Longitudinal speed estimation method and system for distributed electric drive articulated vehicle

Abstract

The invention relates to the technical field of vehicle speed estimation, and provides a longitudinal speed estimation method and system for a distributed electrically-driven articulated vehicle, and the method comprises the steps: collecting data such as the wheel linear speed, the linear acceleration and the frame turning angle through a sensor, obtaining the longitudinal acceleration through an inertial sensor, and carrying out the integration to obtain the longitudinal speed; and projecting the centroid acceleration and the speed to a wheel grounding point in combination with the hinge geometric relationship and the turning angle to obtain a projection value. And calculating an absolute difference value between the measured value and the projection value, and judging whether the wheel completely slips or not according to the difference value through a finite-state machine. The vehicle speed is estimated through fuzzy logic self-adaptive Kalman filtering during non-complete slipping, and the vehicle speed is estimated through inertial measurement unit integration during complete slipping. According to the method, high precision, continuity and system robustness of vehicle speed estimation under all working conditions are guaranteed through a scheme of combining filtering and integration of adaptive working conditions.

Description

Technical Field

[0001] This invention relates to the field of vehicle speed estimation technology, and more specifically, to a method and system for estimating the longitudinal speed of a distributed electric drive articulated vehicle. Background Technology

[0002] The content in this section only provides background information related to this invention and may not constitute prior art.

[0003] Currently, research on longitudinal speed estimation technology is mostly focused on passenger cars and commercial vehicles. However, a mature and systematic method for estimating vehicle speed has not yet been developed for specialized engineering machinery such as articulated special vehicles. This research gap stems primarily from the technical challenges posed by the unique operating conditions of articulated special vehicles. These vehicles often operate in unstructured environments such as mines and material yards. Their dynamic digging and lifting operations cause real-time changes in vehicle load, exacerbating slippage and impact between the tires and complex road surfaces. This significantly degrades the accuracy and reliability of traditional methods based on wheel speed or inertial measurement.

[0004] Typical methods for estimating vehicle longitudinal speed include wheel speed-based methods, which calculate wheel speed by feeding back motor rotational speed and use the average value as the vehicle speed estimate. While simple and economical, this method suffers from a significant drop in accuracy when wheel slippage occurs. Another common approach is to integrate longitudinal acceleration using an inertial measurement unit (IMU) to obtain vehicle speed. Although suitable for four-wheel drive, this method suffers from inherent accumulation of integration errors, making it difficult to guarantee long-term estimation accuracy. Existing research has also attempted to integrate wheel speed, IMU, and Global Positioning System (GPS) signals for comprehensive estimation and correction. However, in complex environments such as mines and material sheds where articulated special vehicles often operate, conventional GPS signals are easily blocked or subject to multipath effects, resulting in insufficient stability and making it difficult to meet the requirements for continuous and reliable vehicle speed estimation. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for estimating the longitudinal velocity of a distributed electric-driven articulated vehicle, thereby improving the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows: In a first aspect, this application provides a method for estimating the longitudinal velocity of a distributed electric-driven articulated vehicle, comprising: Real-time data is acquired through corresponding sensors to obtain the linear velocity, linear acceleration, vehicle acceleration, and turning angle of the front and rear frames of the vehicle. The longitudinal acceleration of the vehicle is obtained by inertial sensors, and the longitudinal velocity is obtained by integrating the longitudinal acceleration. Based on the vehicle's folding steering structure and real-time vehicle attitude, the longitudinal acceleration and longitudinal velocity are projected onto the vehicle's center of mass. According to the folding angle and the spatial position of the four wheels, the longitudinal acceleration and longitudinal velocity located at the center of mass are projected onto the contact points of each wheel, respectively, to obtain the projected velocity and projected acceleration of the four wheels of the vehicle. Calculate the first difference between linear velocity and projected velocity, and the second difference between linear acceleration and projected acceleration; both the first and second differences are absolute values. The current state of the vehicle is identified by a finite state machine, and each wheel is determined to be a reliable wheel according to preset conditions. The preset conditions are that the first difference and the second difference simultaneously meet the corresponding preset range. Based on the number of reliable wheels within a corresponding time period, it is determined whether the vehicle is in a state of complete slippage within that time period. When the judgment result is a non-complete skidding state, the fuzzy logic adaptive Kalman filter method is used to estimate the longitudinal vehicle speed; when the vehicle is in a complete skidding state, the inertial measurement unit integration is used to estimate the longitudinal vehicle speed.

[0006] Furthermore, the linear velocity and linear acceleration of the four wheels are obtained, specifically including: The rotational speeds of the four wheels are obtained using corresponding sensors. Based on the rotational speeds and radii of the four wheels, the linear velocities of the four wheels are calculated. The corresponding linear acceleration is obtained by numerically differentiating each linear velocity.

[0007] Furthermore, the step of projecting the longitudinal acceleration and longitudinal velocity located at the center of mass to the contact points of each wheel specifically includes: Based on the articulated geometry, the relative orientational relationship between the front and rear frames of the vehicle, defined by the folding angle, is determined. Based on the relative orientation and the preset geometric position parameters of each wheel relative to the vehicle's hinge point, the longitudinal acceleration and longitudinal velocity at the center of mass are converted to the corresponding wheel contact points.

[0008] Furthermore, the step of determining whether the time period is a complete slippage state based on the number of reliable wheels within that time period specifically includes: Set a fixed-length time window and count the number of all reliable wheels within the time window. If a reliable wheel is continuously present within the window, the vehicle is determined to be in a non-complete skidding state; if there is no reliable wheel at any moment within the time window, the vehicle is determined to be in a complete skidding state.

[0009] Furthermore, a fuzzy logic adaptive Kalman filter method is used for longitudinal vehicle speed estimation, including: Based on the speed information obtained from the wheel speed sensor, a system state space model is constructed; the vehicle speed is used as the state vector, and the estimated vehicle speed at the previous moment is used as the predicted vehicle speed at the current moment. The difference between the theoretical residual variance and the actual residual variance during the Kalman filtering process is calculated, and the difference is input to the fuzzy logic controller. The fuzzy logic controller performs real-time reasoning and decision-making based on preset fuzzy rules, and adaptively adjusts the observation noise parameters in the Kalman filter. The Kalman gain is updated using the adjusted observation noise parameter, and then the predicted vehicle speed is corrected by combining the wheel speed observation value to obtain the longitudinal vehicle speed estimate at the current moment.

[0010] Furthermore, the formula for longitudinal vehicle speed estimation using the fuzzy logic adaptive Kalman filter method is as follows: In the formula, The prediction error covariance matrix; for The covariance matrix at time -1; for The error covariance matrix updated at time -1; For system state from Transfer to The state transition matrix; State transition matrix The transpose of the matrix; The process noise covariance matrix; Kalman gain; It is the identity matrix; This is the updated error covariance matrix; for The transpose of the matrix; The observation matrix; To measure the noise covariance matrix; To measure and update the system state vector, that is, the final system state at the current moment; For the updated state estimate; These are actual observed values.

[0011] Furthermore, the step of estimating longitudinal vehicle speed using inertial measurement unit integration specifically includes: Based on the preset integration time and initial velocity, the longitudinal acceleration measured by the inertial measurement unit is integrated over time to obtain the estimated vehicle speed at the current moment.

[0012] Furthermore, the initial speed is the estimated vehicle speed output at the previous moment.

[0013] Secondly, this application also provides a longitudinal velocity estimation system for a distributed electric-driven articulated vehicle, comprising: The data acquisition module is used to acquire real-time data through corresponding sensors, including the linear velocity and linear acceleration of the four wheels, the vehicle acceleration, and the turning angle of the front and rear frames of the vehicle. The projection module is used to acquire the longitudinal acceleration of the vehicle through an inertial sensor, integrate the longitudinal acceleration to obtain the longitudinal velocity; based on the vehicle's folding steering structure and real-time vehicle attitude, the longitudinal acceleration and the longitudinal velocity are projected onto the vehicle's center of gravity; according to the folding angle and the spatial position of the four wheels, the longitudinal acceleration and longitudinal velocity located at the center of gravity are projected onto the contact points of each wheel respectively to obtain the projected velocity and projected acceleration of the four wheels of the vehicle. The difference calculation module is used to calculate the first difference between linear velocity and projected velocity, and the second difference between linear acceleration and projected acceleration; both the first and second differences are absolute values. The judgment module is used to identify the current state of the vehicle through a finite state machine and determine whether each wheel is a reliable wheel according to preset conditions. The preset conditions are that the first difference and the second difference simultaneously meet the corresponding preset range. Based on the number of reliable wheels within a corresponding time period, it is determined whether the vehicle is in a complete slippage state within that time period. The longitudinal speed estimation module is used to estimate the longitudinal speed by employing a fuzzy logic adaptive Kalman filter method when the judgment result is not a complete skid state; when the vehicle is in a complete skid state, it uses inertial measurement unit integration to estimate the longitudinal speed.

[0014] The beneficial effects of this invention are as follows: This invention collects real-time data from various corresponding sensors, covering wheel linear velocity, linear acceleration, vehicle acceleration, and the turning angles of the front and rear frames. Simultaneously, it uses an inertial sensor positioned at the vehicle's center of gravity to acquire longitudinal acceleration, which is then integrated to obtain longitudinal velocity. Considering the vehicle's zigzag steering structure, and combining the articulated geometry with real-time vehicle attitude, based on the turning angle and the spatial position information of each wheel, the longitudinal acceleration and longitudinal velocity at the center of gravity are projected onto the contact points of each wheel, thus obtaining the projected velocity and projected acceleration of each wheel. Subsequently, the absolute differences between the measured linear velocity and projected velocity, and the absolute differences between the measured linear acceleration and projected acceleration of the wheels are calculated. A finite state machine is used to identify the vehicle's current operating state. Based on the condition that both sets of absolute differences simultaneously meet a corresponding preset range, each wheel is determined to be a reliable wheel. The number of reliable wheels within a given time period determines whether the vehicle is in a state of complete slippage during that time period. In a non-complete slippage state, a fuzzy logic adaptive Kalman filter method is used for longitudinal velocity estimation; in a complete slippage state, an inertial measurement unit integration method is used for longitudinal velocity estimation. This ensures that the filter parameters can match the real-time changing working conditions of articulated special vehicles and output the optimal high-precision longitudinal speed estimate, while minimizing the growth of integral error and guaranteeing the continuity of speed estimate output and the overall robustness of the system under all working conditions. Attached Figure Description

[0015] Figure 1 A schematic flowchart of a method for estimating the longitudinal velocity of a distributed electric-driven articulated vehicle provided by the present invention; Figure 2 This is a schematic diagram of the right-turning bending steering structure in this invention; Figure 3 This is a schematic diagram of the left turn of the bending steering structure in this invention; Figure 4 This invention provides a schematic diagram of the longitudinal speed estimation system for a distributed electric-driven articulated vehicle. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0017] Example 1: like Figure 1 As shown in the embodiment of the present invention, a longitudinal speed estimation method for a distributed electric drive articulated vehicle includes: S101 acquires real-time data through corresponding sensors, including the linear velocity and linear acceleration of the four wheels, the vehicle acceleration, and the turning angle of the front and rear frames of the vehicle.

[0018] Specifically, real-time data is acquired through corresponding sensors, including the linear velocity and linear acceleration of the four wheels, vehicle acceleration, and the turning angles of the front and rear frames. These corresponding sensors are dedicated sensor assemblies matched to each measurement parameter, specifically including: speed sensors for each of the four wheels, an inertial sensor mounted on the rear frame, and angle sensors mounted at the vehicle's articulation points. The principle is that the longitudinal velocity estimation of a distributed electric drive articulated vehicle relies on the coordinated support of multi-dimensional motion parameters. Wheel speed parameters reflect the local motion state of each wheel, vehicle acceleration reflects the motion trend of the vehicle's center of gravity, and the turning angles of the front and rear frames are adapted to the vehicle's unique swerving structure, compensating for errors caused by neglecting frame deflection in conventional vehicle estimation methods.

[0019] In obtaining the linear velocity and linear acceleration of the four wheels, one approach is to use wheel speed sensors to obtain the wheel rotation speed, calculate the linear velocity using the rotation speed and wheel radius, and then obtain the linear acceleration by differential conversion of the linear velocity. Alternatively, sensors can be used directly to obtain the data. For example, the linear velocity of the wheels can be obtained using a Doppler radar velocity sensor, which emits microwaves or millimeter waves towards the ground. By measuring the Doppler frequency shift of the reflected wave due to relative motion, the longitudinal linear velocity and linear acceleration of the sensor relative to the ground can be directly calculated.

[0020] Then, the linear velocity is numerically differentiated to obtain the corresponding linear acceleration. The principle is based on the definition of acceleration in kinematics. Acceleration is the rate of change of velocity with respect to time. By performing numerical differentiation on the linear velocity signal that is collected and calculated in real time, the trend of linear velocity change with time, i.e., linear acceleration, can be obtained.

[0021] The linear velocity of each wheel is expressed as follows: (1) The linear acceleration expressions for each wheel are: (2) In the formula, , , , These are the rotational speeds of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively, obtained from the wheel speed sensor signals. , , , These are the linear velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively, calculated from the wheel speed sensor signals. , , , These are the reciprocals of the corresponding linear velocities; , , , These are the accelerations of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively, calculated from the wheel speed sensor signals. S102: The longitudinal acceleration of the vehicle is obtained through an inertial sensor, and the longitudinal velocity is obtained by integrating the longitudinal acceleration. Based on the vehicle's folding steering structure and real-time vehicle attitude, the longitudinal acceleration and longitudinal velocity are projected onto the vehicle's center of mass. According to the folding angle and the spatial position of the four wheels, the longitudinal acceleration and longitudinal velocity located at the center of mass are projected onto the contact points of each wheel respectively, so as to obtain the projected velocity and projected acceleration of the four wheels of the vehicle.

[0022] Specifically, the longitudinal acceleration of the vehicle is first acquired using an inertial sensor mounted on the rear frame. Second, the longitudinal acceleration is integrated to obtain the longitudinal velocity. This process is based on classical kinematics, where acceleration is the rate of change of velocity with respect to time. Given the initial velocity, by integrating the longitudinal acceleration signal over time, the velocity change process can be reversed, yielding the longitudinal velocity parameters. Then, based on the vehicle's flexing steering structure and real-time vehicle attitude, the longitudinal acceleration and velocity are projected onto the vehicle's center of mass. Specifically, the vehicle's articulation geometry, composed of fixed structural parameters such as the distance from the center of mass to the front axle, the distance from the center of mass to the rear axle, the front axle track width, the rear axle track width, and the relative position of the articulation point to the center of mass, serves as the static basis for the projection calculation. Simultaneously, the real-time vehicle attitude is sensed collaboratively by inertial and angle sensors. The turning angle collected by the angle sensor directly reflects the relative deflection direction and degree of the front and rear frames. Based on the aforementioned fixed structural parameters and dynamic parameters, a coordinate relationship system between the front and rear frames in a two-dimensional plane is established. The conversion coefficients between the local motion parameters of the rear frame and the global motion parameters of the center of mass are clarified. Then, through vector projection calculation in rigid body kinematics, the longitudinal acceleration of the rear frame measured by the inertial sensor and the longitudinal velocity obtained by integration are accurately converted to the center of mass of the vehicle, thus obtaining the longitudinal acceleration and longitudinal velocity corresponding to the center of mass.

[0023] Simultaneously, regarding the vehicle's flexing steering structure, based on the articulated geometry of the flexing steering structure and the real-time vehicle attitude, such as... Figure 2 and Figure 3As shown, by combining the turning angle and the spatial position of the four wheels, the longitudinal acceleration and longitudinal velocity located at the center of gravity are projected onto the contact points of each wheel to obtain the projected velocity and projected acceleration of the four wheels of the vehicle. First, it is important to clarify that the core structural feature of a distributed electric drive articulated vehicle is the relative deflection of the front and rear frames through a central articulation pin, i.e., a folding steering structure. This structure results in a significant geometric difference between the motion trajectory at the center of gravity and the motion trajectory of each wheel contact point during vehicle operation (especially under steering conditions). The acceleration and velocity at the center of gravity reflect the overall motion state of the vehicle, while the motion state of each wheel contact point is affected by factors such as the frame deflection angle and wheel mounting position, exhibiting localized characteristics. If the motion parameters at the center of gravity are directly compared with the linear velocity and linear acceleration of the wheels themselves, a large error will be introduced due to the geometric mismatch, leading to distortion in subsequent reliable wheel identification and vehicle speed estimation results.

[0024] The specific implementation of the step of projecting the longitudinal acceleration and longitudinal velocity located at the center of mass to the contact points of each wheel is as follows: Based on the articulated geometry, the relative orientation relationship between the front and rear frames of the vehicle, defined by the turning angle, is determined. Fixed structural parameters in the articulated geometry (such as the distance from the center of mass to the front and rear axles, wheelbase, etc.) form the static basis for constructing the relative orientation relationship between the front and rear frames, while the turning angle is the core variable for dynamically adjusting this relationship. When the vehicle is traveling straight, the turning angle is zero, the front and rear frames are collinear, and the relative orientation relationship is parallel or coaxial. When the vehicle turns left or right, the turning angle exhibits the corresponding angular value, the front and rear frames form an angle, and the relative orientation relationship shows a deflection state. The principle is to establish a coordinate association system between the front and rear frames in a two-dimensional plane through the collaborative calculation of fixed geometric parameters and real-time turning angles, clarifying the transformation coefficients of the center of mass motion parameters in the respective coordinate systems of the front and rear frames. Then, based on the relative orientation and the preset geometric position parameters of each wheel relative to the vehicle's hinge point, the longitudinal acceleration and longitudinal velocity at the center of mass are transformed to the corresponding wheel contact points. The preset geometric position parameters of each wheel relative to the vehicle's hinge point are fixed values ​​determined through calibration during the vehicle design phase. These parameters reflect the precise coordinates of each wheel (front left, front right, rear left, rear right) relative to the hinge point in the vehicle's coordinate system, ensuring the uniqueness and traceability of each wheel's spatial position. The principle of the transformation process is to use the determined relative orientation of the front and rear frames as the coordinate transformation reference, specifically decomposing and mapping the acceleration and velocity vectors at the center of mass according to the preset coordinate positions of each wheel. Through vector projection operations in rigid body kinematics, the global center of mass motion parameters are transformed into local motion parameters (i.e., projected velocity and projected acceleration) at each wheel contact point.

[0025] The articulation geometry refers to the kinematic relationship system composed of fixed structural parameters determined during the vehicle design phase. Specifically, it includes preset parameters such as the distance from the center of gravity to the front axle, the distance from the center of gravity to the rear axle, the front axle track width, the rear axle track width, and the relative position of the articulation point and the center of gravity. These parameters form the basis for establishing the kinematic mapping relationship between the center of gravity and the wheel contact points. Real-time vehicle attitude is perceived collaboratively by inertial sensors and angle sensors, comprehensively reflecting the vehicle's driving posture in three-dimensional space (such as straight driving, left turning, right turning, tilting, etc.), ensuring that the projection calculation can adapt to the vehicle's dynamically changing driving state. The turning angle, as a core dynamic parameter, is collected in real-time by angle sensors installed at the articulation points. It directly determines the relative deflection direction and degree of the front and rear frames, and is crucial for correcting the geometric deviation between the center of gravity and the wheel motion parameters. The four-wheel spatial position, i.e., the preset coordinate position of each wheel contact point relative to the articulation point and the center of gravity, clarifies the fixed installation orientation of each wheel on the entire vehicle, providing a precise spatial reference for the projection calculation.

[0026] The specific calculation process is as follows: Taking the left turn of an articulated special vehicle as an example, the acceleration at the mounting point is obtained based on the inertial sensor. The expression is: (3) In the formula, , These are the longitudinal acceleration and lateral acceleration at the inertial sensor mounting location, i.e., the rear frame.

[0027] The speed at the installation location is obtained based on inertial sensors. The expression is: (4) In the formula, The turning angle of an articulated special vehicle at the articulation point; The initial velocity is ; , These are the distances from the center of mass to the front and rear axles, respectively. This refers to the rear axle track.

[0028] The acceleration of the center of mass can be obtained through geometric analysis and calculation. and center of mass velocity for: (5) (6) Based on the acceleration of the center of mass, the accelerations and velocities of the four wheels are calculated as follows: (7) (8) In the formula, This refers to the front axle track. , , , These are the accelerations of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively, calculated by the inertial sensors. , , , These are the linear velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively, calculated by the inertial sensors.

[0029] S103, calculate the first difference between linear velocity and projected velocity, and the second difference between linear acceleration and projected acceleration; both the first and second differences are absolute values.

[0030] The design of using absolute values ​​for both the first and second differences follows the core logic of slippage judgment. The principle is that the difference between linear velocity and projected velocity, and the difference between linear acceleration and projected acceleration, can be positive or negative. A positive value indicates that the actual wheel motion parameters are greater than the theoretical parameters (e.g., slippage caused by excessive acceleration of the drive wheels), while a negative value indicates that the actual parameters are less than the theoretical parameters (e.g., slippage caused by wheel lock-up during braking). However, the core of slippage judgment is the "degree of deviation" rather than the "direction of deviation." Regardless of whether the actual parameters are larger or smaller, it indicates that the wheel motion state is inconsistent with the ideal state driven by the entire vehicle, and both belong to unreliable wheel scenarios that need to be identified. Using absolute values ​​effectively removes the directional attributes of the differences, retaining only the quantitative information of the degree of deviation, ensuring the uniformity and fairness of subsequent judgment standards, avoiding misjudgments due to different deviation directions, and allowing the differences to directly reflect the severity of the deviation, providing a reasonable basis for setting a unified judgment threshold.

[0031] S104, the current state of the vehicle is identified by a finite state machine, and each wheel is determined to be a reliable wheel according to preset conditions; the preset conditions are that the first difference and the second difference simultaneously meet the corresponding preset range; based on the number of reliable wheels within the corresponding time period, it is determined whether the vehicle is in a complete slippage state within the time period.

[0032] Specifically, the operating conditions of distributed electric-driven articulated vehicles are highly complex and dynamic, with wheel states frequently switching between slippage and non-slippage. Finite state machines (FSMs), with their clear state divisions, explicit state transition logic, and robust decision-making characteristics, can accurately capture the dynamic changes in wheel states, providing a reliable basis for adaptive state determination in subsequent estimation algorithms. Here, the current vehicle state specifically refers to the wheel slippage category, i.e., whether each wheel is slipping and the overall degree of slippage of the vehicle. The identification process uses the differences in motion parameters of each wheel as the core input, and achieves a precise mapping from local wheel states to the overall vehicle state through the logical judgment of the finite state machine. Determining whether each wheel is a reliable wheel based on preset conditions is essentially a consistency check between the actual wheel motion and the ideal motion state of the vehicle—a reliable wheel is defined as one that has not slipped and whose motion parameters accurately reflect the vehicle's driving state, while the preset conditions are the quantitative standards for achieving this consistency check.

[0033] In detail, the preset conditions are that both the first and second differences must simultaneously meet their corresponding preset ranges. The motion state of the wheel needs to be comprehensively judged from two core dimensions: velocity and acceleration. Differences in parameters of a single dimension cannot fully and accurately identify the slippage state. Specifically, the first difference is the absolute difference between the wheel's linear velocity and its projected velocity. Linear velocity, as a direct representation parameter of wheel motion, directly reflects the synchronicity between the wheel's rolling state and the vehicle's movement when compared to the projected velocity (the wheel's velocity under ideal vehicle motion). The second difference is the absolute difference between the wheel's linear acceleration and its projected acceleration. Acceleration, as the rate of change of velocity, reflects the consistency between the wheel's acceleration and the vehicle's acceleration trend when compared to the projected acceleration. Only when both differences are within the preset reasonable range can it be concluded that the wheel has not slipped, its motion parameters are consistent with the ideal vehicle motion state, and it possesses the reliability to serve as a benchmark for vehicle speed estimation. The “preset range” here is not a subjective setting, but is calibrated based on the dynamic characteristics of distributed electric drive articulated vehicles, the measurement accuracy of sensors, and a large amount of measured data in typical operating scenarios such as mining areas and material yards. Its upper and lower limits can effectively distinguish between “normal motion deviation” and “abnormal deviation caused by slippage”, which not only ensures the sensitivity of slippage state recognition, but also avoids misjudgment caused by accidental factors such as sensor noise and slight road bumps, providing a quantitative basis for the accurate identification of reliable wheels.

[0034] The expression corresponding to this state is: (9) (10) (11) (12) In the formula, The threshold corresponding to acceleration, The threshold corresponding to the speed.

[0035] Finally, based on the number of reliable wheels within a given time period, it is determined whether the vehicle is in a complete slippage state. The design principle of this step is that when distributed electric-driven articulated vehicles operate in unstructured environments, the wheels may be subjected to instantaneous disturbances such as road bumps, gravel impacts, and transient load changes, causing brief anomalies in the difference at a single moment. If the judgment is based solely on the result of a single moment, such instantaneous disturbances can easily be misjudged as slippage, leading to unnecessary switching of the estimation algorithm and compromising the stability and continuity of vehicle speed estimation. Therefore, introducing statistical analysis over time, by accumulating the number of reliable wheels over a period of time, can effectively filter out random errors caused by instantaneous disturbances, more accurately reflect the vehicle's true slippage state, and ensure the robustness of the state judgment results. The specific implementation method is as follows: Firstly, a fixed-length time window is set, and the number of all reliable wheels within the time window is counted. The continuous time series is divided into discrete time intervals. By centrally statistically analyzing the state data within each interval, a phased assessment of the vehicle's state is achieved. The length of the time window is comprehensively calibrated based on the typical slippage characteristics of distributed electric drive articulated vehicles, sensor sampling frequency, and real-time performance indicators of vehicle speed estimation—assuming a sampling frequency of 100Hz, the time window length is set to capture at least 3-5 sampling points, ensuring both sufficient statistical samples and keeping the delay in state judgment within an acceptable range. Counting the number of all reliable wheels within the time window is to comprehensively grasp the overall distribution trend of the wheel states during that time period, rather than relying on isolated data from individual moments. Cumulative statistics amplify the characteristics of the actual slippage state, weaken the impact of instantaneous interference, and further improve the accuracy of state judgment.

[0036] If a reliable wheel is consistently present within the time window, the vehicle is determined to be in a state of incomplete slippage. The principle behind this is that the "consistent presence of a reliable wheel" indicates that at least one wheel remains slip-free within that time window, and its wheel speed signal remains consistent with the ideal motion state of the entire vehicle. This provides a reliable benchmark for vehicle speed estimation and stable, effective data support for wheel speed-based fusion estimation algorithms. Therefore, it is determined to be in a state of incomplete slippage. Conversely, if no reliable wheels are present at any point within the time window, the vehicle is determined to be in a state of complete slippage. The principle behind this is that the "absence of reliable wheels at any point" indicates that all four wheels are continuously slipping within that time window. Their wheel speed signals are completely detached from the actual driving state of the vehicle and cannot be used as a valid basis for vehicle speed estimation. Continuing to rely on wheel speed signals for estimation would lead to significant estimation errors or even estimation divergence. Therefore, it is determined to be a state of complete slippage.

[0037] The expression corresponding to this state is: (13) (14) (15) (16) S105, when the judgment result is not a complete skidding state, the fuzzy logic adaptive Kalman filter method is used to estimate the longitudinal vehicle speed; when the vehicle is in a complete skidding state, the inertial measurement unit integration is used to estimate the longitudinal vehicle speed.

[0038] Specifically, when the vehicle is determined to be in a "partially slipping state," the system employs a fuzzy logic adaptive Kalman filter method for longitudinal vehicle speed estimation. The core basis for this selection is that in a partially slipping state, there is at least one reliable wheel, whose wheel speed signal can accurately reflect the core motion characteristics of the vehicle, providing reliable data support for fusion estimation based on wheel speed information. Furthermore, the fuzzy logic adaptive Kalman filter method, through optimization of the traditional Kalman filter, effectively overcomes the insufficient adaptability of fixed-parameter filtering under time-varying noise conditions. The specific implementation process, principle, and beneficial effects are as follows: First, a system state-space model is constructed based on the speed information obtained from wheel speed sensors. Vehicle speed is used as the state vector, and the estimated vehicle speed from the previous moment is used as the predicted vehicle speed for the current moment. From a theoretical perspective, the system state-space model is the theoretical basis for Kalman filtering to achieve optimal estimation. For the longitudinal speed estimation requirements of distributed electric-driven articulated vehicles, vehicle speed is chosen as the core state vector because, as the target estimation parameter of this invention, its dynamic changes directly characterize the core motion law of vehicle driving, simplifying the complexity of the state model while ensuring the focus of the estimation target. Using the estimated vehicle speed from the previous moment as the predicted value for the current moment is based on the dynamic continuity characteristics of vehicle driving—at the high sampling frequency used in this invention, the vehicle driving state at adjacent moments has a smooth transition characteristic, and the vehicle speed will not experience sudden changes without warning. This setting can provide a physically reasonable initial prediction benchmark for the current vehicle speed estimation, effectively reducing the impact of initial prediction deviation on the final estimation result.

[0039] Secondly, the difference between the theoretical residual variance and the actual residual variance during the Kalman filtering process is calculated, and this difference is input into the fuzzy logic controller. The principle is that the estimation performance of traditional Kalman filtering highly depends on the accurate modeling of the observed noise parameters. However, the operating scenarios of distributed electric articulated vehicles (such as unstructured environments like mines and material yards) involve complex factors such as road impacts, real-time load changes, and dynamic adjustments in the wheel-ground interaction relationship, leading to significant time-varying characteristics in the observed noise. Fixed observed noise parameters cannot adapt to these dynamically changing conditions. The theoretical residual variance is an ideal value calculated based on a preset system model and initial noise parameters, reflecting the expected performance of the filtering algorithm under ideal conditions. The actual residual variance is the true value calculated based on the deviation between the measured data and predicted values ​​from the wheel speed sensor, directly reflecting the degree of fit between the noise characteristics under actual conditions and the system model. The difference between the two can quantitatively characterize the degree of matching between the current observed noise parameters and the actual working conditions. The reason for using this difference as the input of the fuzzy logic controller is that the fuzzy logic controller has the unique advantage of handling nonlinear and uncertain problems. It can accurately interpret and make decisions on the difference signal by simulating the empirical reasoning logic of experts in the field, and provide an objective basis for adjusting the observed noise parameters.

[0040] Secondly, the fuzzy logic controller performs real-time reasoning and decision-making based on preset fuzzy rules to adaptively adjust the observed noise parameters in the Kalman filter. The principle is that the preset fuzzy rules are a reasoning system built based on typical operating conditions of distributed electric-driven articulated vehicles, sensor measurement characteristics, and a large amount of measured data, combined with the engineering experience of domain experts. The core is to establish a nonlinear mapping relationship between the "degree of difference in residual variance" and the "adjustment magnitude of observed noise." When the difference in residual variance is small, it indicates that the current observed noise parameters are well-suited to the actual operating conditions, requiring only minor adjustments or remaining unchanged. When the difference in residual variance is large, it indicates that the current observed noise parameters have deviated from the actual operating conditions, requiring targeted adjustments (reducing or increasing the observed noise parameters) based on the direction of the difference (theoretical residual variance greater than actual residual variance or vice versa). The fuzzy logic controller performs a series of processes, including fuzzification of the input residual variance difference, rule matching reasoning, and defuzzification output, ultimately generating specific adjustment amounts for the observed noise parameters, achieving dynamic adaptive adjustment of the observed noise parameters according to changes in operating conditions.

[0041] Finally, the Kalman gain is updated using the adjusted observation noise parameter, and then the predicted vehicle speed is corrected by combining the wheel speed observations to obtain the longitudinal vehicle speed estimate at the current moment. The principle is that the Kalman gain is a key parameter for balancing the weight allocation between predicted and observed values, and its value is negatively correlated with the observation noise parameter. When the observation noise parameter is small, it indicates that the reliability of the wheel speed observations is high, and the Kalman gain increases accordingly, making the filtering result more reliant on the wheel speed observations for correction. When the observation noise parameter is large, it indicates that the wheel speed observations are severely affected by noise, reducing their reliability, and the Kalman gain decreases accordingly, making the filtering result more reliant on the predicted values ​​to suppress the adverse effects of observation noise. By updating the Kalman gain with the adjusted observation noise parameter, the weight allocation can be adapted in real time to the reliability of the wheel speed signal under the current operating conditions. Combined with the wheel speed observations, the initial predicted vehicle speed is iteratively corrected, achieving an optimal fusion of the continuity of the predicted value and the accuracy of the observed value.

[0042] The calculation process corresponding to this state is as follows: First, state prediction is performed, and its calculation formula is as follows: (17) (18) In the formula, In order to be in Time based Information at any given time is used to predict the state vector; This is the state transition matrix; This is the optimal estimate of the state vector at time k-1 based on information from time k-1 and earlier. To control the input matrix; The control input is known; In order to be in Time's up Process noise at any given moment; In order to be in Time based The observed value at time; For observing the transition matrix; This is a noise parameter.

[0043] The state vector at the current moment is obtained by balancing the weights between GPS observations and inertial sensor measurement predictions using the Kalman gain parameter. Finally, the covariance matrix at that moment is updated using a formula to complete the prediction process of optimal state estimation. The specific calculation steps are as follows: (19) (20) (twenty one) (twenty two) In the formula, The prediction error covariance matrix; for The covariance matrix at time -1; for The error covariance matrix updated at time -1; For system state from Transfer to The state transition matrix; State transition matrix The transpose of the matrix; The process noise covariance matrix; Kalman gain; It is the identity matrix; This is the updated error covariance matrix; The observation matrix; for The transpose of the matrix; To measure the noise covariance matrix; To measure and update the system state vector, that is, the final system state at the current moment; For the updated state estimate; These are actual observed values.

[0044] In the Kalman filter design, the state vector is the vehicle speed. Since the actual sampling frequency is 100Hz and the sampling period is short, the vehicle speed can be approximated as constant during this time, i.e.: (twenty three) In the formula, for Vehicle speed at any given time; for The vehicle speed at the moment preceding the current moment.

[0045] Wheel speed directly acquired from sensors can be used to estimate vehicle speed, but slippage and changes in tire radius introduce significant errors. Integrating the vehicle body acceleration directly acquired from sensors can also provide a speed estimate, but various disturbances and inherent acceleration biases also lead to substantial errors. Observational noise constantly changes under different operating conditions, but Kalman filtering algorithms typically use a fixed value for this noise, which introduces errors into the estimation results. A fuzzy controller can be used to adjust the observational noise in real time. This further compensates for errors in the Kalman filter, improving the filter's filtering performance. Definition The system observation noise at time t is: (twenty four) In the formula, This is the new system observation noise covariance matrix updated at time k+1; Let k be the system observation noise covariance matrix at time k; This is the adjustment amount of the observation noise covariance matrix calculated by the fuzzy controller.

[0046] The theoretical value of the residual covariance of the Kalman filter is: (25) In the formula, This represents the theoretical value of the residual covariance. The observation matrix; This is the transpose of the observation matrix; Let K+1 be the state prediction error covariance matrix. This is the adjusted observation noise covariance matrix.

[0047] The actual residual is: (26) In the formula, For residuals; Let k be the measurement vector at time k; This is the observation matrix at time k+1.

[0048] The variance of the actual residuals is: (27) for The estimated value; for The transpose of the matrix; Let be the covariance matrix at time k-1; For observing the transition matrix; for The transpose of .

[0049] The fuzzy controller uses the difference between the theoretical variance and the actual variance of the residuals. As input, with This is the output. When the theoretical variance of the residuals is very close to the actual variance, i.e. Then keep Unchanged. When the theoretical variance of the residuals is less than the actual variance, i.e. It should be increased .

[0050] Furthermore, when the vehicle is determined to be in a "complete skid state," the system uses inertial measurement unit (IMU) integration for longitudinal vehicle speed estimation. The core rationale for this choice is that wheel speed signals from all wheels are unreliable in a complete skid state and cannot be used as a valid basis for speed estimation. The IMU, however, has the characteristic of providing continuous longitudinal acceleration information without relying on external signals, ensuring the continuity of vehicle speed estimation. The specific implementation process, principles, and beneficial effects are as follows: Based on a preset integration time and initial velocity, the longitudinal acceleration measured by the inertial measurement unit is integrated over time to obtain the vehicle speed estimate at the current moment; the initial velocity is the vehicle speed estimate output at the previous moment. The principle is that, according to classical kinematics, acceleration is the rate of change of velocity with respect to time. Given the initial velocity, by performing time integration on the longitudinal acceleration signal, the velocity change process can be reconstructed, thereby calculating the vehicle speed at the current moment. The preset integration time is set based on the typical duration of a fully slipped state in a distributed electric drive articulated vehicle. This integration time is relatively short, and the initial value is updated with each integration. This effectively avoids misjudgment problems caused by local noise or wheel speed fluctuations, ensuring the continuity and stability of the longitudinal velocity estimation results. The initial velocity is the vehicle speed estimate output at the previous moment because the vehicle speed estimate obtained through fuzzy logic adaptive Kalman filtering when the vehicle was in a partially slipped state at the previous moment has high accuracy. Using this as the initial integration value can minimize the initial offset of the integration error while ensuring the continuity of the vehicle speed estimation during the switching of operating conditions. Its beneficial effects are that, in the state of complete slippage where all wheel speed signals fail, continuous output of vehicle speed estimation is achieved through IMU (inertial sensor) integration, avoiding the problem of estimation interruption caused by unreliable wheel speed signals; the preset integration time and the initial speed design based on the high-precision estimate of the previous moment effectively suppress the inherent error accumulation defect of IMU integration, ensuring the accuracy and stability of vehicle speed estimation in the state of complete slippage, and guaranteeing the robustness of the estimation system under all working conditions.

[0051] Example 2: like Figure 2 As shown, based on the same inventive concept, this embodiment provides a longitudinal velocity estimation system for a distributed electric drive articulated vehicle, including: The data acquisition module 201 is used to acquire real-time data through corresponding sensors, including the linear velocity and linear acceleration of the four wheels, the vehicle acceleration, and the turning angle of the front and rear frames of the vehicle. The projection module 202 is used to acquire the longitudinal acceleration of the vehicle through an inertial sensor, integrate the longitudinal acceleration to obtain the longitudinal velocity; based on the vehicle's folding steering structure and real-time vehicle attitude, the longitudinal acceleration and longitudinal velocity are projected onto the vehicle's center of mass; according to the folding angle and the spatial position of the four wheels, the longitudinal acceleration and longitudinal velocity located at the center of mass are projected onto the contact points of each wheel respectively to obtain the projected velocity and projected acceleration of the four wheels of the vehicle. The difference calculation module 203 is used to calculate a first difference between the linear velocity and the projected velocity, and a second difference between the linear acceleration and the projected acceleration; both the first difference and the second difference are absolute values. The judgment module 204 is used to identify the current state of the vehicle through a finite state machine, and to determine whether each wheel is a reliable wheel according to preset conditions; the preset conditions are that the first difference and the second difference simultaneously meet the corresponding preset range; and to determine whether the time period is a complete slippage state according to the number of reliable wheels in the corresponding time period. The longitudinal speed estimation module 205 is used to estimate the longitudinal speed by using a fuzzy logic adaptive Kalman filter method when the judgment result is not a complete skid state; and to estimate the longitudinal speed by using inertial measurement unit integration when the vehicle is in a complete skid state.

[0052] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0053] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for estimating the longitudinal velocity of a distributed electric-driven articulated vehicle, characterized in that, include: Real-time data is acquired through corresponding sensors to obtain the linear velocity, linear acceleration, vehicle acceleration, and turning angle of the front and rear frames of the vehicle. The longitudinal acceleration of the vehicle is obtained by an inertial sensor, and the longitudinal velocity is obtained by integrating the longitudinal acceleration. Based on the vehicle's flexing steering structure and real-time vehicle attitude, the longitudinal acceleration and the longitudinal velocity are projected onto the vehicle's center of mass. Based on the turning angle and the spatial position of the four wheels, the longitudinal acceleration and longitudinal velocity located at the center of mass are projected onto the contact points of each wheel to obtain the projected velocity and projected acceleration of the four wheels of the vehicle. Calculate the first difference between the linear velocity and the projected velocity, and the second difference between the linear acceleration and the projected acceleration; both the first difference and the second difference are absolute values. The current state of the vehicle is identified by a finite state machine, and each wheel is determined to be a trustworthy wheel based on preset conditions. The preset condition is that the first difference and the second difference simultaneously satisfy the corresponding preset range; based on the number of reliable wheels within the corresponding time period, it is determined whether the time period is a complete slippage state; When the judgment result is a non-complete skidding state, the fuzzy logic adaptive Kalman filter method is used to estimate the longitudinal vehicle speed; when the vehicle is in a complete skidding state, the inertial measurement unit integration is used to estimate the longitudinal vehicle speed.

2. The longitudinal velocity estimation method for a distributed electric drive articulated vehicle according to claim 1, characterized in that, The acquisition of the linear velocity and linear acceleration of the four wheels specifically includes: The rotational speeds of the four wheels are obtained using corresponding sensors. Based on the rotational speeds and radii of the four wheels, the linear velocities of the four wheels are calculated respectively. The corresponding linear acceleration is obtained by numerical differentiation of each linear velocity.

3. The longitudinal velocity estimation method for a distributed electric drive articulated vehicle according to claim 1, characterized in that, The step of projecting the longitudinal acceleration and longitudinal velocity located at the center of mass to the contact points of each wheel specifically includes: Based on the articulated geometry, the relative orientation between the front and rear frames of the vehicle, as defined by the folding angle, is determined. Based on the relative orientation relationship and the preset geometric position parameters of each wheel relative to the vehicle hinge point, the longitudinal acceleration and longitudinal velocity at the center of mass are respectively converted to the corresponding wheel contact point.

4. The longitudinal velocity estimation method for a distributed electric drive articulated vehicle according to claim 1, characterized in that, The step of determining whether the time period is a complete slippage state based on the number of reliable wheels within the corresponding time period specifically includes: A fixed-length time window is set, and the number of all reliable wheels within the time window is counted. If a reliable wheel is continuously present within the window, the vehicle is determined to be in a non-complete skidding state; if there is no reliable wheel at any moment within the time window, the vehicle is determined to be in a complete skidding state.

5. The longitudinal velocity estimation method for a distributed electric drive articulated vehicle according to claim 1, characterized in that, The longitudinal vehicle speed estimation using the fuzzy logic adaptive Kalman filter method includes: Based on the speed information obtained from the wheel speed sensor, a system state space model is constructed; the vehicle speed is used as the state vector, and the estimated vehicle speed at the previous moment is used as the predicted vehicle speed at the current moment. The difference between the theoretical residual variance and the actual residual variance during the Kalman filtering process is calculated, and the difference is input to the fuzzy logic controller. The fuzzy logic controller performs real-time reasoning and decision-making based on preset fuzzy rules, and adaptively adjusts the observation noise parameters in the Kalman filter. The Kalman gain is updated using the adjusted observation noise parameter, and then the predicted vehicle speed is corrected by combining the wheel speed observation value to obtain the longitudinal vehicle speed estimate at the current moment.

6. The longitudinal velocity estimation method for a distributed electric drive articulated vehicle according to claim 5, characterized in that, The formula for longitudinal vehicle speed estimation using the fuzzy logic adaptive Kalman filter method is as follows: In the formula, The prediction error covariance matrix; for The covariance matrix at time -1; for The error covariance matrix updated at time -1; For system state from Transfer to The state transition matrix; State transition matrix The transpose of the matrix; The process noise covariance matrix; Kalman gain; It is the identity matrix; This is the updated error covariance matrix; The observation matrix; for The transpose of the matrix; To measure the noise covariance matrix; To measure and update the system state vector, that is, the final system state at the current moment; For the updated state estimate; These are actual observed values.

7. The longitudinal velocity estimation method for a distributed electric drive articulated vehicle according to claim 1, characterized in that, The step of estimating longitudinal vehicle speed using inertial measurement unit integration specifically includes: Based on the preset integration time and initial velocity, the longitudinal acceleration measured by the inertial measurement unit is integrated over time to obtain the estimated vehicle speed at the current moment.

8. The longitudinal velocity estimation method for a distributed electric drive articulated vehicle according to claim 7, characterized in that, The initial speed is the estimated vehicle speed output at the previous moment.

9. A longitudinal velocity estimation system for a distributed electric drive articulated vehicle, based on the longitudinal velocity estimation method for a distributed electric drive articulated vehicle as described in claim 1, characterized in that, include: The data acquisition module is used to acquire real-time data through corresponding sensors, including the linear velocity and linear acceleration of the four wheels, the vehicle acceleration, and the turning angle of the front and rear frames of the vehicle. The projection module is used to acquire the longitudinal acceleration of the vehicle through an inertial sensor, integrate the longitudinal acceleration to obtain the longitudinal velocity, and project the longitudinal acceleration and the longitudinal velocity onto the vehicle's center of mass based on the vehicle's slant steering structure and real-time vehicle attitude. Based on the turning angle and the spatial position of the four wheels, the longitudinal acceleration and longitudinal velocity located at the center of mass are projected onto the contact points of each wheel to obtain the projected velocity and projected acceleration of the four wheels of the vehicle. The difference calculation module is used to calculate the first difference between the linear velocity and the projected velocity, and the second difference between the linear acceleration and the projected acceleration; Both the first difference and the second difference are expressed in absolute value. The judgment module is used to identify the current state of the vehicle through a finite state machine and determine whether each wheel is a trustworthy wheel based on preset conditions. The preset condition is that the first difference and the second difference simultaneously satisfy the corresponding preset range; based on the number of reliable wheels within the corresponding time period, it is determined whether the time period is a complete slippage state; The longitudinal speed estimation module is used to estimate the longitudinal speed by employing a fuzzy logic adaptive Kalman filter method when the judgment result is not a complete skid state; when the vehicle is in a complete skid state, it uses inertial measurement unit integration to estimate the longitudinal speed.

Images