Search

CN-121979225-A - Autonomous mimicry control and disturbance rejection method for intelligent shuttle vehicle of refrigeration house

CN121979225ACN 121979225 ACN121979225 ACN 121979225ACN-121979225-A

Abstract

The invention discloses an autonomous mimicry control and disturbance rejection method of an intelligent shuttle vehicle for a refrigeration house. The method comprises the steps of constructing an environment space-time gradient field through a distributed optical fiber sensing network, inverting heat flow disturbance to update a dynamic environment model, synchronously analyzing driving current harmonic waves to identify a ground phase change state, compensating a hysteresis effect of a mechanism by utilizing a digital twin model, adopting a reflection-adaptation-learning three-layer decision-making architecture, adjusting motion parameters in a millisecond level of a reflection layer, learning a minute level planning mimicry navigation path in a layer, and finally predicting, controlling and fusing instructions of each layer through the model to output a cooperative control signal. The invention enables the shuttle to actively sense, forecast and adapt to the non-uniform dynamic environment of the refrigeration house, and improves the operation safety, positioning precision and operation efficiency under the working conditions of low temperature, wet and slippery and multiple disturbance.

Inventors

  • ZHOU TAO
  • WANG JIAXIN

Assignees

  • 江苏欧标智能储存科技有限公司

Dates

Publication Date
20260505
Application Date
20260408

Claims (9)

  1. 1. An autonomous mimicry control and disturbance rejection method of an intelligent shuttle vehicle for a refrigeration house is characterized by comprising the following steps: Acquiring microscale physical quantity data of a vehicle body-environment interface, and constructing an environment space-time gradient field through non-structural grid gradient calculation and space-time interpolation; Based on the environmental space-time gradient field, solving the inverse problem of heat conduction and fusing the inverse problem with multi-source data, and dynamically updating an environmental dynamic embedded model containing risk characteristic prediction; acquiring primary state data of a driving system, and identifying the phase change state of a wheel-ground contact interface in real time through motor current harmonic analysis and multi-modal feature fusion; updating a hysteresis nonlinear digital twin model on line based on historical track deviation data of a motion mechanism, and generating a feedforward compensation instruction by using the model; and fusing the environment dynamic embedded model, the phase change state and the feedforward compensation instruction, generating a cooperative control instruction through a reflection-adaptation-learning three-layer decision framework, and driving the shuttle to execute mimicry navigation and disturbance rejection operation.
  2. 2. The method of claim 1, wherein constructing an ambient spatiotemporal gradient field comprises: Demodulating and calculating physical quantity of original wavelength data of each node of the distributed fiber bragg grating sensing network to generate a multi-node physical quantity instantaneous snapshot data table; performing space-time alignment and wild point elimination on the data in the physical quantity instantaneous snapshot table to generate clean physical quantity field grid data; based on the unstructured triangulated grid and the linear shape function theory, calculating a temperature gradient vector and a strain gradient tensor at the center of each grid cell, and generating unstructured grid cell gradient vector data; And performing anisotropic diffusion filtering and node Zhang Liangchong construction on the unit gradient vector data, and generating high-resolution space-time gradient field map spectrum data covering the vehicle body and the near field space through Kerling space interpolation.
  3. 3. The method of claim 1, wherein dynamically updating the environmental dynamic embedded model including risk feature predictions comprises: the method comprises the steps of constructing a heat flow inversion optimization problem by taking a temperature field in a high-resolution space-time gradient field map as a boundary condition and combining a parameterized vehicle body thermal model, solving by adopting an accompanying method, and inverting out vehicle body surface heat flow density vector distribution data; Fusing the heat flux density vector distribution data with the strain gradient data, identifying and marking a high heat flux mutation area, a heat-force coupling risk area and a potential phase change interface through a preset rule engine, and generating environment high risk feature list data; And updating the environment dynamic embedded model by adopting a data assimilation algorithm according to the currently observed high risk characteristic list data and the historical gradient field sequence.
  4. 4. The method of claim 1, wherein identifying in real time the phase change state of the wheel-ground contact interface comprises: Collecting three-phase high-frequency current waveforms, rotor positions and vehicle body vibration data of each driving motor to form a driving system primary state data stream; Carrying out short-time Fourier transform on the current waveform data, extracting the amplitude and phase characteristics of fractional harmonic waves and sideband frequency components which are in non-integer multiple relation with the fundamental frequency, and forming current nonlinear harmonic characteristic spectrum data; Carrying out wavelet packet decomposition on the vehicle body vibration data to obtain vibration characteristic data; And fusing harmonic characteristic spectrum data, vibration characteristic data, local ground temperature data and dynamic load data of the wheels into characteristic vectors, inputting the characteristic vectors into an online integrated learning classifier for real-time reasoning, and outputting interface phase probability distribution vectors through probability calibration.
  5. 5. The method of claim 1, wherein generating a feed forward compensation command comprises: Recording actual track feedback after the motion mechanism executes the theoretical instruction track, and calculating to obtain a high-fidelity track tracking error profile data set; Inputting the error profile data set, the corresponding theoretical instruction sequence and the corresponding environmental temperature data into a hysteresis nonlinear digital twin model, and updating model parameters by adopting an online parameter identification algorithm to obtain an updated hysteresis model parameter vector; Inputting new motion instruction data to be executed into the updated digital twin model for forward simulation, predicting the execution deviation of the digital twin model, and calculating and generating a feedforward compensation moment/current instruction sequence through a mechanism inverse dynamics model; and performing low-pass filtering and safe limiting on the feedforward compensation moment/current command sequence, and then overlapping the feedforward compensation moment/current command sequence with new motion command data in a command domain to generate final execution command data after predistortion optimization.
  6. 6. The method of claim 1, wherein generating the cooperative control instruction comprises: Continuously monitoring the unsteady transition signals in the highest-level alarm and interface phase state probability distribution vectors in the environment high-risk feature list data, and directly matching and outputting the energy level emergency deviation rectifying instruction data through a hardware logic rule base; Based on the short-term prediction of the updated dynamic environment model data and the continuous interface phase state probability distribution vector, dynamically calculating and outputting dynamic chassis parameter adjustment data for adjusting torque distribution, steering sensitivity and suspension parameters; Constructing a multi-level environment cost map based on the updated dynamic environment model data, searching a path with the lowest comprehensive cost through a space-time three-dimensional algorithm, optimizing a speed curve, and generating global mimicry navigation stream data; and according to the emergency deviation rectifying instruction data, the dynamic chassis parameter adjusting data, the final execution instruction data and the global mimicry navigation stream data, carrying out conflict arbitration and multi-objective optimization through a model prediction controller, solving and outputting a final cooperative control instruction data packet.
  7. 7. The method of claim 6, wherein generating global mimicry navigation stream data comprises: The gradient intensity, the risk area distribution, the static obstacle and the task semantic information in the updated dynamic environment model data are respectively mapped into multi-level cost maps such as a thermal disturbance cost layer, a basic passing layer and the like, and a multi-level fusion environment cost map is generated through weighted fusion; On the fusion environment cost map, a space-time three-dimensional search algorithm is operated to find a primary space-time path sequence with optimal accumulated cost in space and time dimensions from a starting point to an ending point; And (3) performing spline curve smoothing on the initially selected space-time path sequence, planning a smoothed speed-time curve based on vehicle dynamics constraint and path curvature, and packaging to form final global mimicry navigation flow data.
  8. 8. A method according to claim 3, wherein solving the heat flow inversion optimization problem using an accompanying method comprises: Constructing an inversion problem framework which takes the distribution of the heat flux density on the surface of the vehicle body as an optimized variable, takes the minimum difference between the predicted temperature and the observed temperature of a forward thermal model as an objective function and comprises Tikhonov regularization terms; starting iteration from an initial heat flow hypothesis, in each iteration, firstly carrying out forward heat calculation once to obtain a predicted temperature field, and then running accompanying calculation once to efficiently obtain gradients of an objective function on all heat flow parameters; updating the heat flow distribution by using the obtained gradient and adopting an L-BFGS optimization algorithm; And determining an optimal regularization parameter based on an L-curve method, taking a solution under the parameter as final heat flow density distribution data, and simultaneously calculating the variance of a principal component of the solution to carry out uncertainty quantification.
  9. 9. The method of claim 5, wherein updating model parameters using an online parameter identification algorithm comprises: constructing parameter updating of the hysteresis nonlinear digital twin model as a recursive least square optimization problem; The method aims at minimizing the difference between a model predicted track and an actual track reversely deduced by superimposing error profile data by a theoretical instruction sequence; After each motion is completed, the model parameter vector and covariance matrix thereof are recursively updated by using the newly acquired error data, so that the hysteresis characteristics of the model dynamic tracking mechanism are changed due to low temperature and abrasion.

Description

Autonomous mimicry control and disturbance rejection method for intelligent shuttle vehicle of refrigeration house Technical Field The invention belongs to the technical field of automatic warehouse logistics, and particularly relates to an autonomous mimicry control and disturbance rejection method of an intelligent shuttle vehicle for a refrigeration house. Background The refrigeration house is used as a core node of modern cold chain logistics, and the internal automation level of the refrigeration house is directly related to the storage efficiency and the operation cost. Shuttle systems have become a mainstream solution for automated freezer due to their high density storage and flexible scheduling capabilities. However, the extreme and dynamically changing environment of the refrigerator constitutes a serious challenge for the stable operation of the shuttle, and the prior art solutions still have significant bottlenecks in terms of environmental adaptability, operation safety and operation accuracy. First, the freezer environment is not uniformly steady. The uneven air supply of the refrigeration equipment, the heat exchange caused by the opening of the warehouse door and the local heat source generated by the storage of goods form a complex three-dimensional temperature and airflow gradient field together. The existing shuttle generally relies on a limited point type temperature and humidity sensor to perform environment sensing, and the discretization measurement cannot capture continuous gradient distribution and microscopic disturbance, so that a control system regards the environment as a static background or can only respond to large-amplitude and low-frequency macroscopic changes, and the sensing and coping capability of progressive influences such as nonlinear deformation of materials, sensor drift and the like caused by gradients is lacking. Second, the low temperature and high humidity environment causes a series of unique physical interaction problems. The ground thin ice layer may generate solid-quasi-liquid phase change under the coupling of the wheel pressure and the temperature, so that the adhesion coefficient is dynamically suddenly changed, the traditional anti-slip control based on wheel speed difference or vehicle body posture feedback has hysteresis, and the occurrence of slip is difficult to prevent. Meanwhile, the lubricant of the mechanical part is sticky and the elastic modulus of the polymer material is changed due to low temperature, so that a motion mechanism (such as fork lifting and steering) generates hysteresis nonlinearity related to temperature and motion history, repeated positioning accuracy is seriously influenced, and the existing calibration method is mostly off-line static compensation and cannot adapt to dynamic characteristic drift in operation. Furthermore, the control architecture of the existing shuttle vehicle mostly follows a linear paradigm of 'perception-planning-execution', and the processing of the environmental disturbance by each module is relatively isolated. For example, the navigation module plans a geometrically shortest path, but may traverse a region of strong airflow disturbance, and the stability control module passively intervenes after slip is detected, but cannot proactively adjust torque distribution. The responsive control strategy lacks the capability of overall cognition and collaborative optimization of an environment-vehicle coupling system, and is stiff and inefficient in dealing with sudden and variable freezer disturbances. In summary, the current technology of the refrigerator shuttle cannot sufficiently look at the fundamental influence of space-time heterogeneity of the environment, lacks fine measurement on microscopic dynamics such as gradient fields, phase change interfaces and the like at a perception level, and lacks autonomous adaptation and collaborative decision-making capability of fusing multi-source prospective information at a control level. Therefore, development of a novel intelligent control system capable of deeply understanding and actively adapting to the dynamic environment of a complex refrigeration house is needed to realize high-reliability and high-precision autonomous operation of the shuttle under extreme working conditions. Disclosure of Invention In view of the problems in the prior art, the application provides an autonomous mimicry control and disturbance rejection method for an intelligent shuttle vehicle for a refrigeration house. The technical scheme is that the autonomous mimicry control and disturbance rejection method of the intelligent shuttle of the refrigeration house comprises the following steps: Acquiring microscale physical quantity data of a vehicle body-environment interface, and constructing an environment space-time gradient field through non-structural grid gradient calculation and space-time interpolation; Based on the environmental space-time gradient field, solving the inverse