that using MSE, while MSE is better than the detector based on eRatio alone.

[0227] Figure 17 Shown in Figure 8 The decision line of MSE and SVM on the 2D plane. Figure 17 The line 1710 in FIG is the SVM decision line separating the two states, while the line 1720 is the MSE decision line. As can be seen, the SVM line gives the largest gap between the two data sets, for example, as seen by the -1 line 1730 and the +1 line 1740.

[0228] Practical Example 2

[0229] In another example, the outputs R(i) and E(i) of each of the three BPF pairs specified in Table 1 are L (i) For training SVM. The training data is provided in Table 3 of Appendix B below. Accordingly, the new decision variable is specified according to Equation 16 above.

[0230] After training with SVM, the weighted coefficients obtained are w = [-2.942, 3.433, 1.439, 2.098, 0.063, 0.407, -1.369] T .

[0231] The relevant data and results are listed in the last column of Table 4 in Appendix C below.

[0232] Figure 18A and Figure 18B The distribution of decision variables with 3 BPF pairs relative to the decision variables with BPF pairs based on eRatio of a single BPF pair is shown. In particular, Figure 18A A histogram distribution 1810 based only on eRatio is shown. Figure 18B A histogram distribution 1820 based on the weighted sum of the six features is shown. Figure 18B One benefit of an embodiment of is as follows: there is no need to select the passband of the BPF, now the passband is automatically determined by training.

[0233] server

[0234] A server such as servers 340, 342 or data service provider 350 may be any network node. Figure 19 , provides a simplified server that can execute the above embodiments.

[0235] exist Figure 19 In the embodiment, the server 1910 includes a processor 1920 and a communication subsystem 1930 , wherein the processor 1920 and the communication subsystem 1930 cooperate to perform the methods of the embodiments described herein.

[0236] The processor 1920 is configured to execute programmable logic that may be stored on the server 1910 along with data and Figure 19 In the example of FIG, 1920 is shown as memory 1940. Memory 1940 can be any tangible, non-transitory computer-readable storage medium, such as DRAM, flash memory, optical media (e.g., CD, DVD, etc.), magnetic media (e.g., tape), flash drive, hard drive, or other memory known in the art. In one embodiment, processor 1920 can also be implemented entirely in hardware and does not require any stored program to perform logical functions.

[0237] Alternatively or in addition to memory 1940 , server 1910 may access data or programmable logic from an external storage medium, such as through communication subsystem 1930 .

[0238] The communication subsystem 1930 allows the server 1910 to communicate with other devices or network elements.

[0239] In one embodiment, communication between the various elements of server 1910 may occur via internal bus 1960. However, other forms of communication may also occur.

[0240] The embodiments described herein are examples of structures, systems or methods with elements corresponding to the elements of the technology of the present application. This written description can enable those skilled in the art to manufacture and use embodiments with alternative elements corresponding equally to the elements of the technology of the present application. Therefore, the intended scope of the technology of the present application includes other structures, systems or methods that are not different from the technology of the present application described herein, and also includes other structures, systems or methods that are not substantially different from the technology of the present application as described herein.

[0241] In some embodiments, the present invention provides a plurality of program components and a plurality of software products, and the plurality of software products can be integrated into a single software product or packaged into a plurality of software products. In some cases, the function can be performed in hardware, and the plurality of software products can be integrated into a single software product or packaged into a plurality of software products.

[0242] In addition, the techniques, systems, subsystems, and methods described and illustrated as discrete or separate in various implementations may be combined or integrated with other systems, modules, techniques, or methods. Other items shown or discussed as coupled or directly coupled or communicating with each other may be indirectly coupled or communicating through an interface, device, or intermediate component electrically, mechanically, or otherwise. Other examples of changes, substitutions, and modifications may be determined and made by those skilled in the art.

[0243] While the above detailed description has shown, described, and pointed out the basic novel features of the present disclosure as applied to various implementations, it should be understood that various omissions, substitutions, and changes in form and detail of the illustrated system may be made by those skilled in the art. In addition, the order of method steps is not implied by the order in which they appear in the claims.

[0244] When messages are sent to or from an electronic device, such operations may not be immediate or directly from a server. They may be delivered synchronously or asynchronously from a server or other computing system infrastructure that supports the devices / methods / systems described herein. The aforementioned steps may include, in whole or in part, synchronous / asynchronous communications to / from the device / infrastructure. Moreover, the electronic device may communicate with one or more endpoints on the network. These endpoints may be served by servers, distributed computing systems, stream processors, etc. A content distribution network (CDN) may also provide communication with the electronic device. For example, in addition to typical server responses, the server may also provide or indicate data for a content delivery network (CDN) for the electronic device to download at a later time, such as subsequent activity of the electronic device. Thus, data may be sent directly from a server or other infrastructure (such as a distributed infrastructure or CDN) as part of the system or separately from the system.

[0245] In general, the storage medium may include any one or some combination of the following: semiconductor memory devices, such as dynamic or static random access memory (DRAM or SRAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), and flash memory; magnetic disks, such as fixed disks, floppy disks, and removable disks; another magnetic medium, including magnetic tape; optical media, such as compact disks (CDs) or digital video disks (DVDs); or other types of storage devices. Note that the instructions discussed above may be provided on one computer-readable storage medium or one machine-readable storage medium, or alternatively, may be provided on multiple computer-readable storage media or machine-readable storage media distributed in a large system having multiple nodes. Such one or more computer-readable storage media or machine-readable storage media are considered part of an article (or product). An article or product may refer to any manufactured single component or multiple components. The one or more storage media may be located in the machine that runs the machine-readable instructions, or may be located at a remote site from which the machine-readable instructions can be downloaded over a network for execution.

[0246] In the foregoing description, many details have been set forth to provide an understanding of the subject matter disclosed herein. However, implementations may be practiced without some of these details. Other implementations may include modifications and variations of the details discussed above. It is intended that the appended claims cover such modifications and variations.

[0247] Appendix A

[0248]

[0249]

[0250]

[0251] 2: Training data of a single BPF pair ((2~5.5)+(5.5-9)Hz) Appendix B

[0252]

[0253]

[0254]

[0255]

[0256]

[0257] Table 3: Training data for three BPF pairs Appendix C

[0258]

[0259]

[0260]

[0261]

[0262]

[0263] Table 4: Training results.

Claims

1. A method at a computing device, the method comprising: acquiring sensor data for the vehicle, the sensor data providing vibration frequency and amplitude; calculating an energy for each of a low frequency passband and a high frequency passband of the bandpass filter pair; determining an energy ratio based on the energy for the low frequency passband and the energy for the high frequency passband; applying a weighting constant to each of the following terms to calculate a decision variable: the energy of the low frequency passband, the energy of the high frequency passband, and the energy ratio; as well as If the decision variable is lower than a threshold, it is determined that the vehicle is unloaded, and if the decision variable is higher than a threshold, it is determined that the vehicle is loaded; The energy ratios include: Wherein eRatio is the energy ratio; E L is the energy for the low frequency passband; and E H is the energy for the high frequency passband.

2. The method according to claim 1, further comprising: The calculating, the determining, and the applying are repeated for a plurality of bandpass filter pairs.

3. The method according to claim 1, further comprising: Prior to the acquiring, the weighting constants are derived.

4. The method of claim 3, wherein the deriving comprises: A machine learning algorithm is applied to derive a plurality of known energies and energy ratios and known loading states for each of the low frequency passband and the high frequency passband pairs.

5. The method of claim 4, wherein the machine learning algorithm is a minimum mean square error algorithm.

6. The method of claim 5 , wherein each of the plurality of known energies and energy ratios and known loading conditions for the low-frequency passband and the high-frequency passband pair is represented as: y n =w0+w1x n1 +w2x n2 +…+w j x nj And wherein multiple equations can be solved for: y=Xw in and y=[y1 y2 … y n ] T and w=[w1 w2 ... w j ]。 7. The method of claim 4, wherein the machine learning algorithm is a support vector machine algorithm. The method of claim 1 , wherein the computing device is a sensor device on a vehicle.

9. The method of claim 1, wherein the computing device is a server remote from the vehicle.

10. A computing device comprising: processor; as well as Communication subsystem, The computing device is configured to: acquiring sensor data for the vehicle, the sensor data providing vibration frequency and amplitude; calculating an energy for each of a low frequency passband and a high frequency passband of the bandpass filter pair; determining an energy ratio based on the energy for the low frequency passband and the energy for the high frequency passband; applying a weighting constant to each of the following terms to calculate a decision variable: the energy for the low frequency passband, the energy for the high frequency passband, and the energy ratio; as well as If the decision variable is lower than a threshold, it is determined that the vehicle is unloaded, and if the decision variable is higher than a threshold, it is determined that the vehicle is loaded; The energy ratios include: Wherein eRatio is the energy ratio; E L is the energy for the low frequency passband; and E H is the energy for the high frequency passband. 11 . The computing device of claim 10 , wherein the computing device is further configured to repeat the calculating, the determining, and the applying for a plurality of bandpass filter pairs.

12. The computing device of claim 10, wherein the computing device is further configured to derive the weighting constant.

13. The computing device of claim 12, wherein the computing device is configured to derive a plurality of known energies and energy ratios and known loading states for each of the low-frequency passband and the high-frequency passband pairs by applying a machine learning algorithm.

14. The computing device of claim 13, wherein the machine learning algorithm is a minimum mean square error algorithm.

15. The computing device of claim 14 , wherein each of the plurality of known energies and energy ratios and known loading states of the low-frequency passband and the high-frequency passband pair is represented as: y n =w0+w1x n1 +w2x n2 +…+w j x nj And wherein multiple equations can be solved for: y=Xw in and y=[y1 y2 ... y n ] T and in=[in1 in2 … in j ] T 。 16. The computing device of claim 13, wherein the machine learning algorithm is a support vector machine algorithm.

17. The computing device of claim 10, wherein the computing device is a sensor device on a vehicle.

18. The computing device of claim 10, wherein the computing device is a server remote from the vehicle.

19. A computer-readable medium storing instruction codes which, when executed by a processor on a computing device, cause the computing device to: acquiring sensor data for the vehicle, the sensor data providing vibration frequency and amplitude; calculating an energy for each of a low frequency passband and a high frequency passband of the bandpass filter pair; determining an energy ratio based on the energy for the low frequency passband and the energy for the high frequency passband; applying a weighting constant to each of the following terms to calculate a decision variable: the energy for the low frequency passband, the energy for the high frequency passband, and the energy ratio; as well as If the decision variable is lower than a threshold, it is determined that the vehicle is unloaded, and if the decision variable is higher than a threshold, it is determined that the vehicle is loaded; The energy ratios include: Wherein eRatio is the energy ratio; E L is the energy for the low frequency passband; and E H is the energy for the high frequency passband.

Images