A sweeping vehicle driving strategy self-adaptive adjustment system in response to a garbage overflow state

Abstract

The application discloses a sweeping vehicle driving strategy self-adaptive adjustment system responding to a garbage full overflow state, relates to the technical field of vehicle motion control, and comprises the following steps: taking a probability error as a sliding mode surface variable, calculating a speed scaling factor and a speed variable, constructing a target function and a constraint condition, using DSQP to optimize and solve the target function, obtaining an execution vehicle speed and a steering angle instruction, and issuing the execution vehicle speed and the steering angle instruction to a vehicle-mounted controller through an API interface for execution; through multi-source current and loading rate fusion modeling, high-sensitivity identification of the garbage full overflow state is realized; through disturbance spectrum intensity triggering and noise adaptive updating, loading state estimation stability is improved; and through overflow probability constraint driving strategy optimization, safety and efficiency collaborative control is realized.

Description

Technical Field

[0001] This invention relates to the field of vehicle motion control technology, and in particular to an adaptive adjustment system for the driving strategy of a road sweeper in response to a garbage overflow state. Background Technology

[0002] Existing road sweepers use speed sensors, garbage bin level or mass sensors, and motor current sensors to monitor the sweeping and loading status. They then combine preset operating parameters to complete sweeping, vacuuming, and garbage collection operations. Furthermore, they incorporate onboard controllers and back-end management platforms, enabling the uploading of operating status, alarms for abnormalities, and remote adjustment of operating routes via communication interfaces. This has made progress in improving operating efficiency and reducing manual intervention, and has become an important development direction for current smart sanitation equipment.

[0003] However, existing technologies still have shortcomings. Current loading state estimation methods use fixed noise parameter adjustment, which cannot analyze the frequency domain characteristics of disturbance energy during operation. This leads to over-response or long-term deviation in state estimation after disturbance, making it difficult to achieve fast and stable estimation recovery after disturbance. Current sweeper driving strategy adjustment is based on rule triggering and does not introduce the risk of garbage overflow into the control target in a probabilistic form. It is difficult to achieve coordinated optimization control of speed and steering while ensuring vehicle stability, safety and operation efficiency. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an adaptive adjustment system for the driving strategy of a sweeper truck in response to the state of overflowing garbage. This solves the problem that existing loading state estimation methods use fixed noise parameter adjustment, which cannot analyze the frequency domain characteristics of disturbance energy during operation. This leads to over-response or long-term deviation in state estimation after a disturbance occurs, making it difficult to achieve fast and stable estimation recovery after a disturbance. Existing sweeper truck driving strategy adjustment is based on rule triggering and does not introduce the risk of overflowing garbage into the control objective in a probabilistic form. This makes it difficult to achieve coordinated optimization control of speed and steering while ensuring vehicle stability, safety and operation efficiency.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides an adaptive adjustment system for the driving strategy of a road sweeper in response to a garbage overflow state, comprising:

[0008] The disturbance sensing module is used to collect data from the vehicle controller and calibration parameters of the vehicle controller, obtain the original disturbance observations and the fused loading rate observations, construct an autocorrelation estimation sequence for the original disturbance observations, and calculate the spectral intensity index based on the autocorrelation estimation sequence.

[0009] The noise conditioning module is used to calculate residual energy based on the fused loading rate observations, obtain the noise covariance through the vehicle controller, construct the final noise update increment by combining the spectral intensity index, make update judgments based on the sampling period, update the final noise update increment and perform upper and lower bound clipping to obtain the final noise covariance, combine the predicted state vector generated by the UKF method and the predicted covariance to obtain the innovation quantity and innovation covariance, and calculate the loading rate mean and loading rate standard deviation.

[0010] The probability correction module is used to calculate the dimensionless threshold distance based on the mean and standard deviation of the load rate, construct the intercept correction term based on the innovation quantity, obtain skewness and kurtosis through the quantile regression equation, correct the dimensionless threshold distance, and input the standard normal CDF to obtain the probability error.

[0011] The strategy optimization module is used to calculate the speed scaling factor and speed variable by using the probability error as the sliding surface variable, construct the objective function and constraints, use DSQP to optimize and solve the objective function, obtain the vehicle speed and steering angle commands, and send them to the vehicle controller for execution through the API interface.

[0012] As a preferred embodiment of the adaptive adjustment system for the sweeper's driving strategy in response to a garbage overflow state as described in this invention, the step of calculating residual energy based on fused loading rate observations and obtaining noise covariance through the onboard controller includes:

[0013] The difference between the fused loading rate observations in sampling period t-1 and sampling period t is calculated to obtain the innovation residual;

[0014] Based on statistical analysis, the productivity parameter is set, the relaxation sampling period constant is calculated, the product of the relaxation sampling period constant and the sampling frequency is calculated, and the result is rounded up to obtain the number of window length periods.

[0015] Read the noise covariance of the vehicle controller sampling period t-1, calculate the product with the scaling factor, and obtain the noise covariance of the sampling period t.

[0016] Calculate the ratio of residual energy to noise covariance of sampling period t, obtain the standard deviation and squared residuals, calculate the consistency loss, and calculate the allowable update coefficient;

[0017] Read the calibrated maximum spectral intensity of the vehicle controller, calculate the ratio of the spectral intensity index to the maximum spectral intensity, obtain the normalized spectral intensity, and calculate the membership degree.

[0018] As a preferred embodiment of the adaptive adjustment system for the sweeper's driving strategy in response to a garbage overflow state as described in this invention, wherein: the construction of the final noise update increment includes:

[0019] Based on membership, the basic noise increment is calculated, the variable point sampling period and relaxation window length of the vehicle controller are read, and the attenuation factor is generated through exponential decay.

[0020] The final noise update increment is obtained by multiplying the attenuation factor, the allowable update coefficient, and the basic noise increment.

[0021] As a preferred embodiment of the adaptive adjustment system for the sweeper's driving strategy in response to a garbage overflow state as described in this invention, the step of updating the final noise update increment and pruning its upper and lower bounds to obtain the final noise covariance includes:

[0022] If the sampling period t is less than or equal to the sum of the variable point sampling period and the relaxation window length, then the window flag is set to 1; otherwise, it is set to 0.

[0023] If it is 0, no noise adjustment is performed. The noise covariance of the sampling period t is set to be equal to the noise covariance of the sampling period t-1, and it is marked as frozen covariance. If it is 1, the noise update amount is adjusted to obtain the measured noise covariance.

[0024] The frozen covariance and the measurement noise covariance are clipped by upper and lower bounds to obtain the final noise covariance.

[0025] As a preferred embodiment of the adaptive adjustment system for the sweeper's driving strategy in response to a garbage overflow state as described in this invention, wherein: the predicted state vector and predicted covariance generated using the UKF method include:

[0026] Read the state estimation vector and covariance of the vehicle controller during the sampling period t-1;

[0027] The unscented Kalman filter (UKF) method is used to generate the predicted state vector and predicted covariance for the sampling period t-1.

[0028] The predicted state vector is mapped to the measurement space through a measurement function to obtain the predicted measurement mean.

[0029] As a preferred embodiment of the adaptive adjustment system for the sweeper's driving strategy in response to garbage overflow as described in this invention, the step of obtaining the innovation quantity and innovation covariance, and calculating the mean and standard deviation of the loading rate, includes:

[0030] The innovation quantity is obtained by calculating the difference between the observed value and the predicted mean of the fusion loading rate, and the innovation covariance is obtained by calculating the sum of the predicted covariance and the final noise covariance.

[0031] Based on the innovation covariance and innovation quantity, the mean and standard deviation of the loading rate are calculated.

[0032] As a preferred embodiment of the adaptive adjustment system for the sweeper's driving strategy in response to garbage overflow, as described in this invention, the following steps are taken: skewness and kurtosis are obtained through quantile regression equations, the dimensionless threshold distance is corrected, and a standard normal CDF is input to obtain the probability error, including:

[0033] Based on historical experimental experience, an overflow boundary threshold is set. The difference between the overflow boundary threshold and the mean loading rate is divided by the standard deviation of the loading rate to obtain the dimensionless threshold distance.

[0034] The innovation quantity is standardized using innovation covariance to obtain the standardized innovation quantity. The intercept of the standardized innovation quantity is corrected to obtain the intercept correction term. The empirical quantiles are obtained through the quantile regression equation. Skewness and kurtosis are constructed. The dimensionless threshold distance is corrected to obtain the corrected threshold distance.

[0035] The corrected threshold distance is input into the standard normal CDF to obtain the corrected tail probability. The difference between the corrected tail probability and the target overflow probability is calculated to obtain the probability error.

[0036] As a preferred embodiment of the adaptive adjustment system for the sweeper truck's driving strategy in response to a garbage overflow state as described in this invention, the step of using DSQP to optimize the objective function to obtain the commands for vehicle speed and steering angle, and then sending them to the onboard controller for execution via an API interface, includes:

[0037] The probability error is set as a sliding surface variable, the speed scaling factor is calculated to obtain the speed variable, the objective function is constructed and the constraints are set, the objective function is optimized and solved using DSQP to obtain the vehicle speed and steering angle commands, which are then issued and executed through the API interface.

[0038] As a preferred embodiment of the adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in this invention, the step of constructing an autocorrelation estimation sequence for the original disturbance observations and calculating the spectral intensity index based on the autocorrelation estimation sequence includes:

[0039] Set a fixed window length of N, collect the original perturbation observations within the fixed window, obtain the mean-removed perturbation window sequence, calculate the autocorrelation estimation sequence, and calculate the power spectral density function value;

[0040] Within the vehicle controller, a set of frequency bands is set based on engineering calibration. The power spectral density function values ​​are summed within the set of frequency bands to obtain the spectral intensity index.

[0041] As a preferred embodiment of the adaptive adjustment system for the sweeper's driving strategy in response to a garbage overflow state as described in this invention, the step of collecting onboard controller data and onboard controller calibration parameters to obtain the original disturbance observation and the fused loading rate observation includes:

[0042] By obtaining vehicle controller data and calibration parameters through the API interface, the loading rate variable and the height channel loading rate are calculated and weighted summed to obtain the fused loading rate observation value.

[0043] The beneficial effects of this invention are as follows: This invention uses probability error as a sliding surface variable, calculates speed scaling factor and speed variable, constructs objective function and constraints, uses DSQP to optimize and solve the objective function, obtains the execution speed and steering angle commands, and sends them to the vehicle controller for execution through API interface; it achieves highly sensitive identification of garbage overflow state, improves the stability of loading state estimation, and realizes coordinated control of safety and efficiency. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a schematic diagram of the adaptive adjustment system for the sweeper's driving strategy in response to the overflowing garbage state in Example 1.

[0046] Figure 2 This is a schematic diagram of the adaptive noise adjustment system of the sweeper truck driving strategy adaptive adjustment system in response to the garbage overflow state in Example 1. Detailed Implementation

[0047] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0050] Example 1, referring to Figure 1 and Figure 2This is the first embodiment of the present invention, which provides an adaptive adjustment system for the driving strategy of a road sweeper in response to a garbage overflow state, including the following steps:

[0051] S1, Disturbance Sensing Module, is used to collect data from the vehicle controller and calibration parameters of the vehicle controller, obtain the original disturbance observations and the fused loading rate observations, construct an autocorrelation estimation sequence for the original disturbance observations, and calculate the spectral intensity index based on the autocorrelation estimation sequence;

[0052] Specifically, data from the onboard controller and its calibration parameters are collected to obtain the original disturbance observations and fused load rate observations, including:

[0053] Data from the vehicle controller is obtained through the API interface, including the garbage bin mass, garbage height, vehicle speed, lateral acceleration, roller motor current, and suction fan current.

[0054] The sum of the current of the roller motor and the current of the suction fan is calculated to obtain the original observation of the disturbance;

[0055] Obtain the vehicle controller calibration parameters through the API interface, including empty box reference mass, maximum safe mass, empty box reference height, and maximum safe height;

[0056] Set the idle reference mass to zero and the maximum safe mass to the full-scale endpoint.

[0057] The mass and height of the trash can were normalized using calibration coefficients to obtain the loading rate variable and the height channel loading rate.

[0058] We obtain the fused load rate observation by weighted summation of the load rate variable and the height channel load rate.

[0059] By integrating the data collected by the vehicle controller and its calibration parameters, the original disturbance observation and the fused loading rate observation are constructed, which significantly improves the accuracy and real-time response of vehicle load status perception. By summing multiple sensor parameters, such as the current of the roller motor and the suction fan, a single indicator for disturbance observation is formed, which effectively avoids the error superposition problem caused by multi-variable coupling. The normalization of garbage quality and height not only enhances the universality between different vehicles, but also provides a stable foundation for the accurate quantification of subsequent loading status. The weighted composition mechanism of the fused loading rate further avoids the problem of single channel dominance and realizes the structural optimization of multi-source indicators, which is conducive to load status estimation in dynamic driving tasks.

[0060] Furthermore, an autocorrelation estimation sequence is constructed for the original perturbation observations, and spectral intensity indices are calculated based on the autocorrelation estimation sequence, including:

[0061] Set a fixed window length of N, collect the raw observations of disturbances within the fixed window, and obtain the disturbance window sequence, as shown in the formula:

[0062] ;

[0063] in, For the perturbation window sequence, The sampling sequence number at the end of the window. For fixed window length, To disturb the original observations;

[0064] Based on the perturbation window, the mean of the original perturbation observations is calculated, and the mean is subtracted from the original perturbation observations to obtain the mean-free perturbation window sequence.

[0065] Setting the maximum lag order to M, the autocorrelation estimate sequence is calculated based on the mean-removed perturbation window sequence, using the following formula:

[0066] ;

[0067] in, The value is the autocorrelation estimate, representing an element in the autocorrelation estimation sequence. The lag order is... The element values ​​in the mean-free perturbation window sequence, The index of the element position in the mean-free perturbation window sequence;

[0068] Based on the autocorrelation estimation sequence, the power spectral density function value is calculated using the following formula:

[0069] ;

[0070] in, The power spectral density function value. The angular frequency variable is obtained by taking values ​​point by point from the preset discrete frequency table of the on-board controller. The maximum lag order, The base of the natural constant is... The imaginary unit;

[0071] Within the vehicle controller, a set of frequency bands is set based on engineering calibration. The power spectral density function values ​​are summed within the set of frequency bands to obtain the spectral intensity index.

[0072] By introducing autocorrelation estimation and spectral intensity index, a quantitative tool is provided for frequency domain identification of dynamic disturbances, realizing an effective transition from the time domain to the frequency domain. By constructing a fixed window disturbance sequence and combining it with mean removal processing, trend interference in the original signal is eliminated, making the calculated autocorrelation sequence more realistically reflect the periodic characteristics. The introduction of the power spectral density function provides an evaluation standard for the energy concentration of the frequency band. Especially with the assistance of the calibration frequency band set, specific disturbance modes can be identified in a targeted manner. The spectral intensity index serves as the core basis for subsequent disturbance triggering and update logic, and its accuracy and robustness are crucial to the stability of the system response mechanism.

[0073] S2, Noise Adjustment Module, is used to calculate residual energy based on fused loading rate observations, obtain noise covariance through the vehicle controller, construct the final noise update increment by combining spectral intensity index, make update judgments based on sampling period, update the final noise update increment and perform upper and lower bound clipping to obtain the final noise covariance, combine the predicted state vector and predicted covariance generated by the UKF method to obtain innovation quantity and innovation covariance, and calculate the loading rate mean and loading rate standard deviation;

[0074] Specifically, based on the fused loading rate observations, the residual energy is calculated, and the noise covariance is obtained through the onboard controller, including:

[0075] The difference between the fused loading rate observations of sampling period t-1 and sampling period t is calculated to obtain the innovation residual, and the square of the innovation residual is set as the residual energy.

[0076] If the spectral intensity index is greater than or equal to the calibration threshold (set based on empirical rules), the event flag is set to 1, indicating that subsequent update operations will be performed; otherwise, it is set to 0, the noise covariance of the sampling period t-1 is retained, and no subsequent update operations are performed.

[0077] Based on statistical analysis, productivity parameters are set, and the relaxation sampling period constant is calculated using the following formula:

[0078] ;

[0079] in, To relax the sampling period constant, For productivity parameters, For sampling period index;

[0080] Calculate the product of the relaxation sampling period constant and the sampling frequency, and round up to obtain the number of window length periods;

[0081] Deploy a family of candidate noise models in the vehicle controller and set the model task to generate candidate noise scaling factors.

[0082] Read the noise covariance of the vehicle controller sampling period t-1, calculate the product with the scaling factor (based on engineering calibration settings, such as conservative, moderate and aggressive levels, corresponding to the three scaling factors), and obtain the noise covariance of the sampling period t.

[0083] Calculate the ratio of residual energy to the noise covariance of the sampling period t to obtain the standard deviation and squared residuals, and then calculate the consistency loss using the following formula:

[0084] ;

[0085] ;

[0086] in, For the first Measurement noise covariance of each candidate model For the first Scaling factor of each candidate model Sampling period The noise covariance, For the first The standard deviation and squared residuals of the candidate models For the first The standard deviation and squared residuals of the candidate models This results in a loss of consistency.

[0087] Based on the consistency loss, the allowable update coefficient is calculated using the following formula:

[0088] ;

[0089] in, To allow for coefficient updates, The consistency loss threshold is set based on rules of thumb;

[0090] Read the calibrated maximum spectral intensity of the vehicle controller, calculate the ratio of the spectral intensity index to the maximum spectral intensity, and obtain the normalized spectral intensity.

[0091] Based on the aforementioned endpoints, including the zero point and the full-scale endpoint, and combined with the normalized spectral intensity, the membership degree is calculated using the following formula:

[0092] ;

[0093] ;

[0094] in, To ensure stable membership, For intense membership, Zero point The full-scale endpoint, This represents the normalized spectral intensity.

[0095] By introducing a coupling judgment mechanism between residual energy and noise covariance, the update action of disturbance events becomes more conditionally based. By setting a spectral intensity trigger threshold, the covariance is dynamically updated only when high-intensity disturbances occur, avoiding frequent ineffective adjustments and reducing system energy consumption and computational burden. At the same time, by introducing a relaxation sampling period constant and window length period, the update rhythm is further made time-controlled. The deployment of the noise model family provides configurable model alternatives for different driving strategies, while the calculation of consistency loss ensures the optimality and stability of the selected model, realizing closed-loop optimization from observation bias to model adjustment.

[0096] Furthermore, the final noise update increment is constructed, including:

[0097] Based on membership degree, the fundamental noise increment is calculated using the following formula:

[0098] ;

[0099] in, Based on the basic noise increment, The minimum basic increment is set based on the stability test of the on-board controller. This is the endpoint for amplifying severe disturbance noise (engineering calibration parameter).

[0100] The variable-point sampling period and relaxation window length of the vehicle controller are read, and an attenuation factor is generated through exponential decay, as shown in the formula:

[0101] ;

[0102] in, As the attenuation factor, For variable sampling period, For the number of periods of the window length;

[0103] The final noise update increment is obtained by multiplying the attenuation factor, the allowable update coefficient, and the basic noise increment.

[0104] By modeling the membership of stable and violent disturbances, piecewise weighted adjustment of noise increments is achieved, avoiding runaway response under violent disturbances. The introduction of the basic noise increment allows the system to make incremental adjustments based on the disturbance properties, taking into account both the controller's response speed and control stability. The introduction of the exponential decay factor constructs a dynamic penalty mechanism based on the timing logic of variable point sampling and relaxation window, so that the impact of early disturbances on the current increment gradually decreases, effectively suppressing the disturbance memory effect and enhancing the foresight and dynamic adaptability of noise adjustment.

[0105] Furthermore, the final noise update increment is updated and its upper and lower bounds are clipped to obtain the final noise covariance, including:

[0106] If the sampling period t is less than or equal to the sum of the variable point sampling period and the relaxation window length, then the window flag is set to 1; otherwise, it is set to 0.

[0107] If the value is 0, no noise adjustment is performed, and the noise covariance of the sampling period t is set to be equal to the noise covariance of the sampling period t-1, marked as frozen covariance. If the value is 1, the noise update amount is adjusted to obtain the measured noise covariance, and the formula is:

[0108] ;

[0109] in, To measure the noise covariance, The final noise update increment;

[0110] By clipping the upper and lower bounds of the frozen covariance and the measurement noise covariance, the final noise covariance is obtained, as shown in the formula:

[0111] ;

[0112] ;

[0113] in, For the final noise covariance, and For minimum noise constraints and maximum noise constraints, To freeze the covariance and measurement noise covariance, The noise covariance is clipped to the upper bound.

[0114] By using logical judgments based on the indicators within the window, the dynamic determination of the covariance update rhythm is achieved, enhancing the stability of the overall system during the disturbance recovery period. The pruning mechanism not only prevents the covariance from exceeding the safety limit but also ensures that the system maintains sufficient control robustness in the early stages of the disturbance response. Freezing the covariance setting provides a stable buffer after severe disturbances, preventing the noise covariance from being amplified by multiple disturbances and causing system oscillations. Through the pruning mechanism of upper and lower limits, the noise covariance is ultimately controlled within the engineering acceptable range, providing a reliable prior for subsequent filtering and prediction operations.

[0115] The next step involves using the predicted state vector and predicted covariance generated by the UKF method, including:

[0116] Read the state estimation vector (e.g., loading rate state component and loading entry rate component) and covariance of the vehicle controller at sampling period t-1;

[0117] The unscented Kalman filter (UKF) method is used to generate the predicted state vector and predicted covariance for the sampling period t-1.

[0118] The predicted state vector is mapped to the measurement space through a measurement function to obtain the predicted measurement mean, as shown in the formula:

[0119] ;

[0120] ;

[0121] in, To predict the measurement mean, For measurement function, This is the predicted state vector.

[0122] By using the unscented Kalman filter (UKF) method to update the state estimation vector, the accuracy of state prediction under nonlinear systems can be significantly improved. Compared with traditional linear filters, UKF can more accurately fit the prediction error and system dynamics in highly nonlinear scenarios. It is particularly suitable for the high volatility characteristics of loading rate and entry rate state variables in vehicle dynamic scenarios. The predicted measurement mean generated after mapping the state vector to the measurement space provides a stable benchmark for the construction of subsequent innovative quantities, reduces the uncertainty of the error transformation path, and ensures the sustainable closed-loop update of system state estimation.

[0123] Finally, the innovation quantity and innovation covariance are obtained, and the mean and standard deviation of the loading rate are calculated, including:

[0124] The innovation quantity is obtained by calculating the difference between the observed value and the predicted mean of the fusion loading rate, and the innovation covariance is obtained by calculating the sum of the predicted covariance and the final noise covariance.

[0125] Based on the innovation covariance and innovation quantity, the mean loading rate (extracted from the updated state estimate) and the standard deviation of the loading rate (obtained by extracting the variance corresponding to the loading rate from the update uncertainty and taking the square root) are calculated using the following formulas:

[0126] ;

[0127] ;

[0128] ;

[0129] in, For Kalman gain, The state-measurement cross-covariance is calculated from the predicted state vector and the predicted measurement mean using the formula for the cross-covariance between the predicted state and the predicted measurement through unscented Kalman filtering. To innovate covariance, To update the state estimate, For innovation quantity, To update the uncertainty, To predict covariance.

[0130] The calculation of innovation quantity and innovation covariance constitutes the core path of observation update in the filter. The difference between the observation and prediction results is quantified into innovation quantity. Combined with the innovation covariance formed by prediction uncertainty and noise covariance, the state update reflects the actual disturbance impact while maintaining tolerance to system noise. The linkage extraction of the mean and standard deviation of the loading rate provides probabilistic boundary information for the state output, enabling the upper-level control decision to formulate adjustment strategies based on confidence intervals, thus better meeting the dual goals of safety control and energy saving in practical applications.

[0131] S3, Probability Correction Module, is used to calculate the dimensionless threshold distance based on the mean and standard deviation of the loading rate, construct the intercept correction term based on the innovation quantity, obtain skewness and kurtosis through the quantile regression equation, correct the dimensionless threshold distance, and input the standard normal CDF to obtain the probability error;

[0132] Specifically, skewness and kurtosis are obtained through the quantile regression equation, the dimensionless threshold distance is corrected, and the result is input into the standard normal CDF to obtain the probability error, including:

[0133] Based on historical experimental experience, an overflow boundary threshold is set. The difference between the overflow boundary threshold and the mean loading rate is divided by the standard deviation of the loading rate to obtain the dimensionless threshold distance.

[0134] The innovation quantity is standardized using the innovation covariance to obtain the standardized innovation quantity, as shown in the formula:

[0135] ;

[0136] in, To standardize innovation volume, For innovation quantity, For sampling period index, For innovative covariance;

[0137] The intercept correction term is obtained by correcting the standardized innovation quantity using the following formula:

[0138] ;

[0139] in, For the intercept correction term, The in-window offset index ranges from 0 to... Decreasing generation;

[0140] Based on the intercept correction term, empirical quantiles are obtained through the quantile regression equation, and skewness and kurtosis are constructed using the following formulas:

[0141] ;

[0142] ;

[0143] ;

[0144] ;

[0145] ;

[0146] in, For empirical quantiles, For the first quantile levels (defined based on statistical analysis). The intercept is estimated using OLS and calculated by OLS. For designing vectors, For transpose, To revert to residuals, The non-intercept coefficient vector For skewness, For kurtosis, For the first The normal distribution with quantile levels is the standard normal cumulative distribution function. , as well as Based on the design vector, the first, second, and third coefficient components are calculated using OLS, corresponding to... The three items in It is an integer. This is the multiplier factor. The coefficients of the coefficient components;

[0147] Extract the coefficient components and calculate the constraint test metric. If the constraint test metric is greater than or equal to 0, the skewness and kurtosis are accepted; otherwise, constraints are applied to the coefficient components, and the skewness and kurtosis are recalculated until the constraint test metric is satisfied. The final skewness and kurtosis are obtained using the following formula:

[0148] ;

[0149] ;

[0150] in, and The coefficients of the coefficient components. As a constraint test metric, it is used to determine whether skewness and kurtosis are acceptable;

[0151] The dimensionless threshold distance is corrected using the final skewness and kurtosis to obtain the corrected threshold distance, as shown in the formula:

[0152] ;

[0153] in, To correct the threshold distance, The dimensionless threshold distance. and For the final skewness and kurtosis;

[0154] Input the corrected threshold distance into the standard normal CDF to obtain the corrected tail probability. Calculate the difference between the corrected tail probability and the target overflow probability (set based on empirical rules) to obtain the probability error.

[0155] The introduction of dynamic corrections for skewness and kurtosis provides asymmetric information modeling capabilities for error boundary prediction. By standardizing the innovation quantity and inputting it into the quantile regression model, the non-Gaussian deviation characteristics of the system response can be estimated. The intercept correction term enables historical data to adaptively adjust to the current prediction. The skewness and kurtosis constructed at the end not only reflect the current disturbance intensity but also provide dynamic curve correction capabilities for subsequent probabilistic error modeling. The tail probability obtained by inputting the correction threshold distance into the standard normal cumulative distribution function reflects the accurate probabilistic expression of spillover risk, providing quantitative input for objective constraint optimization.

[0156] S4, Strategy Optimization Module, is used to use probability error as sliding surface variable, calculate speed scaling factor and speed variable, construct objective function and constraints, use DSQP to optimize and solve objective function, obtain execution speed and steering angle commands, and send them to vehicle controller for execution through API interface;

[0157] Specifically, DSQP is used to optimize and solve the objective function, resulting in commands for vehicle speed and steering angle. These commands are then sent to the vehicle controller for execution via an API interface, including:

[0158] By setting the probability error as a variable of the sliding surface, calculating the velocity scaling factor, obtaining the velocity variable, and constructing the objective function, the formula is as follows:

[0159] ;

[0160] ;

[0161] ;

[0162] in, For velocity variables, and The minimum and maximum safe speeds were determined based on low-speed stable operation tests. The nominal operating speed is based on the operating mode parameter settings. The objective function value, The candidate execution speed (to be solved). The penalty coefficient for speed change, The change in speed is obtained by subtracting the vehicle speed from the candidate execution value. This is the steering penalty coefficient. Candidate steering angle command (to be solved). This is the speed scaling factor. The mapping gain (obtained based on experimental calibration) is given. For sliding surface variables;

[0163] Constraints are set based on lateral acceleration and velocity variables, including lateral acceleration less than the maximum permissible value of lateral acceleration (obtained from vehicle stability tests), (acceleration variables calculated from vehicle speed using the difference method), absolute value of acceleration variables less than or equal to the maximum permissible value of acceleration variables (obtained from comfort calibration tests), absolute value of candidate steering angle commands less than or equal to the maximum permissible steering angle (obtained from steering gear structural limits), and absolute value of velocity variables less than or equal to the maximum velocity change (based on drive system response capability settings).

[0164] Based on the constraints, the objective function is optimized and solved using DSQP to obtain the commands for vehicle speed and steering angle, which are then issued and executed via the API interface.

[0165] The objective function construction and DSQP optimization solution mechanism provide real-time and feasibility guarantees for the generation of system control variables. By mapping the probability error to a sliding surface variable and introducing a speed scaling factor to form a speed variable, the logical closed-loop expression of the control objective is realized. The objective function combines the speed change and steering angle penalty factor to make the optimization results more physically feasible and driving smooth. The constraints set by the vehicle dynamic parameters ensure that the output command achieves the optimal trade-off between structural safety and ride comfort. Under the premise of satisfying the constraints, the DSQP method efficiently searches for the optimal execution strategy and finally generates control commands that can be directly executed through the API interface, realizing the engineering implementation of adaptive adjustment of the closed-loop path.

[0166] In summary, this invention constructs an objective function and constraints by using probability error as a sliding surface variable, calculating the speed scaling factor and speed variable, and using DSQP to optimize and solve the objective function to obtain the execution speed and steering angle commands, which are then sent to the on-board controller for execution via an API interface. This achieves highly sensitive identification of the garbage overflow state, improves the stability of loading state estimation, and realizes coordinated control of safety and efficiency.

[0167] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A sweeper truck driving strategy adaptive adjustment system in response to garbage overflow, characterized in that: include, The disturbance sensing module is used to collect data from the vehicle controller and calibration parameters of the vehicle controller, obtain the original disturbance observations and the fused loading rate observations, construct an autocorrelation estimation sequence for the original disturbance observations, and calculate the spectral intensity index based on the autocorrelation estimation sequence. The noise conditioning module is used to calculate residual energy based on the fused loading rate observations, obtain the noise covariance through the vehicle controller, construct the final noise update increment by combining the spectral intensity index, make update judgments based on the sampling period, update the final noise update increment and perform upper and lower bound clipping to obtain the final noise covariance, combine the predicted state vector generated by the UKF method and the predicted covariance to obtain the innovation quantity and innovation covariance, and calculate the loading rate mean and loading rate standard deviation. The probability correction module is used to calculate the dimensionless threshold distance based on the mean and standard deviation of the load rate, construct the intercept correction term based on the innovation quantity, obtain skewness and kurtosis through the quantile regression equation, correct the dimensionless threshold distance, and input the standard normal CDF to obtain the probability error. The strategy optimization module is used to calculate the speed scaling factor and speed variable by using the probability error as the sliding surface variable, construct the objective function and constraints, use DSQP to optimize and solve the objective function, obtain the vehicle speed and steering angle commands, and send them to the vehicle controller for execution through the API interface.

2. The adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in claim 1, characterized in that: The process of calculating residual energy based on fused loading rate observations and obtaining noise covariance through the on-board controller includes: The difference between the fused loading rate observations in sampling period t-1 and sampling period t is calculated to obtain the innovation residual; Based on statistical analysis, the productivity parameter is set, the relaxation sampling period constant is calculated, the product of the relaxation sampling period constant and the sampling frequency is calculated, and the result is rounded up to obtain the number of window length periods. Read the noise covariance of the vehicle controller sampling period t-1, calculate the product with the scaling factor, and obtain the noise covariance of the sampling period t. Calculate the ratio of residual energy to noise covariance of sampling period t, obtain the standard deviation and squared residuals, calculate the consistency loss, and calculate the allowable update coefficient; Read the calibrated maximum spectral intensity of the vehicle controller, calculate the ratio of the spectral intensity index to the maximum spectral intensity, obtain the normalized spectral intensity, and calculate the membership degree.

3. The adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in claim 2, characterized in that: The construction of the final noise update increment includes: Based on membership, the basic noise increment is calculated, the variable point sampling period and relaxation window length of the vehicle controller are read, and the attenuation factor is generated through exponential decay. The final noise update increment is obtained by multiplying the attenuation factor, the allowable update coefficient, and the basic noise increment.

4. The adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in claim 3, characterized in that: The step of updating the final noise update increment and pruning its upper and lower bounds to obtain the final noise covariance includes: If the sampling period t is less than or equal to the sum of the variable point sampling period and the relaxation window length, then the window flag is set to 1; otherwise, it is set to 0. If it is 0, no noise adjustment is performed. The noise covariance of the sampling period t is set to be equal to the noise covariance of the sampling period t-1, and it is marked as frozen covariance. If it is 1, the noise update amount is adjusted to obtain the measured noise covariance. The frozen covariance and the measurement noise covariance are clipped by upper and lower bounds to obtain the final noise covariance.

5. The adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in claim 4, characterized in that: The predicted state vector and predicted covariance generated using the UKF method include: Read the state estimation vector and covariance of the vehicle controller during the sampling period t-1; The unscented Kalman filter (UKF) method is used to generate the predicted state vector and predicted covariance for the sampling period t-1. The predicted state vector is mapped to the measurement space through a measurement function to obtain the predicted measurement mean.

6. The adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in claim 5, characterized in that: The process of obtaining the innovation quantity and innovation covariance, and calculating the mean and standard deviation of the loading rate, includes: The innovation quantity is obtained by calculating the difference between the observed value and the predicted mean of the fusion loading rate, and the innovation covariance is obtained by calculating the sum of the predicted covariance and the final noise covariance. Based on the innovation covariance and innovation quantity, the mean and standard deviation of the loading rate are calculated.

7. The adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in claim 6, characterized in that: The skewness and kurtosis are obtained through the quantile regression equation, the dimensionless threshold distance is corrected, and the standard normal CDF is input to obtain the probability error, including: Based on historical experimental experience, an overflow boundary threshold is set. The difference between the overflow boundary threshold and the mean loading rate is divided by the standard deviation of the loading rate to obtain the dimensionless threshold distance. The innovation quantity is standardized using innovation covariance to obtain the standardized innovation quantity. The intercept of the standardized innovation quantity is corrected to obtain the intercept correction term. The empirical quantiles are obtained through the quantile regression equation. Skewness and kurtosis are constructed. The dimensionless threshold distance is corrected to obtain the corrected threshold distance. The corrected threshold distance is input into the standard normal CDF to obtain the corrected tail probability. The difference between the corrected tail probability and the target overflow probability is calculated to obtain the probability error.

8. The adaptive adjustment system for the sweeper truck driving strategy in response to a garbage overflow state as described in claim 7, characterized in that: The process of using DSQP to optimize the objective function, obtaining the commands for vehicle speed and steering angle, and then sending them to the vehicle controller for execution via an API interface includes: The probability error is set as a sliding surface variable, the speed scaling factor is calculated to obtain the speed variable, the objective function is constructed and the constraints are set, the objective function is optimized and solved using DSQP to obtain the vehicle speed and steering angle commands, which are then issued and executed through the API interface.

9. The adaptive adjustment system for the driving strategy of a road sweeper in response to a garbage overflow state as described in claim 8, characterized in that: The process of constructing an autocorrelation estimation sequence from the original perturbation observations and calculating the spectral intensity index based on the autocorrelation estimation sequence includes: Set a fixed window length of N, collect the original perturbation observations within the fixed window, obtain the mean-removed perturbation window sequence, calculate the autocorrelation estimation sequence, and calculate the power spectral density function value; Within the vehicle controller, a set of frequency bands is set based on engineering calibration. The power spectral density function values ​​are summed within the set of frequency bands to obtain the spectral intensity index.

10. The adaptive adjustment system for the driving strategy of a sweeper truck in response to a garbage overflow state as described in claim 9, characterized in that: The process of collecting on-board controller data and on-board controller calibration parameters to obtain the original disturbance observations and fused load rate observations includes: By obtaining vehicle controller data and calibration parameters through the API interface, the loading rate variable and the height channel loading rate are calculated and weighted summed to obtain the fused loading rate observation value.

Images