Complex terrain crawler excavator operation action decoupling cooperative optimization control method and system

Abstract

The application relates to the technical field of industrial control systems, and discloses a complex-terrain crawler excavator operation action decoupling and cooperative optimization control method and system. The hydraulic network topological impedance model is established to convert the hydraulic system flow distribution into a calculable optimization space, the time delay matching decoupling is realized by using the phase alignment of the pressure wave front propagation time delay, the contact impact is avoided by automatically switching between the open chain and the closed chain mode, the working condition transition jitter is eliminated by means of the continuous probability density interpolation, the centralized algorithm bottleneck is broken through by relying on the distributed rolling optimization, the long-term self-adaptation of the control strategy is realized by using the mapping weight coefficient closed-loop evolution driven by the kernel function smoothing method, and the flow distribution is changed from passive lag to active prediction, the coupled interference is changed from unquantifiable to directional cancellation, the control architecture is changed from centralized global solution to distributed local cooperation, the working condition adaptability is changed from discrete table lookup jump to continuous smooth transition, and the system performance is changed from fixed preset parameters to continuous evolution along with the operation process.

Description

Technical Field

[0001] This application relates to the field of industrial control system technology, and in particular to a decoupling and collaborative optimization control method and system for the operation of a tracked excavator in complex terrain. Background Technology

[0002] Hydraulic excavators, as the most widely used earthmoving machinery in the construction machinery field, rely heavily on intelligent and precise control for their operational efficiency and construction quality. The core working device of this equipment consists of three actuators: the boom, the stick, and the bucket. Driven by a hydraulic system, they achieve complex actions such as digging, lifting, and slewing. In actual operation, the boom is responsible for vertical lifting and coarse positioning, while the stick is responsible for horizontal extension and precise digging. Both need to achieve smooth, combined movements in complex terrain. However, these actuators share the same hydraulic pump source, leading to severe flow competition when flow demands reach their peak simultaneously. Furthermore, boom extension alters the effective lever arm of the stick, and stick retraction affects the load distribution on the boom, resulting in strong coupling kinematic and dynamic interference between the multiple actuators. This makes it difficult for traditional independent control strategies to guarantee the trajectory accuracy and coordination of the combined movements.

[0003] To address the aforementioned issues, existing technologies often employ distributed drive or load-sensing pressure compensation schemes. CN112648244A discloses a pump-valve coordinated multi-actuator electro-hydraulic control system, which uses a distributed pump control unit to drive each actuator separately and adjusts the swashplate angle of the main pump to achieve flow matching. However, this system requires each actuator to be configured with an independent power source based on its peak power, resulting in a significant increase in total installed power, weight, and volume. Furthermore, additional compensation is needed for the flow difference between the two chambers of asymmetric hydraulic cylinders, leading to high system complexity. ZL201810342465.3 discloses a distributed direct-drive excavator hydraulic system with a servo motor-driven single-variable pump. This system uses single-pump, single-actuator displacement control to largely eliminate throttling losses. However, the independent drive results in high installed power, a non-compact structure, and does not solve the problems of coupling interference between multiple actuators and adaptive operation under complex working conditions. The above solutions still rely on fixed priority or passive pressure compensation, and do not establish a complete network topology between the pump source, valve port, and actuator chamber, making it impossible to explicitly quantify the transmission path of coupling interference. At the same time, they do not consider the physical time delay caused by the limited propagation speed of pressure waves in hydraulic oil, nor do they identify the kinematic closed chain formed by the contact of the bucket teeth tip with the ground. Furthermore, the storage of operating parameters mostly adopts the discrete lookup table method, and parameter jumps at the boundary cause actuator vibration. Summary of the Invention

[0004] This application proposes a decoupling and collaborative optimization control method and system for tracked excavator operation in complex terrain. It has the advantages of adaptive operation in complex terrain, directional cancellation of coupling interference of multiple actuators, and continuous evolution of control strategy. It is used to solve the problems of passive response of flow distribution, unquantifiable coupling interference, failure of parameters in complex working conditions, and centralized computing power bottleneck in the prior art.

[0005] To achieve the above objectives, this application adopts the following technical solution: a decoupling and collaborative optimization control method for the operation of a tracked excavator in complex terrain, the specific steps of which are as follows: S1. Multi-source sensing field construction: Sensors are set up on the chassis and working device to collect slope, cylinder pressure and displacement, main pump flow and oil temperature. An extended state observer is constructed to identify oil temperature drift, component wear and pipeline elastic deformation nonlinear disturbances. A hydraulic network topology impedance model is established to map the pump source, valve port and chamber as node channels. The channel impedance and flow distribution are initialized to form the initial distribution of the working condition-control parameter mapping weight coefficient. S2. Active traffic arbitration: Based on the data from step S1, the expected traffic demand is determined according to the actuator speed and load. The weighted interpolation of the mapping weight coefficients of adjacent operating condition clusters is used to determine the continuous weights, and the optimal traffic allocation path is reconstructed. S3. Decoupling and Cooperative Optimization: According to the instructions in step S2, establish the coupling relationship of the three actuators, map the disturbance to the topological model disturbance branch, determine the pressure wave propagation delay based on the channel length, oil elastic modulus and oil temperature, and superimpose the feedforward compensation after phase alignment; allocate to four distributed sub-controllers to independently execute single-step local optimization, and use rolling time domain optimization to ensure that the calculation cycle matches the hydraulic system control cycle; S4. Closed-loop evolution of mapping weight coefficients: Record the operation cycle flow deviation, trajectory tracking error and pressure fluctuation, construct a multi-dimensional comprehensive performance evaluation function, and trigger the update of mapping weight coefficients when the comprehensive performance score is lower than the preset proportion of the historical best score, evolve the probability density of the parameter space, and write the updated parameters into storage to form a knowledge base.

[0006] Furthermore, in step S1, the specific steps for multi-source sensing field establishment are as follows: S1.1 Multi-source sensing topology model construction: Sensors are deployed on the chassis and working device to collect data on slope, support roller reaction force, cylinder pressure and displacement, pump flow and oil temperature. An extended state observer is constructed to identify oil temperature drift, component wear and nonlinear disturbances of pipeline elastic deformation online, output terrain feature parameters and load values, establish a hydraulic network topology impedance model, and map the pump source outlet, valve port and chamber as node channels. The nodes include source nodes, intermediate nodes and terminal nodes. The channel impedance includes liquid resistance, liquid capacity and liquid sensing. The channel impedance and flow distribution are initialized. S1.2 Initialization of Working Condition Mapping Weight Coefficients: Historical flow distribution is associated with terrain feature parameters to form an initial distribution of mapping weight coefficients. A closed-chain detection layer is added to the topological impedance model. The contact state is inferred based on the bucket attitude angle and pressure changes. It is determined whether the boom, stick, and bucket form a kinematic closed chain. The closed-chain or open-chain mode is marked. The terrain load parameter space is constructed with slope, ground stiffness, and subsidence as coordinate axes, so that the mapping weight coefficients are stored according to the probability density distribution in the space.

[0007] Furthermore, in step S2, the specific steps of the traffic active arbitration are as follows: S2.1 Mode Switching and Flow Demand Prediction: Receive the S1 topological impedance model status, mapping weight coefficient distribution and mode label. In open chain mode, establish a flow demand prediction model based on the estimated speed, acceleration and load of the boom, stick, bucket, and slewing mechanism, and calculate the expected flow demand of each actuator. In closed chain mode, introduce a fixed constraint on the tooth tip position, and jointly solve the flow demand of the three actuators into the minimum energy consumption coordination control quantity that satisfies the closed chain constraint. S2.2 Mapping Weight Coefficient Interpolation and Path Reconstruction: Based on the coordinates of the current working point in the terrain load parameter space, the mapping weight coefficient distribution of the neighboring working clusters is weighted and interpolated to generate continuous flow allocation weights. This gives priority to paths with high mapping weight coefficient values ​​and those that match the terrain load. The optimal flow allocation path is reconstructed in the topology impedance model, and active flow allocation instructions for each actuator are generated.

[0008] Furthermore, in step S2.1, the specific steps for mode switching and traffic demand prediction are as follows: S2.11 Open Chain Flow Demand Prediction: Receive the S1 topology impedance model status, mapping weight coefficient distribution and mode label. When the topology impedance model is in open chain mode, based on the estimated values ​​of the speed, acceleration and load of the boom, stick, bucket and slewing mechanism, combined with terrain feature parameters, establish a multi-actuator joint flow demand prediction model, calculate the expected flow demand of each actuator in the next control cycle, and form the independent flow allocation benchmark value of each actuator in open chain mode. S2.12, Closed-chain constraint coordination solution: When the topological impedance model is in closed-chain mode, the fixed constraint of the tooth tip position is introduced based on the closed-chain detection layer mark, and the kinematic closed-chain constraint equations of the boom, stick and bucket three actuators are established. The flow demand of the three actuators is incorporated into the joint optimization framework, and the minimum energy consumption coordination control quantity that satisfies the closed-chain constraint is solved to form the benchmark value of the coupled flow distribution of each actuator in closed-chain mode.

[0009] Furthermore, in step S2.2, the specific steps for mapping weight coefficient interpolation and path reconstruction are as follows: S2.21, Weighted Interpolation of Mapping Weight Coefficients and Weight Generation: Receives the traffic allocation baseline value output from S2.1, and performs weighted interpolation on the distribution of mapping weight coefficients of neighboring working condition clusters based on the coordinates of the current working point in the terrain load joint parameter space. This generates continuously changing traffic allocation weights, giving priority to paths with high mapping weight coefficient values ​​and matching terrain load, and determining the set of candidate paths for traffic allocation of each actuator, providing a screening basis for path reconstruction. S2.22 Optimal Path Reconstruction and Command Generation: In the topological impedance model, the optimal flow distribution path is screened and reconstructed from the candidate path set according to the selection probability. The path is optimized by combining the impedance of each channel. The reconstruction result is converted into active flow distribution commands for the boom, stick, bucket and slewing mechanism and output to step S3.

[0010] Furthermore, in step S3, the specific steps of decoupling and collaborative optimization are as follows: S3.1 Coupled Disturbance Mapping and Wavefront Compensation: Receive the S2 active flow distribution command, establish a kinematic and dynamic coupling model of the three actuators (boom, stick, and bucket), identify lever arm changes, load offsets, and attitude feedback disturbances in real time, map the coupled disturbances into topological model disturbance branches, calculate the pressure wavefront propagation delay based on channel length, oil elastic modulus, and oil temperature, and then align the feedforward compensation amount with the phase of the delay and superimpose it onto the control command. S3.2 Distributed Cooperative Optimization and Feedback Correction: The topology model reconstruction and compensation are decomposed into four distributed sub-controllers. Each sub-controller independently performs single-step local optimization based on local sensor information. The calculation cycle of the single-step local optimization is no greater than the hydraulic system control cycle. The topology model state and compensation parameters are shared through a high-speed bus. Rolling time-domain optimization is used to ensure that the calculation cycle matches the hydraulic system control cycle. A feedback correction link is introduced to correct the feedforward compensation amount in real time, generating precise control commands for each actuator after decoupling.

[0011] Furthermore, in step S3.1, the specific steps for coupled perturbation mapping and wavefront compensation are as follows: S3.11, Coupled Model Construction and Disturbance Tributary Mapping: Receive the S2 active flow allocation command, establish a kinematic and dynamic coupling model between the three actuators of boom, stick, and bucket, identify in real time the feedback disturbances caused by boom extension, stick effective lever arm change, boom load distribution offset caused by stick retraction, and bucket attitude angle change, quantify the coupling disturbances into the interference intensity of each tributary on the main channel in the topology model, and establish the disturbance transmission topology; S3.12, Wavefront Delay Calculation and Phase Alignment Compensation: Based on the length of each channel in the topology model, the elastic modulus of the oil, and the oil temperature, the pressure wavefront propagation delay is calculated in real time. The feedforward compensation is then phase-aligned according to this delay and superimposed onto the control command so that the compensation and disturbance arrive synchronously at the target actuator.

[0012] Furthermore, in step S3.2, the specific steps of distributed collaborative optimization and feedback correction are as follows: S3.21, Node Decomposition and Rolling Optimization: The topology model reconstruction and compensation are decomposed into four distributed sub-controllers. Each sub-controller independently performs single-step local optimization based on local sensor information. The topology model state and compensation parameters are shared through a high-speed bus. Rolling time-domain optimization is used to transform long-term time-domain prediction into a local single-step optimization and a central coordinator constraint check iterative structure. The central coordinator is responsible for task synchronization and hard constraint arbitration. S3.22 Feedback Correction and Command Generation: Feedback correction based on the deviation between actual displacement and target trajectory is introduced to correct the feedforward compensation amount after phase alignment in step S3.1 in real time, eliminate residual deviation and unmodeled disturbance, and generate the final decoupled precise control commands for each actuator based on the correction results.

[0013] Furthermore, in step S4, the specific steps of the closed-loop evolution of the mapping weight coefficients are as follows: S4.1 Operation Data Recording and Performance Decay Triggering: During the continuous operation of steps S1 to S3, the data recording unit collects and stores the flow distribution deviation, trajectory tracking error, pressure fluctuation variance and valve action frequency of each operation cycle, constructs a multi-dimensional comprehensive performance evaluation function, and monitors the control performance decay under complex terrain conditions in real time based on this function. When the evaluation index is lower than the preset ratio of the historical best performance, the mapping weight coefficient update mechanism is automatically triggered, and the parameter space evolution process is started. S4.2 Evolution of Mapping Weight Coefficient Kernel Function and Knowledge Base Evolution: The probability density of the mapping weight coefficients at corresponding coordinate points in the terrain load joint parameter space is evolved using a kernel function smoothing method to make the distribution of mapping weight coefficients form a smooth and continuous manifold. Based on the evaluation results, the mapping weight coefficients of the flow allocation path in step S2 are enhanced or attenuated, and the mapping weight coefficients of the decoupling compensation path in step S3 are modified in a linked manner. The updated topology model parameters and terrain features are associated and written into a non-volatile storage unit to form a working condition knowledge base.

[0014] This application also provides a decoupled collaborative optimization control system for tracked excavator operation in complex terrain, including: a perception and site establishment module, a flow arbitration module, a decoupled collaborative module, a parameter evolution module, and a storage module; The sensing field module is used to collect chassis slope, cylinder pressure and displacement, main pump flow and oil temperature, build an extended state observer to identify nonlinear disturbances, establish a hydraulic network topology impedance model to map pump source, valve port and chamber as node channels, initialize channel impedance and flow distribution, and form the initial distribution of mapping weight coefficients. The flow arbitration module is used to determine the expected flow demand based on the actuator speed and load, determine the continuous weight by weighted interpolation of the mapping weight coefficients of adjacent operating condition clusters, and reconstruct the optimal flow allocation path. The decoupling and coordination module is used to establish the coupling relationship of the three actuators, map the disturbance to the topological model disturbance branch, determine the pressure wave propagation delay based on the channel length, oil elastic modulus and oil temperature, and superimpose the feedforward compensation after phase alignment; it is distributed to four distributed sub-controllers to independently execute single-step local optimization, and rolling time-domain optimization is used to ensure that the calculation cycle matches the hydraulic system control cycle; The parameter evolution module is used to record the operation cycle flow deviation, trajectory error and pressure fluctuation, and to evolve the probability density of the parameter space by triggering the update of the mapping weight coefficient based on the performance evaluation. The storage module is used to store the operating condition knowledge base and mapping weights.

[0015] The beneficial effects of this invention are as follows: This application provides a decoupling and collaborative optimization control method and system for tracked excavator operations in complex terrain. It transforms hydraulic system flow distribution into a calculable optimization space by establishing a hydraulic network topology impedance model. It achieves time delay matching decoupling through phase alignment of pressure wavefront propagation delay. Automatic switching between open-chain and closed-chain modes avoids contact impact. Continuous probability density interpolation eliminates operational condition transition jitter. Distributed rolling optimization overcomes the bottleneck of centralized computing power. A closed-loop evolution of mapping weight coefficients driven by a kernel function smoothing method achieves long-term adaptive control strategy. This results in flow distribution evolving from passive lag to active prediction, coupling interference evolving from unquantifiable to directional cancellation, control architecture evolving from centralized global solution to distributed local collaboration, operational condition adaptability evolving from discrete lookup table jumps to continuous smooth transitions, and system performance evolving from fixed preset parameters to continuous evolution throughout the operation process. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort: Figure 1 This is a diagram illustrating the overall method steps of this application.

[0017] Figure 2 This is an extended diagram of step S2 of this application.

[0018] Figure 3 This is an extended diagram of step S2.1 of this application.

[0019] Figure 4This is an extended diagram of step S2.2 of this application.

[0020] Figure 5 This is an extended diagram of step S3 of this application.

[0021] Figure 6 This is an extended diagram of step S3.1 of this application.

[0022] Figure 7 This is an extended diagram of step S3.2 of this application.

[0023] Figure 8 This is the system flowchart for this application. Detailed Implementation

[0024] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Example 1, as Figure 1 A decoupling and collaborative optimization control method for tracked excavator operation in complex terrain is proposed, with the following specific steps: S1. Multi-source sensing field construction: Sensors are set up on the chassis and working device to collect slope, cylinder pressure and displacement, main pump flow and oil temperature. An extended state observer is constructed to identify oil temperature drift, component wear and pipeline elastic deformation nonlinear disturbances. A hydraulic network topology impedance model is established to map the pump source, valve port and chamber as node channels. The channel impedance and flow distribution are initialized to form the initial distribution of the working condition-control parameter mapping weight coefficient. S2. Active traffic arbitration: Based on the data from step S1, the expected traffic demand is determined according to the actuator speed and load. The weighted interpolation of the mapping weight coefficients of adjacent operating condition clusters is used to determine the continuous weights, and the optimal traffic allocation path is reconstructed. S3. Decoupling and Cooperative Optimization: According to the instructions in step S2, establish the coupling relationship of the three actuators, map the disturbance to the topological model disturbance branch, determine the pressure wave propagation delay based on the channel length, oil elastic modulus and oil temperature, and superimpose the feedforward compensation after phase alignment; allocate to four distributed sub-controllers to independently execute single-step local optimization, and use rolling time domain optimization to ensure that the calculation cycle matches the hydraulic system control cycle; S4. Closed-loop evolution of mapping weight coefficients: Record the operation cycle flow deviation, trajectory tracking error and pressure fluctuation, construct a multi-dimensional comprehensive performance evaluation function, and trigger the update of mapping weight coefficients when the comprehensive performance score is lower than the preset proportion of the historical best score, evolve the probability density of the parameter space, and write the updated parameters into storage to form a knowledge base.

[0026] In this embodiment: Step S1 uses multi-source sensors to collect key parameters such as chassis slope, actuator pressure and displacement, main pump flow and oil temperature in real time. An extended state observer is constructed to identify nonlinear disturbances such as oil temperature drift, component wear and pipeline elastic deformation online. Based on this, a hydraulic network topology impedance model is established and the pump source, valve port and chamber are mapped as node channels. At the same time, the mapping weight coefficient distribution is initialized, a closed-chain detection layer and terrain load joint parameter space are constructed. The task of comprehensive perception of actuator state and working environment under complex working conditions and construction of topology impedance model is completed. This achieves the purpose of avoiding strong dependence on the accurate mathematical model of hydraulic system, realizing automatic identification of open chain and closed chain operation modes, and eliminating the hidden danger of boundary parameter jump in traditional discrete working condition clusters. It provides an accurate and low-latency state basis for subsequent active arbitration and decoupling compensation.

[0027] Step S2 receives the topological impedance model status and mode marker. In open-chain mode, a flow demand prediction model is established based on the actuator speed acceleration and load estimate. In closed-chain mode, a fixed constraint on the tooth tip position is introduced for joint optimization solution. At the same time, based on the spatial coordinates of the terrain load parameters, the weight coefficients of the adjacent working condition clusters are weighted and interpolated to generate continuous weights. The optimal flow allocation path is reconstructed and an active command is generated. This completes the transformation of flow allocation from passive response to active prediction and the task of smooth switching of parameters in the working condition transition area. It achieves the purpose of eliminating the response delay of the traditional fixed priority strategy, suppressing actuator jitter during working condition switching, and avoiding the excessive impact load at the moment of bucket contact. This makes the flow allocation match the complex terrain features in real time and the transition is smooth.

[0028] Step S3 establishes the kinematic and dynamic coupling relationship of the three actuators, mapping lever arm changes, load offsets, and attitude disturbances to topological model disturbance branches. Based on channel length, oil elastic modulus, and oil temperature, the pressure wave propagation delay is calculated and phase alignment compensation is performed. Simultaneously, the topological model reconstruction and compensation are decomposed into four distributed sub-controllers for independent single-step local optimization. Rolling time-domain optimization is used to ensure matching of calculation cycles, and a feedback correction loop is introduced to correct residual deviations in real time. This completes the task of directional cancellation of strong coupling interference between multiple actuators and breaking through the bottleneck of centralized computing power. It achieves the goals of eliminating the compensation phase misalignment caused by the propagation delay of hydraulic medium pressure waves, ensuring the synergy of the boom and stick composite action force, and improving the trajectory tracking accuracy in complex terrain while meeting the real-time requirements of the control cycle.

[0029] Step S4 continuously records the flow distribution deviation, trajectory tracking error, pressure fluctuation variance, and valve action frequency for each work cycle. It constructs a comprehensive performance evaluation function and triggers a mapping weight coefficient update mechanism when performance degrades. The probability density of the mapping weight coefficient in the terrain load joint parameter space is evolved using a kernel function smoothing method. The mapping weight coefficients of the flow distribution and decoupling compensation paths are simultaneously corrected. The updated parameters are written into storage to form a working condition knowledge base. This completes the task of continuous optimization and knowledge accumulation of the control strategy throughout the work process, achieving the goals of eliminating inefficient control strategies, solidifying historical successful experiences, and eliminating the failure of preset parameters under different terrains. This enables the system to have long-term adaptive evolution capabilities.

[0030] Compared with the limitations of existing technologies, such as the passive response mechanism of fixed priority or load sensing pressure compensation, the computational bottleneck of centralized controller global optimization, strong dependence on the precise mathematical model of the hydraulic system, boundary parameter jumps caused by discrete working condition lookup, compensation phase misalignment caused by ignoring the propagation delay of hydraulic medium pressure waves, and contact shock and actuator forced pulling caused by the failure to identify kinematic closed-chain modes, this solution transforms the hydraulic system flow distribution into a computable optimization space through a hydraulic network topology impedance model and a mapping weight coefficient distribution mechanism. It achieves time delay matching decoupling by using phase alignment of the pressure wave front propagation delay, avoids contact shock by automatically switching between open-chain and closed-chain modes, eliminates working condition transition jitter by using continuous probability density interpolation, overcomes the centralized computing power bottleneck by relying on distributed rolling optimization, and achieves long-term adaptive control strategy by using the closed-loop evolution of mapping weight coefficients driven by the kernel function smoothing method. The above improvements enable the flow distribution response to change from passive lag to active prediction, the multi-actuator coupling interference to change from unquantifiable to directional cancellation, the control architecture to change from centralized global solution to distributed local collaboration, the working condition adaptability to change from discrete lookup table jumps to continuous smooth transitions, and the system performance to change from fixed preset parameters to continuous evolution with the operation process, thereby improving the operating efficiency, trajectory accuracy, operation smoothness and long-term operation stability of tracked excavators in complex terrain.

[0031] Example 2, as Figure 1 In step S1, the specific steps for multi-source sensing field establishment are as follows: S1.1 Multi-source sensing topology model construction: Sensors are deployed on the chassis and working device to collect data on slope, support roller reaction force, cylinder pressure and displacement, pump flow and oil temperature. An extended state observer is constructed to identify oil temperature drift, component wear and nonlinear disturbances of pipeline elastic deformation online, output terrain feature parameters and load values, establish a hydraulic network topology impedance model, and map the pump source outlet, valve port and chamber as node channels. The nodes include source nodes, intermediate nodes and terminal nodes. The channel impedance includes liquid resistance, liquid capacity and liquid sensing. The channel impedance and flow distribution are initialized. S1.2 Initialization of Working Condition Mapping Weight Coefficients: Historical flow distribution is associated with terrain feature parameters to form an initial distribution of mapping weight coefficients. A closed-chain detection layer is added to the topological impedance model. The contact state is inferred based on the bucket attitude angle and pressure changes. It is determined whether the boom, stick, and bucket form a kinematic closed chain. The closed-chain or open-chain mode is marked. The terrain load parameter space is constructed with slope, ground stiffness, and subsidence as coordinate axes, so that the mapping weight coefficients are stored according to the probability density distribution in the space.

[0032] In this embodiment: Step S1.1 involves deploying various types of sensors at key locations on the chassis and working device to collect parameters such as chassis slope, support roller reaction force, hydraulic cylinder pressure and displacement of each actuator, main pump outlet flow rate, and hydraulic oil temperature in real time. An extended state observer is constructed, and the oil viscosity drift caused by oil temperature changes, the leakage coefficient change caused by component wear, and the volumetric efficiency decay caused by pipeline elastic deformation are used as extended state variables. These variables are then jointly identified online with the load disturbance of each actuator and the coupling interference term between actuators. The distributed state identification result, which includes terrain feature parameters and real-time load estimates of each actuator, is output. Based on this, a hydraulic network topology impedance model is established, mapping the pump source outlet, multi-way valve port, and each actuator chamber as node channels, and initializing the impedance and flow distribution state of each channel. This step completes the task of comprehensively perceiving the actuator state and working environment under complex working conditions and constructing a topological impedance model. It achieves the goals of avoiding strong dependence on the precise mathematical model of the hydraulic system, incorporating nonlinear factors that are difficult to model in traditional methods into the extended state for real-time estimation, and providing accurate and low-latency state information for subsequent control links, thus alleviating the dilemma of traditional control strategies' dependence on model accuracy.

[0033] Step S1.2 involves associating and mapping verified flow distribution records from historical operations with terrain feature parameters to form an initial distribution of working condition-control parameter mapping weight coefficients with positive feedback guidance. A closed-chain detection layer is added to the topological impedance model. Based on the bucket attitude angle and hydraulic cylinder pressure change characteristics, the tooth tip contact state is inferred in real time to determine whether the boom, stick, and bucket form a kinematic closed chain with the ground or material pile. When a closed chain is detected, the topological impedance model is marked as a closed-chain mode; otherwise, it is marked as an open-chain mode. Simultaneously, a terrain load joint parameter space is constructed, using slope, ground stiffness, and subsidence as continuous coordinate axes, and the mapping weight coefficients are stored according to the continuous probability density distribution in this parameter space. This step completes the initial association mapping between historical successful experience and current terrain working conditions, automatic identification of operating modes, and continuous storage of parameter space. It achieves the goals of establishing a positive association between historical successful experience and terrain working conditions, realizing automatic identification of open-chain and closed-chain operating modes, and eliminating the potential for boundary parameter jumps caused by traditional discrete working condition cluster classification. This provides a smooth and continuous prior basis with optimization guidance for subsequent active flow arbitration and decoupling compensation.

[0034] The hydraulic network topology impedance model is established as follows: the main pump outlet is taken as the source node, the valve ports of the multi-way valve are taken as intermediate nodes, and the rodless and rod chambers of the boom hydraulic cylinder, the rodless and rod chambers of the stick hydraulic cylinder, the rodless and rod chambers of the bucket hydraulic cylinder, and the two chambers of the swing motor are taken as terminal nodes. The hydraulic pipelines and valve orifices between the nodes are taken as channels. The channel impedance consists of hydraulic resistance, hydraulic capacity, and hydraulic induction. Hydraulic resistance reflects the influence of valve opening and oil viscosity on flow resistance. Hydraulic capacity reflects the influence of chamber volume and oil elastic modulus on pressure accumulation. Hydraulic induction reflects the influence of pipeline length and oil density on pressure fluctuation inertia. Each node satisfies the conservation relationship that the inflow flow is equal to the outflow flow. The flow distribution is solved based on the impedance of each channel and the pressure relationship of each node. The oil viscosity is corrected according to the viscosity-temperature characteristic curve based on the oil temperature. The oil elastic modulus is corrected according to the oil temperature and gas content using empirical formulas.

[0035] The working condition-control parameter mapping weight coefficient is initialized as follows: The verified and effective flow allocation ratios from historical operations are associated and stored with terrain feature parameters. The flow allocation ratio is the ratio of the target flow of each actuator to the total flow of the main pump outlet. Terrain feature parameters include slope angle, ground stiffness grade, and track roller subsidence. The mapping weight coefficient represents the priority of flow allocation for each actuator at the corresponding working condition point. Its initial value is determined based on the calibration data before the first operation or the factory preset MAP map, and the value range is 0 to 1. The flow allocation ratio is determined based on the relative magnitude of the mapping weight coefficients of each actuator at the current working condition point. Actuators with higher mapping weight coefficient values ​​receive a relatively larger flow allocation ratio. The horizontal axis of the MAP map is the slope angle, the vertical axis is the ground stiffness grade, and the table values ​​are the flow allocation ratios of each actuator.

[0036] The extended state observer is constructed as follows: the oil viscosity drift caused by oil temperature changes, the leakage coefficient change caused by component wear, and the volumetric efficiency decay caused by pipeline elastic deformation are taken as extended state variables. The initial value of the extended state variables is 1.0, indicating that the system is in an ideal, undisturbed state. The observer uses the displacement and velocity of the hydraulic cylinder as measurable outputs. By introducing the above extended state variables, the original system state space is expanded. The observer gain matrix is ​​used to correct the output estimation deviation. The oil temperature viscosity drift, leakage coefficient change, and volumetric efficiency decay are identified online. The terrain feature parameters and load estimates are output. The oil temperature viscosity drift is corrected by calculating the correction coefficient according to the viscosity-temperature characteristic curve based on the deviation between the measured oil temperature and the standard oil temperature. The leakage coefficient change is determined by the initial value based on the leakage test data of the hydraulic cylinder or the factory calibration data and corrected according to the wear rate over time. The volumetric efficiency decay is calculated based on the pipeline elastic modulus and pressure fluctuation amplitude.

[0037] The closed-chain detection layer operates as follows: It monitors the pressure change rate of the rodless chamber of the bucket hydraulic cylinder and the bucket attitude angle change rate in real time. When the pressure change rate exceeds a preset proportion of the rated working pressure within a single control cycle and the attitude angle change rate is lower than a preset threshold, it is determined that the bucket tooth tip is in contact with the ground or material pile, and the topological impedance model is marked as closed-chain mode. Otherwise, it is marked as open-chain mode. The preset proportion is determined based on the rated pressure of the hydraulic system and the impact characteristics of the bucket operation, and the value range is 15% to 25%. The preset threshold is determined based on the normal no-load swing angular velocity of the bucket, and the value range is 5% to 10% of the rated angular velocity.

[0038] The terrain load joint parameter space is constructed as follows: the slope angle, ground stiffness grade and track roller sinking are used as three-dimensional coordinate axes. The coordinate axis range is determined according to the working conditions of the excavator. The mapping weight coefficients are stored according to the continuous probability density distribution in the parameter space. The probability density distribution is initialized by kernel density estimation based on the number of records falling into the neighborhood of each coordinate point in the historical operation cycle.

[0039] Compared with the limitations of existing technologies, such as relying on precise mathematical models for centralized global calculations, using discrete working condition lookup tables leading to boundary parameter jumps, and failing to identify kinematic closed-chain modes resulting in contact impacts and forced actuator tension, step S1 transforms the hydraulic system flow and pressure distribution into a calculable spatial topological relationship through the synergistic effect of multi-source sensing and topological impedance model initialization. It uses an extended state observer to identify nonlinear disturbances such as oil temperature drift, component wear, and pipeline elastic deformation online, avoiding the strong dependence on precise modeling in traditional methods. Through the closed-chain detection layer, it realizes the automatic identification of open-chain and closed-chain operation modes, enabling the control system to switch control strategies according to the actual operation state, avoiding excessive impact loads at the moment of bucket contact. By storing the continuous probability density distribution of the terrain load combined with the parameter space, it replaces the traditional discrete working condition cluster classification, eliminating parameter jumps and actuator jitter in the working condition transition area. These improvements enable state perception to shift from relying on offline model identification to online real-time estimation, working condition classification to shift from discrete point set jumps to continuous smooth transitions, and operation mode recognition to automatic state inference, thereby improving the environmental adaptability, state estimation accuracy, and decision reliability of subsequent control links of the tracked excavator control system in complex terrain.

[0040] Example 3, as Figure 2 In step S2, the specific steps of the traffic active arbitration are as follows: S2.1 Mode Switching and Flow Demand Prediction: Receive the S1 topological impedance model status, mapping weight coefficient distribution and mode label. In open chain mode, establish a flow demand prediction model based on the estimated speed, acceleration and load of the boom, stick, bucket, and slewing mechanism, and calculate the expected flow demand of each actuator. In closed chain mode, introduce a fixed constraint on the tooth tip position, and jointly solve the flow demand of the three actuators into the minimum energy consumption coordination control quantity that satisfies the closed chain constraint. S2.2 Mapping Weight Coefficient Interpolation and Path Reconstruction: Based on the coordinates of the current working point in the terrain load parameter space, the mapping weight coefficient distribution of the neighboring working clusters is weighted and interpolated to generate continuous flow allocation weights. This gives priority to paths with high mapping weight coefficient values ​​and those that match the terrain load. The optimal flow allocation path is reconstructed in the topology impedance model, and active flow allocation instructions for each actuator are generated.

[0041] In this embodiment: Step S2.1 receives the topological impedance model state, mapping weight coefficient distribution, and mode label. When the topological impedance model is in open-chain mode, a multi-actuator joint flow demand prediction model is established based on the speed, acceleration, and load estimates of the boom, stick, bucket, and slewing mechanism. The expected flow demand of each actuator in the next control cycle is calculated. When the topological impedance model is in closed-chain mode, a fixed constraint on the bucket tooth tip position is introduced, and the flow demand of the three actuators—boom, stick, and bucket—is incorporated into the joint optimization framework to solve for the minimum energy consumption coordinated control quantity that satisfies the closed-chain constraint. This step completes the differentiated prediction and coordinated solution of flow demand under different operating modes, achieving the goals of eliminating the passive response delay after flow saturation in traditional fixed priority strategies, avoiding excessive impact loads caused by independent control when the bucket contacts the ground or the material pile, and preventing forced pulling between actuators. It transforms flow distribution from passive overflow diversion to active prediction distribution, realizing adaptive switching of operating conditions and energy optimization coordination.

[0042] Step S2.2, based on the coordinates of the current operating point in the terrain-load joint parameter space, performs weighted interpolation on the distribution of mapping weight coefficients of neighboring operating point clusters to generate continuously changing flow allocation weights. This ensures that paths with high mapping weight coefficients and matching terrain loads receive priority selection probability. The optimal flow allocation path is then screened and reconstructed in the topological impedance model, and the reconstruction result is converted into active flow allocation commands for each actuator. This step completes the dynamic reconstruction task of the optimal flow allocation path based on a continuous probability density distribution. It achieves the goals of eliminating boundary parameter jumps and actuator jitter caused by traditional discrete operating point lookup tables, smoothing the flow allocation strategy in complex terrain transition areas, and suppressing flow allocation oscillations among multiple actuators. It replaces the traditional passive response mechanism of fixed priority or load sensing pressure compensation with an active prediction allocation mechanism guided by mapping weight coefficients.

[0043] Weighted interpolation is performed as follows: In the joint parameter space of terrain load where the current operating point is located, the three historical operating point clusters with the closest Euclidean distance to the current operating point are selected as neighboring clusters. The interpolation weight of the mapping weight coefficient of the current operating point to each neighboring cluster is calculated according to the inverse distance weighting principle. The inverse distance weighting principle is that the closer the neighboring cluster is, the greater its interpolation weight. The interpolated mapping weight coefficient is used as the correction coefficient of the flow distribution ratio of each actuator to generate continuously changing flow distribution weight.

[0044] Compared to the limitations of existing technologies, such as passive response mechanisms using fixed priorities or load sensing pressure compensation, boundary parameter jumps caused by discrete operating condition lookups, and contact shocks and energy waste due to unrecognized closed-chain modes, step S2 achieves a fundamental shift in flow allocation from passive lag to active prediction through the synergistic effect of mode switching, flow demand prediction, mapping weight coefficient interpolation, and path reconstruction. In open-chain mode, a prediction model is established based on the dynamic characteristics of the actuator, enabling accurate prediction of flow demand at the beginning of the control cycle. In closed-chain mode, joint optimization is performed by introducing fixed position constraints, avoiding impact loads and forced tension caused by independent control. Simultaneously, continuous probability density interpolation replaces discrete operating condition lookups, allowing flow allocation weights to change smoothly in complex terrain transition areas, eliminating parameter jumps and actuator jitter at the junction of slopes and flat ground, and at the transition between soft and hard surfaces, as in traditional methods. The above improvements transform the flow distribution response from passive mechanical diversion after overflow to active prediction within the control cycle, the working condition adaptability from rigid switching of discrete point sets to smooth transition of continuous manifolds, and the energy consumption from disordered competition of independent flow taking by each actuator to coordinated optimization under closed-chain constraints. These improvements enhance the response speed, trajectory accuracy, operational smoothness, and energy utilization efficiency of tracked excavators in complex terrain.

[0045] Example 4, as Figure 3 In step S2.1, the specific steps for mode switching and traffic demand prediction are as follows: S2.11 Open Chain Flow Demand Prediction: Receive the S1 topology impedance model status, mapping weight coefficient distribution and mode label. When the topology impedance model is in open chain mode, based on the estimated values ​​of the speed, acceleration and load of the boom, stick, bucket and slewing mechanism, combined with terrain feature parameters, establish a multi-actuator joint flow demand prediction model, calculate the expected flow demand of each actuator in the next control cycle, and form the independent flow allocation benchmark value of each actuator in open chain mode. S2.12, Closed-chain constraint coordination solution: When the topological impedance model is in closed-chain mode, the fixed constraint of the tooth tip position is introduced based on the closed-chain detection layer mark, and the kinematic closed-chain constraint equations of the boom, stick and bucket three actuators are established. The flow demand of the three actuators is incorporated into the joint optimization framework, and the minimum energy consumption coordination control quantity that satisfies the closed-chain constraint is solved to form the benchmark value of the coupled flow distribution of each actuator in closed-chain mode.

[0046] In this embodiment: Step S2.11 receives the topological impedance model state, mapping weight coefficient distribution, and mode label. When the topological impedance model is in open-chain mode, a multi-actuator joint flow demand prediction model is established based on the estimated speed, acceleration, and load values ​​of the boom, stick, bucket, and slewing mechanism, combined with terrain feature parameters. The expected flow demand of each actuator in the next control cycle is calculated, forming an independent flow allocation benchmark value for each actuator in open-chain mode. This sub-step completes the task of dynamic prediction and independent benchmark value generation of the flow demand of each actuator in open-chain operation mode. It achieves the goal of eliminating the passive response delay of the traditional fixed priority strategy, which only mechanically diverts flow through overflow valves or pressure reducing valves after flow saturation. It enables the flow allocation to be actively predicted based on the dynamic characteristics of the actuators and the terrain conditions at the beginning of the control cycle, transforming the passive overflow response into active prediction allocation, and providing a flow demand benchmark with working condition adaptability for subsequent path reconstruction.

[0047] Step S2.12: When the topological impedance model is in closed-chain mode, a fixed constraint on the bucket tooth tip position is introduced based on the closed-chain detection layer markers. Kinematic closed-chain constraint equations for the boom, stick, and bucket are established, incorporating the flow demands of the three actuators into the joint optimization framework. The minimum energy consumption coordinated control quantity satisfying the closed-chain constraints is solved, forming the baseline value for coupled flow distribution of each actuator in the closed-chain mode. This sub-step completes the joint coordination solution of the flow demands of multiple actuators in the closed-chain operation mode. It achieves the goals of avoiding excessive impact loads caused by independent control of each actuator at the moment the bucket tooth tip contacts the ground or the material pile, eliminating forced pulling and energy loss of the boom and stick under closed-chain constraints, and ensuring that flow distribution achieves minimum energy consumption coordination while satisfying the fixed position constraints, transforming independent and disordered flow extraction into constrained coordinated optimization.

[0048] Compared to existing technologies, which suffer from limitations such as passive response mechanisms using fixed priorities or load-sensing pressure compensation, contact impacts and forced actuator tension due to unrecognized kinematic closed-chain modes, and energy losses caused by independent flow extraction by each actuator, step S2.1 achieves adaptive switching of operating modes and differentiated optimization of flow allocation through the synergistic effect of open-chain flow demand prediction and closed-chain constraint coordination. In open-chain mode, a prediction model is established based on the dynamic characteristics of the actuators and terrain parameters, enabling accurate prediction of flow demand at the beginning of the control cycle. In closed-chain mode, joint optimization is performed by introducing fixed position constraints, avoiding impact loads and forced tension caused by independent control.

[0049] The above improvements transform flow allocation from passive mechanical diversion after overflow to active prediction, and from disorderly competition where each actuator independently draws flow to coordinated optimization under closed-chain constraints, thereby improving the response speed and energy utilization efficiency of the tracked excavator's compound actions in complex terrain.

[0050] Example 5, as Figure 4 In step S2.2, the specific steps for mapping weight coefficient interpolation and path reconstruction are as follows: S2.21, Weighted Interpolation of Mapping Weight Coefficients and Weight Generation: Receives the traffic allocation baseline value output from S2.1, and performs weighted interpolation on the distribution of mapping weight coefficients of neighboring working condition clusters based on the coordinates of the current working point in the terrain load joint parameter space. This generates continuously changing traffic allocation weights, giving priority to paths with high mapping weight coefficient values ​​and matching terrain load, and determining the set of candidate paths for traffic allocation of each actuator, providing a screening basis for path reconstruction. S2.22 Optimal Path Reconstruction and Command Generation: In the topological impedance model, the optimal flow distribution path is screened and reconstructed from the candidate path set according to the selection probability. The path is optimized by combining the impedance of each channel. The reconstruction result is converted into active flow distribution commands for the boom, stick, bucket and slewing mechanism and output to step S3.

[0051] In this embodiment: Step S2.21 receives the flow allocation benchmark value output from S2.1, and based on the coordinates of the current working point in the terrain-load joint parameter space, performs weighted interpolation on the mapping weight coefficient distribution of neighboring working point clusters to generate continuously changing flow allocation weights. This ensures that paths with high mapping weight coefficients and matching terrain loads receive priority selection probability, thus determining the set of candidate flow allocation paths for each actuator. This sub-step completes the flow allocation weight generation task, transitioning from discrete working point lookup to continuous probability density distribution. It achieves the goals of eliminating parameter jumps and actuator jitter caused by traditional discrete working point cluster classification at boundaries, ensuring that flow allocation weights change continuously and smoothly with the working point coordinates, and providing a working point-adaptive candidate selection basis for subsequent optimal path reconstruction.

[0052] Step S2.22 involves selecting and reconstructing the optimal flow distribution path from the candidate path set based on the selection probability in the topological impedance model. Path optimization is performed by combining the impedance of each channel, and the reconstruction result is transformed into active flow distribution commands for the boom, stick, bucket, and slewing mechanism. This sub-step completes the task of dynamic reconstruction and active command generation of the optimal flow distribution path based on mapping weight coefficients. It achieves the goals of replacing the traditional passive response mechanism of fixed priority or load sensor pressure compensation, enabling real-time matching of flow distribution with complex terrain features, suppressing flow distribution oscillations among multiple actuators, and transforming passive overflow diversion into active predictive distribution.

[0053] Compared to the limitations of existing technologies, such as passive response mechanisms using fixed priorities or load sensing pressure compensation, and boundary parameter jumps and actuator jitter caused by discrete operating condition lookup tables, step S2.2 achieves a fundamental shift in flow allocation strategy from discrete rigid switching to continuous smooth transition through the synergistic effect of weighted interpolation of mapping weight coefficients and optimal path reconstruction. S2.21 performs weighted interpolation of mapping weight coefficients for adjacent operating condition clusters based on the spatial coordinates of the terrain-load joint parameter, ensuring that flow allocation weights change continuously with the operating condition. S2.22 reconstructs the optimal path in the topological impedance model based on selection probability, converting the result into active allocation commands for each actuator. These improvements transform operating condition adaptability from discrete point set jumps to continuous manifold smooth transitions, and flow allocation from passive overflow response to active prediction, enhancing operational smoothness and control stability in complex terrain transition areas.

[0054] Example 6, as Figure 5 In step S3, the specific steps of decoupling and collaborative optimization are as follows: S3.1 Coupled Disturbance Mapping and Wavefront Compensation: Receive the S2 active flow distribution command, establish a kinematic and dynamic coupling model of the three actuators (boom, stick, and bucket), identify lever arm changes, load offsets, and attitude feedback disturbances in real time, map the coupled disturbances into topological model disturbance branches, calculate the pressure wavefront propagation delay based on channel length, oil elastic modulus, and oil temperature, and then align the feedforward compensation amount with the phase of the delay and superimpose it onto the control command. S3.2 Distributed Cooperative Optimization and Feedback Correction: The topology model reconstruction and compensation are decomposed into four distributed sub-controllers. Each sub-controller independently performs single-step local optimization based on local sensor information. The calculation cycle of single-step local optimization is no greater than the hydraulic system control cycle. The topology model state and compensation parameters are shared through a high-speed bus. Rolling time-domain optimization is used to ensure that the calculation cycle matches the hydraulic system control cycle. A feedback correction link is introduced to correct the feedforward compensation amount in real time, generating precise control commands for each actuator after decoupling.

[0055] In this embodiment: Step S3.1 receives the active flow distribution command from step S2, establishes a kinematic and dynamic coupling model of the boom, stick, and bucket actuators, identifies in real time the changes in the effective lever arm of the stick caused by boom extension, the shift in boom load distribution caused by stick retraction, and the feedback disturbances generated by changes in bucket attitude. The coupled disturbances are mapped to disturbance branches in the topology model. Based on the length of each channel, the elastic modulus of the hydraulic fluid, and the oil temperature, the pressure wave front propagation delay is calculated. The feedforward compensation is then aligned with the phase of this delay and superimposed onto the control command. This step completes the explicit quantification and directional cancellation of strong coupling disturbances between multiple actuators. It achieves the goal of transforming coupled disturbances, which are considered external and unmeasurable disturbances in traditional methods, into calculable disturbance transmission paths within the topology model; eliminating the misalignment between the compensation and the arrival time of the disturbance caused by neglecting the propagation delay of the hydraulic medium pressure wave; and ensuring that the feedforward compensation and the disturbance arrive synchronously at the target actuator. This achieves time delay matching decoupling from the source of the control command.

[0056] Step S3.2 decomposes the topology model reconstruction and compensation into four distributed sub-controllers: boom, stick, bucket, and slewing. Each sub-controller independently performs single-step local optimization based on local sensor information. The topology model state and compensation parameters are shared via a high-speed bus. Rolling time-domain optimization transforms long-term prediction into an iterative structure of local single-step optimization and central coordinator constraint checks, ensuring the calculation cycle matches the hydraulic system control cycle. A feedback correction loop is introduced to correct the feedforward compensation amount in real time, generating precise control commands for each decoupled actuator. This step completes the transformation from centralized control to a distributed collaborative architecture and the decomposition of long-term time-domain optimization into short-cycle rolling optimization. It overcomes the computing power bottleneck of centralized controllers, maintains global coordination of multiple actuators while meeting the real-time requirements of the control cycle, eliminates residual deviations and unmodeled disturbances, and achieves force-position collaborative closed-loop control.

[0057] Rolling time-domain optimization is performed as follows: the prediction time domain has 3 to 5 steps, and the control time domain has 1 step, with the step size equal to the hydraulic system control cycle; each distributed sub-controller performs single-step optimization calculations locally, using the deviation between the predicted output and the reference trajectory and the control increment as optimization targets; the central coordinator performs global constraint checks, including the total engine power not exceeding the limit, the system's maximum pressure limit, and closed-chain constraint equations; the central coordinator does not participate in solving real-time control quantities, but is only responsible for task synchronization and hard constraint arbitration.

[0058] Compared to existing technologies, which suffer from limitations such as passive suppression of coupling interference, neglect of hydraulic medium pressure wave propagation delay leading to compensation phase misalignment, and difficulty for centralized controllers to complete long-term time-domain optimization within short periods, step S3 achieves a transformation in multi-actuator decoupling from passive suppression to active directional cancellation, from time delay misalignment to matching, and from centralized global solution to distributed local collaboration through the synergistic effect of coupling disturbance mapping, wavefront compensation, distributed collaborative optimization, and feedback correction. S3.1 maps coupling interference to topological model disturbance branches and aligns the phases based on the pressure wave propagation delay, ensuring that the compensation arrives synchronously with the disturbance. S3.2 overcomes the computational bottleneck through distributed node independent optimization and a rolling time-domain strategy, using feedback correction to eliminate residual deviations. These improvements transform decoupling compensation from conservative estimation to directional cancellation of time delay matching control, and the control architecture from centralized global optimization to distributed local collaboration, improving trajectory tracking accuracy and coordination of complex actions in complex terrain.

[0059] Example 7, as Figure 6 In step S3.1, the specific steps for coupled perturbation mapping and wavefront compensation are as follows: S3.11, Coupled Model Construction and Disturbance Tributary Mapping: Receive the S2 active flow allocation command, establish a kinematic and dynamic coupling model between the three actuators of boom, stick, and bucket, identify in real time the feedback disturbances caused by boom extension, stick effective lever arm change, boom load distribution offset caused by stick retraction, and bucket attitude angle change, quantify the coupling disturbances into the interference intensity of each tributary on the main channel in the topology model, and establish the disturbance transmission topology; S3.12, Wavefront Delay Calculation and Phase Alignment Compensation: Based on the length of each channel in the topology model, the elastic modulus of the oil, and the oil temperature, the pressure wavefront propagation delay is calculated in real time. The feedforward compensation is then phase-aligned according to this delay and superimposed onto the control command so that the compensation and disturbance arrive synchronously at the target actuator.

[0060] In this embodiment: Step S3.11 receives the active flow allocation command from step S2, establishes a kinematic and dynamic coupling model among the three actuators (boom, stick, and bucket), identifies in real time the feedback disturbances caused by the change in the effective lever arm of the stick due to boom extension, the shift in boom load distribution due to stick retraction, and the change in bucket attitude angle. The coupled disturbances are quantified as the interference intensity of each branch on the main channel in the topology model, and a disturbance propagation topology is established. This sub-step completes the explicit quantification of strong coupling disturbances among multiple actuators and the construction of the disturbance propagation topology. It achieves the goal of transforming coupled disturbances, which are considered external and unmeasurable disturbances in traditional methods, into calculable disturbance propagation paths within the topology model. This provides a disturbance source tracing basis for subsequent time delay matching and decoupling, making coupled disturbances transform from unquantifiable to directional cancellation.

[0061] Step S3.12 calculates the pressure wave propagation delay in real time based on the channel lengths, oil elastic modulus, and oil temperature of the topology model. The feedforward compensation is then phase-aligned according to this delay and superimposed onto the control command, ensuring that the compensation arrives synchronously with the disturbance at the target actuator. This sub-step completes the real-time calculation of the hydraulic medium pressure wave propagation delay and the phase alignment of the compensation, achieving the goal of eliminating the misalignment between the compensation and the disturbance arrival time caused by neglecting the hydraulic medium pressure wave propagation delay. It realizes time delay matching decoupling, transforming the feedforward compensation from an instantaneous response to a precise compensation arriving synchronously with the disturbance.

[0062] The wavefront delay calculation and phase alignment compensation are performed as follows: The pressure wavefront propagation delay is calculated in real time based on the length of each channel in the topology model, the elastic modulus of the oil, and the oil temperature. The propagation speed of the pressure wave in the oil is proportional to the square root of the elastic modulus of the oil and inversely proportional to the square root of the oil density. The pressure wavefront propagation delay is equal to the channel length divided by the pressure wave propagation speed. The feedforward compensation is phase aligned according to this delay and then superimposed on the control command so that the compensation and disturbance arrive synchronously at the target actuator.

[0063] Compared to existing technologies that treat coupled interference as an external disturbance and passively suppress it, fail to explicitly map it as an internal energy flow path, and ignore the propagation delay of hydraulic medium pressure waves leading to phase misalignment in compensation, step S3.1 achieves a transformation of coupled interference from an externally unmeasurable disturbance to an internally calculable disturbance transmission topology, and a shift of feedforward compensation from instantaneous response to time-delay matching and phase alignment, through the synergistic effect of coupled model construction, disturbance branch mapping, wavefront delay calculation, and phase alignment compensation. S3.11 establishes the disturbance transmission topology, making coupled interference quantifiable and traceable; S3.12 performs phase alignment based on the pressure wave propagation delay, ensuring that compensation arrives synchronously with the disturbance. These improvements transform decoupling compensation from conservative estimation to directional cancellation through time-delay matching control, enhancing trajectory tracking accuracy and the coordination of complex actions in complex terrain.

[0064] Example 8, as shown in Figure 7, the specific steps of distributed collaborative optimization and feedback correction in step S3.2 are as follows: S3.21, Node Decomposition and Rolling Optimization: The topology model reconstruction and compensation are decomposed into four distributed sub-controllers. Each sub-controller independently performs single-step local optimization based on local sensor information. The topology model state and compensation parameters are shared through a high-speed bus. Rolling time-domain optimization is used to transform long-term time-domain prediction into a local single-step optimization and a central coordinator constraint check iterative structure. The central coordinator is responsible for task synchronization and hard constraint arbitration. S3.22 Feedback Correction and Command Generation: Feedback correction based on the deviation between actual displacement and target trajectory is introduced to correct the feedforward compensation amount after phase alignment in step S3.1 in real time, eliminate residual deviation and unmodeled disturbance, and generate the final decoupled precise control commands for each actuator based on the correction results.

[0065] In this embodiment: Step S3.21 decomposes the topology model reconstruction and compensation into four distributed sub-controllers: boom, stick, bucket, and slewing. Each sub-controller independently performs single-step local optimization based on local sensor information. The topology model state and compensation parameters are shared via a high-speed bus. Rolling time-domain optimization transforms long-term prediction into an iterative structure of local single-step optimization and central coordinator constraint checks. The central coordinator is responsible for task synchronization and hard constraint arbitration. This sub-step completes the transformation from a centralized control architecture to a distributed collaborative architecture and the decomposition of long-term prediction optimization into short-cycle rolling optimization. It achieves the goals of overcoming the computing power bottleneck of centralized controllers, maintaining global coordination of multiple actuators while meeting the real-time requirements of the hydraulic system control cycle, and avoiding system loss of control due to single-point failure of the central node, thus matching the calculation cycle of each sub-controller with the hydraulic system control cycle.

[0066] Step S3.22 introduces a feedback correction loop based on the deviation between the actual displacement and the target trajectory. This loop corrects the feedforward compensation amount after phase alignment in step S3.1 in real time, eliminating residual errors and unmodeled disturbances. Based on the correction results, it generates the final, precise control commands for each actuator after decoupling. This sub-step completes the tasks of real-time elimination of feedforward compensation residual errors and generation of precise control commands. It achieves the goals of compensating for extended state observer identification errors and unmodeled dynamic factors, realizing force-position coordinated closed-loop control, and improving trajectory tracking accuracy and operational stability in complex terrain. This allows the decoupled control to simultaneously possess the dual advantages of rapid feedforward response and robust feedback correction.

[0067] Compared to existing technologies, which suffer from limitations such as insufficient centralized global optimization computing power leading to control cycle timeouts, difficulty in real-time solution of long-term predictions, and accumulation of residual errors due to the lack of feedback correction, step S3.2 achieves a transformation from a centralized to a distributed collaborative control architecture and a shift from single feedforward to composite control through the synergistic effect of node decomposition and rolling optimization, as well as feedback correction and instruction generation. S3.21 decomposes long-term prediction into local single-step optimization and central constraint checking, overcoming the computing power bottleneck. S3.22 uses feedback correction to eliminate feedforward residual bias, achieving a force-position collaborative closed loop. These improvements transform the control architecture from centralized to distributed real-time collaborative, and the decoupling from open-loop estimation to closed-loop correction, thereby improving trajectory tracking accuracy and system robustness.

[0068] Example 9, as Figure 1 In step S4, the specific steps of the closed-loop evolution of the mapping weight coefficients are as follows: S4.1 Operation Data Recording and Performance Decay Triggering: During the continuous operation of steps S1 to S3, the data recording unit collects and stores the flow distribution deviation, trajectory tracking error, pressure fluctuation variance and valve action frequency of each operation cycle, constructs a multi-dimensional comprehensive performance evaluation function, and monitors the control performance decay under complex terrain conditions in real time based on this function. When the evaluation index is lower than the preset ratio of the historical best performance, the mapping weight coefficient update mechanism is automatically triggered, and the parameter space evolution process is started. S4.2 Evolution of Mapping Weight Coefficient Kernel Function and Knowledge Base Evolution: The probability density of the mapping weight coefficients at corresponding coordinate points in the terrain load joint parameter space is evolved using a kernel function smoothing method to make the distribution of mapping weight coefficients form a smooth and continuous manifold. Based on the evaluation results, the mapping weight coefficients of the flow allocation path in step S2 are enhanced or attenuated, and the mapping weight coefficients of the decoupling compensation path in step S3 are modified in a linked manner. The updated topology model parameters and terrain features are associated and written into a non-volatile storage unit to form a working condition knowledge base.

[0069] In this embodiment: Step S4.1 During the continuous operation of steps S1 to S3, the data recording unit collects and stores operational data such as flow distribution deviation, trajectory tracking error, pressure fluctuation variance, and valve action frequency for each work cycle, constructing a multi-dimensional comprehensive performance evaluation function. Based on this function, the control performance degradation under complex terrain conditions is monitored in real time. When the evaluation index is lower than the preset proportion of the historical best performance, the mapping weight coefficient update mechanism is automatically triggered, initiating the parameter space evolution process. This sub-step completes the task of continuously collecting control performance data and automatically identifying performance degradation during the operation process, achieving the goal of establishing a performance evaluation benchmark based on actual operating data, avoiding manual intervention to trigger parameter updates, and ensuring that the timing of the mapping weight coefficient evolution accurately matches the actual control performance degradation. This provides a data-driven triggering basis for the subsequent smooth evolution of the kernel function.

[0070] Step S4.2 uses a kernel function smoothing method to evolve the probability density of the mapping weight coefficients at corresponding coordinate points in the terrain load joint parameter space, making the distribution of mapping weight coefficients form a smooth and continuous manifold. Based on the evaluation results, the mapping weight coefficients of the flow allocation path in Step S2 are enhanced or attenuated, and the mapping weight coefficients of the decoupling compensation path in Step S3 are linked and corrected. The updated topology model parameters and terrain features are associated and written into a non-volatile storage unit, forming a working condition knowledge base with positive feedback guidance. This sub-step completes the continuous smooth evolution of the mapping weight coefficient distribution and the multi-path linkage update task, achieving the goal of eliminating inefficient control strategies and solidifying historical successful experiences, making the mapping weight coefficient distribution continuously evolve with the operation process rather than discretely jumping, forming a reusable working condition knowledge base, and realizing the closed-loop self-evolution of control strategies with the operation process.

[0071] The performance evaluation function is constructed as follows: The comprehensive performance score is obtained by weighted summation of the flow distribution deviation score, trajectory tracking error score, pressure fluctuation variance score, and valve action frequency score. The score of each sub-item is determined based on the ratio of the actual value of the index to the target value. The target value is determined based on the factory calibration data or the first operation benchmark data. The preset ratio range is 80% to 90%. When the comprehensive performance score of the current operation cycle is lower than the preset ratio of the historical best score, the mapping weight coefficient update mechanism is triggered.

[0072] The kernel smoothing method is performed as follows: A Gaussian kernel function is used to smooth and update the probability density of the mapping weight coefficients of each coordinate point in the terrain load joint parameter space. The Gaussian kernel function assigns weights based on the distance between the current coordinate point and historical coordinate points. The closer the historical records are, the greater their influence on the mapping weight coefficients of the current coordinate point. When the performance evaluation index is lower than the preset proportion of the historical best value, the mapping weight coefficients corresponding to the historical successful records that fall within the neighborhood of the current coordinate point are enhanced, and the mapping weight coefficients corresponding to the failed records are decayed. The enhancement range is 5% to 15% of the weight of the historical successful records, and the decay range is 10% to 30% of the weight of the historical failed records. The enhancement and decay ranges are determined based on the proportion of the performance evaluation deviation to the rated index.

[0073] Compared to existing technologies, which suffer from limitations such as fixed preset control parameters leading to performance failure under complex operating conditions, reliance on manual calibration for parameter updates in the operating condition lookup method, boundary jumps caused by discrete operating condition cluster classification, and the lack of an automatic performance evaluation and parameter evolution mechanism based on actual operating data, step S4 achieves a fundamental shift in control strategy from fixed presets to data-driven adaptive evolution through the synergistic effect of operation data recording, performance decay triggering, and the evolution of the mapping weight coefficient kernel function and knowledge base. S4.1 establishes a multi-dimensional performance evaluation function and automatically triggers updates based on actual operating data, ensuring precise matching between parameter evolution timing and performance decay. S4.2 employs a kernel function smoothing method to create a smooth and continuous manifold distribution of the mapping weight coefficients, avoiding boundary jumps associated with discrete lookup tables, and eliminates inefficient strategies and solidifies successful experiences through an enhanced decay mechanism. The above improvements transform control parameters from fixed values ​​calibrated manually into a dynamic knowledge base that automatically evolves with the operation process; change working condition adaptability from rigid switching of discrete point sets to smooth transition of continuous manifolds; and change system performance from passive correction after preset parameter failure to active self-evolution based on data-driven processes. These improvements enhance the long-term operational stability, working condition adaptability, and operational efficiency of the tracked excavator control system in complex terrain.

[0074] This application also provides a decoupled collaborative optimization control system for the operation of a tracked excavator in complex terrain, such as... Figure 8It includes: a perception and field establishment module, a traffic arbitration module, a decoupling and collaboration module, a parameter evolution module, and a storage module; The sensing field module is used to collect chassis slope, cylinder pressure and displacement, main pump flow and oil temperature, build an extended state observer to identify nonlinear disturbances, establish a hydraulic network topology impedance model to map pump source, valve port and chamber as node channels, initialize channel impedance and flow distribution, and form the initial distribution of mapping weight coefficients. The flow arbitration module is used to determine the expected flow demand based on the actuator speed and load, determine the continuous weight by weighted interpolation of the mapping weight coefficients of adjacent operating condition clusters, and reconstruct the optimal flow allocation path. The decoupling and coordination module is used to establish the coupling relationship of the three actuators, map the disturbance to the topological model disturbance branch, determine the pressure wave propagation delay based on the channel length, oil elastic modulus and oil temperature, and superimpose the feedforward compensation after phase alignment; it is distributed to four distributed sub-controllers to independently execute single-step local optimization, and rolling time-domain optimization is used to ensure that the calculation cycle matches the hydraulic system control cycle; The parameter evolution module is used to record the operation cycle flow deviation, trajectory error and pressure fluctuation, and to evolve the probability density of the parameter space by triggering the update of the mapping weight coefficient based on the performance evaluation. The storage module is used to store the operating condition knowledge base and mapping weights.

[0075] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for decoupling and collaboratively optimizing the operation actions of a tracked excavator in complex terrain, characterized in that, The specific steps are as follows: S1. Multi-source sensing field construction: Sensors are set up on the chassis and working device to collect slope, cylinder pressure and displacement, main pump flow and oil temperature. An extended state observer is constructed to identify oil temperature drift, component wear and pipeline elastic deformation nonlinear disturbances. A hydraulic network topology impedance model is established to map the pump source, valve port and chamber as node channels. The channel impedance and flow distribution are initialized to form the initial distribution of the working condition-control parameter mapping weight coefficient. S2. Active traffic arbitration: Based on the data from step S1, the expected traffic demand is determined according to the actuator speed and load. The weighted interpolation of the mapping weight coefficients of adjacent operating condition clusters is used to determine the continuous weights, and the optimal traffic allocation path is reconstructed. S3. Decoupling and Coordination Optimization: According to the instructions in step S2, establish the coupling relationship of the three actuators, map the interference into the topological model disturbance branch, determine the pressure wave propagation delay based on the channel length, oil elastic modulus and oil temperature, and superimpose the feedforward compensation amount after phase alignment. The optimization is assigned to four distributed sub-controllers to perform single-step local optimization independently, and rolling time-domain optimization is used to ensure that the calculation cycle matches the hydraulic system control cycle. S4. Closed-loop evolution of mapping weight coefficients: Record the operation cycle flow deviation, trajectory tracking error and pressure fluctuation, construct a multi-dimensional comprehensive performance evaluation function, and trigger the update of mapping weight coefficients when the comprehensive performance score is lower than the preset proportion of the historical best score, evolve the probability density of the parameter space, and write the updated parameters into storage to form a knowledge base.

2. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 1, characterized in that, In step S1, the specific steps for multi-source sensing field establishment are as follows: S1.1 Multi-source sensing topology model construction: Sensors are deployed on the chassis and working device to collect data on slope, support roller reaction force, cylinder pressure and displacement, pump flow and oil temperature. An extended state observer is constructed to identify oil temperature drift, component wear and nonlinear disturbances of pipeline elastic deformation online, output terrain feature parameters and load values, establish a hydraulic network topology impedance model, and map the pump source outlet, valve port and chamber as node channels. The nodes include source nodes, intermediate nodes and terminal nodes. The channel impedance includes liquid resistance, liquid capacity and liquid sensing. The channel impedance and flow distribution are initialized. S1.2 Initialization of Working Condition Mapping Weight Coefficients: Historical flow distribution is associated with terrain feature parameters to form an initial distribution of mapping weight coefficients. A closed-chain detection layer is added to the topological impedance model. The contact state is inferred based on the bucket attitude angle and pressure changes. It is determined whether the boom, stick, and bucket form a kinematic closed chain. The closed-chain or open-chain mode is marked. The terrain load parameter space is constructed with slope, ground stiffness, and subsidence as coordinate axes, so that the mapping weight coefficients are stored according to the probability density distribution in the space.

3. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 2, characterized in that, In step S2, the specific steps of the traffic active arbitration are as follows: S2.1 Mode Switching and Flow Demand Prediction: Receive the S1 topological impedance model status, mapping weight coefficient distribution and mode label. In open chain mode, establish a flow demand prediction model based on the estimated speed, acceleration and load of the boom, stick, bucket, and slewing mechanism, and calculate the expected flow demand of each actuator. In closed chain mode, introduce a fixed constraint on the tooth tip position, and jointly solve the flow demand of the three actuators into the minimum energy consumption coordination control quantity that satisfies the closed chain constraint. S2.2 Mapping Weight Coefficient Interpolation and Path Reconstruction: Based on the coordinates of the current working point in the terrain load parameter space, the mapping weight coefficient distribution of the neighboring working clusters is weighted and interpolated to generate continuous flow allocation weights. This gives priority to paths with high mapping weight coefficient values ​​and those that match the terrain load. The optimal flow allocation path is reconstructed in the topology impedance model, and active flow allocation instructions for each actuator are generated.

4. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 3, characterized in that, In step S2.1, the specific steps for mode switching and traffic demand prediction are as follows: S2.11 Open Chain Flow Demand Prediction: Receive the S1 topology impedance model status, mapping weight coefficient distribution and mode label. When the topology impedance model is in open chain mode, based on the estimated values ​​of the speed, acceleration and load of the boom, stick, bucket and slewing mechanism, combined with terrain feature parameters, establish a multi-actuator joint flow demand prediction model, calculate the expected flow demand of each actuator in the next control cycle, and form the independent flow allocation benchmark value of each actuator in open chain mode. S2.12, Closed-chain constraint coordination solution: When the topological impedance model is in closed-chain mode, the fixed constraint of the tooth tip position is introduced based on the closed-chain detection layer mark, and the kinematic closed-chain constraint equations of the boom, stick and bucket three actuators are established. The flow demand of the three actuators is incorporated into the joint optimization framework, and the minimum energy consumption coordination control quantity that satisfies the closed-chain constraint is solved to form the benchmark value of the coupled flow distribution of each actuator in closed-chain mode.

5. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 4, characterized in that, In step S2.2, the specific steps for mapping weight coefficient interpolation and path reconstruction are as follows: S2.21, Weighted Interpolation of Mapping Weight Coefficients and Weight Generation: Receives the traffic allocation baseline value output from S2.1, and performs weighted interpolation on the distribution of mapping weight coefficients of neighboring working condition clusters based on the coordinates of the current working point in the terrain load joint parameter space. This generates continuously changing traffic allocation weights, giving priority to paths with high mapping weight coefficient values ​​and matching terrain load, and determining the set of candidate paths for traffic allocation of each actuator, providing a screening basis for path reconstruction. S2.22 Optimal Path Reconstruction and Command Generation: In the topological impedance model, the optimal flow distribution path is screened and reconstructed from the candidate path set according to the selection probability. The path is optimized by combining the impedance of each channel. The reconstruction result is converted into active flow distribution commands for the boom, stick, bucket and slewing mechanism and output to step S3.

6. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 5, characterized in that, In step S3, the specific steps of decoupling and collaborative optimization are as follows: S3.1 Coupled Disturbance Mapping and Wavefront Compensation: Receive the S2 active flow distribution command, establish a kinematic and dynamic coupling model of the three actuators (boom, stick, and bucket), identify lever arm changes, load offsets, and attitude feedback disturbances in real time, map the coupled disturbances into topological model disturbance branches, calculate the pressure wavefront propagation delay based on channel length, oil elastic modulus, and oil temperature, and then align the feedforward compensation amount with the phase of the delay and superimpose it onto the control command. S3.2 Distributed Cooperative Optimization and Feedback Correction: The topology model reconstruction and compensation are decomposed into four distributed sub-controllers. Each sub-controller independently performs single-step local optimization based on local sensor information. The calculation cycle of the single-step local optimization is no greater than the hydraulic system control cycle. The topology model state and compensation parameters are shared through a high-speed bus. Rolling time-domain optimization is used to ensure that the calculation cycle matches the hydraulic system control cycle. A feedback correction link is introduced to correct the feedforward compensation amount in real time, generating precise control commands for each actuator after decoupling.

7. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 6, characterized in that, In step S3.1, the specific steps for coupled perturbation mapping and wavefront compensation are as follows: S3.11, Coupled Model Construction and Disturbance Tributary Mapping: Receive the S2 active flow allocation command, establish a kinematic and dynamic coupling model between the three actuators of boom, stick, and bucket, identify in real time the feedback disturbances caused by boom extension, stick effective lever arm change, boom load distribution offset caused by stick retraction, and bucket attitude angle change, quantify the coupling disturbances into the interference intensity of each tributary on the main channel in the topology model, and establish the disturbance transmission topology; S3.12, Wavefront Delay Calculation and Phase Alignment Compensation: Based on the length of each channel in the topology model, the elastic modulus of the oil, and the oil temperature, the pressure wavefront propagation delay is calculated in real time. The feedforward compensation is then phase-aligned according to this delay and superimposed onto the control command so that the compensation and disturbance arrive synchronously at the target actuator.

8. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 7, characterized in that, In step S3.2, the specific steps of distributed collaborative optimization and feedback correction are as follows: S3.21, Node Decomposition and Rolling Optimization: The topology model reconstruction and compensation are decomposed into four distributed sub-controllers. Each sub-controller independently performs single-step local optimization based on local sensor information. The topology model state and compensation parameters are shared through a high-speed bus. Rolling time-domain optimization is used to transform long-term time-domain prediction into a local single-step optimization and a central coordinator constraint check iterative structure. The central coordinator is responsible for task synchronization and hard constraint arbitration. S3.22 Feedback Correction and Command Generation: Feedback correction based on the deviation between actual displacement and target trajectory is introduced to correct the feedforward compensation amount after phase alignment in step S3.1 in real time, eliminate residual deviation and unmodeled disturbance, and generate the final decoupled precise control commands for each actuator based on the correction results.

9. The method for decoupling and coordinating optimization control of tracked excavator operation actions in complex terrain according to claim 8, characterized in that, In step S4, the specific steps of the closed-loop evolution of the mapping weight coefficients are as follows: S4.1 Operation Data Recording and Performance Decay Triggering: During the continuous operation of steps S1 to S3, the data recording unit collects and stores the flow distribution deviation, trajectory tracking error, pressure fluctuation variance and valve action frequency of each operation cycle, constructs a multi-dimensional comprehensive performance evaluation function, and monitors the control performance decay under complex terrain conditions in real time based on this function. When the evaluation index is lower than the preset ratio of the historical best performance, the mapping weight coefficient update mechanism is automatically triggered, and the parameter space evolution process is started. S4.2 Evolution of Mapping Weight Coefficient Kernel Function and Knowledge Base Evolution: The probability density of the mapping weight coefficients at corresponding coordinate points in the terrain load joint parameter space is evolved using a kernel function smoothing method to make the distribution of mapping weight coefficients form a smooth and continuous manifold. Based on the evaluation results, the mapping weight coefficients of the flow allocation path in step S2 are enhanced or attenuated, and the mapping weight coefficients of the decoupling compensation path in step S3 are modified in a linked manner. The updated topology model parameters and terrain features are associated and written into a non-volatile storage unit to form a working condition knowledge base.

10. A decoupled collaborative optimization control system for the operation of a tracked excavator in complex terrain, characterized in that, The complex terrain tracked excavator operation action decoupling and collaborative optimization control system is used to execute the complex terrain tracked excavator operation action decoupling and collaborative optimization control method as described in any of claims 1 to 9. The system includes: a sensing and site establishment module, a flow arbitration module, a decoupling and collaboration module, a parameter evolution module, and a storage module. The sensing field module is used to collect chassis slope, cylinder pressure and displacement, main pump flow and oil temperature, build an extended state observer to identify nonlinear disturbances, establish a hydraulic network topology impedance model to map pump source, valve port and chamber as node channels, initialize channel impedance and flow distribution, and form the initial distribution of mapping weight coefficients. The flow arbitration module is used to determine the expected flow demand based on the actuator speed and load, determine the continuous weight by weighted interpolation of the mapping weight coefficients of adjacent operating condition clusters, and reconstruct the optimal flow allocation path. The decoupling and coordination module is used to establish the coupling relationship of the three actuators, map the disturbance to the topological model disturbance branch, determine the pressure wave propagation delay based on the channel length, oil elastic modulus and oil temperature, and superimpose the feedforward compensation after phase alignment; it is distributed to four distributed sub-controllers to independently execute single-step local optimization, and rolling time-domain optimization is used to ensure that the calculation cycle matches the hydraulic system control cycle; The parameter evolution module is used to record the operation cycle flow deviation, trajectory error and pressure fluctuation, and to evolve the probability density of the parameter space by triggering the update of the mapping weight coefficient based on the performance evaluation. The storage module is used to store the operating condition knowledge base and mapping weights.

Images