A high-speed heavy-load robot scale-structure-drive collaborative optimization design method

Abstract

The present application relates to the technical field of robot optimization design, and more particularly to a high-speed heavy-load robot scale-structure-driving collaborative optimization design method, comprising establishing a coupling optimization model; constructing a trajectory-driven performance quantification interface function, constructing a parameterized performance modeling module, a stiffness proxy model and a binary gating function, and obtaining a compact candidate driving set through type spectrum screening; an outer loop generates a structure-scale candidate scheme and performs rapid feasibility screening; an inner loop iteratively solves the minimum cycle time and demand envelope of the candidate scheme under the trajectory-driven performance quantification interface function on the compact candidate driving set, and performs trajectory-driving closed-loop iteration on the compact candidate driving set to obtain an optimal driving selection; and the outer loop updates the Pareto optimal solution based on a multi-objective optimization function to obtain a multi-objective optimization solution set. The present application solves the problems of low iteration efficiency, combinatorial explosion and performance post-processing of traditional serial design.

Description

Technical Field

[0001] This invention relates to the field of robot optimization design technology, and in particular to a method for collaborative optimization design of scale, structure, and drive of high-speed heavy-duty robots. Background Technology

[0002] High-speed heavy-duty robots face multiple challenges when performing cyclical handling tasks, including short cycle times, high load weights, and frequent operations. Therefore, they have extremely high requirements for structural rigidity, driving capability, and motion performance.

[0003] Traditional robot design methods typically employ a "serial design" approach: first, preliminary designs of dimensional structures and components are completed based on experience or simple statics; then, drive selection is performed based on the design results; and finally, cycle time is checked through trajectory planning to ensure it meets requirements. This approach has the following inherent drawbacks: (1) The scale-structure design and drive selection are separated, and the coupling relationship between drive capability and structural performance is not fully utilized; (2) The trajectory planning stage passively adapts to the determined ontological parameters and driving configuration, making it difficult to achieve pre-optimization of cycle time; (3) The problem of space explosion in design is prominent: the scale parameters, structural parameters and discrete drive selection combinations are huge, the cost of high-fidelity simulation is too high, which leads to serious reliance on experience trial and error in actual engineering, making it difficult to obtain the global optimal design.

[0004] To address the aforementioned issues, existing research has attempted to employ collaborative optimization methods, jointly solving structural optimization and trajectory optimization. However, current collaborative design methods are mostly geared towards continuous design variables, making it difficult to handle discrete decision-making problems such as drive selection; moreover, they often employ single-objective optimization, failing to simultaneously consider the comprehensive optimization of multi-dimensional engineering objectives such as cycle time, overall machine mass, and end-effector stiffness. Furthermore, existing methods lack effective reusable modeling and search space compression mechanisms when facing the dual bottlenecks of high-fidelity evaluation costs and the expansion of discrete combination scale.

[0005] Therefore, there is an urgent need to develop an integrated collaborative design method for robots for high-speed, heavy-load cyclic tasks, which can incorporate scale-structure parameters and discrete drive selection into a unified optimization framework to achieve multi-objective collaborative optimization with controllable computational cost. Summary of the Invention

[0006] This invention aims to solve at least one of the technical problems existing in related technologies. To this end, this invention provides a scale-structure-drive collaborative optimization design method for high-speed heavy-duty robots, which realizes pre-quantization of cycle time, compression of discrete drive combination space, low-cost screening of structural feasibility, and deep coupling optimization of body-drive-trajectory, solving the problems of low iteration efficiency, combinatorial explosion, and a posteriori performance in traditional serial design.

[0007] This invention provides a method for co-optimization design of scale, structure, and drive for high-speed, heavy-duty robots, comprising: S1: Construct a coupled optimization model, which includes optimization variables, constraints, and a multi-objective optimization function. The coupled optimization model aims to optimize the multi-objective optimization function by adjusting the optimization variables while satisfying the constraints. S2: Construct a trajectory-driven performance quantization interface function based on optimization variables, constraints, and multi-objective optimization functions; S3: Construct a parameterized performance modeling module based on the optimization variables. Construct a stiffness surrogate model based on the parameterized performance modeling module through simulation and Gaussian process regression. Construct a binary gating function based on the stiffness surrogate model. Generate the demand envelope of each joint through the performance quantization interface function of trajectory drive. Filter the drive spectrum through the demand envelope of each joint to obtain a compact candidate drive set. S4: The outer loop generates scale-structure candidate solutions through a multi-objective optimization algorithm, and performs rapid feasibility screening on the scale-structure candidate solutions through a binary gating function to obtain a set of feasible candidate solutions and a set of re-evaluation and certification solutions. The inner loop evaluates the set of feasible candidate solutions on a compact candidate driving set to obtain a set of feasible candidate solution solutions. The re-evaluation and certification set is evaluated with high fidelity and the inner loop is executed to obtain a set of certified candidate solution solutions. The set of feasible candidate solution solutions and the set of certified candidate solution solutions are merged to form a multi-objective optimization solution set.

[0008] Furthermore, the requirement envelope for each joint is generated through the trajectory-driven performance quantization interface function, including: S21: Initialize the particle swarm optimization algorithm based on the optimization variables, constraints, and multi-objective optimization function, and set the number of iterations according to the time interval; S22: B-splines are used to parameterize the joint trajectory, and the trajectory parameters are optimized using a particle swarm optimization algorithm; S23: Set smoothness constraints and determine whether the trajectory parameters violate the unified kinematics and smoothness constraints. If they do not comply with the unified kinematics and smoothness constraints, repeat step S22. If they comply with the unified kinematics and smoothness constraints, solve the requirement envelope through the trajectory-driven performance quantization interface function and obtain the minimum cycle time of the current trajectory parameters. S24: Repeat steps S22 and S23 based on the number of iterations to obtain the optimal trajectory parameters, minimum cycle time, and required envelope under the current drive.

[0009] Furthermore, constructing the stiffness proxy model includes: S311: Construct a CAD-FEA parametric joint simulation link based on the parametric performance modeling module. The CAD-FEA parametric joint simulation link is used to automatically solve the scale-structure parameters and end deformation response. S312: Latin hypercube sampling is performed on the scale-structure parameters in the design space to generate a scale-structure parameter sample set. Simulation calculations are performed on each sample through the CAD-FEA parametric joint simulation link to obtain the end deformation dataset. S313: Perform Gaussian process regression on the end deformation dataset to obtain a stiffness surrogate model.

[0010] Furthermore, the drive spectrum is filtered based on the demand envelope of each joint to obtain a compact candidate drive set, including: S321: Select reference parameters and generate the requirement envelope of each joint under the reference parameters through the performance quantization interface function of trajectory driving; S322: Apply an engineering margin to the requirement envelope of each joint to obtain a conservative requirement envelope; S323: Based on conservative requirements, drive selection is performed in the original drive library of the joint through type spectrum screening to obtain a compact candidate drive set; The model selection process compares the requirement envelope of each joint with the torque, speed, power, reduction ratio, and safety margin of the drive in the original drive library, deletes drives that do not meet the requirements, and obtains a compact set of candidate drives.

[0011] Furthermore, in step S4, the working steps of the outer loop include: S411: Initialize the outer loop multi-objective optimization algorithm and generate scale-structure candidate schemes through the outer loop multi-objective optimization algorithm; S412: Use a binary gate function to screen the structural feasibility of scale-structure candidate schemes to obtain a set of feasible candidate schemes and a set of re-evaluation and certification schemes; S413: Perform an inner loop evaluation on a compact candidate driver set for the candidate solutions in the feasible candidate solution set to obtain the minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions; if the driver configuration of the candidate solutions is updated, then re-execute the inner loop evaluation under the updated driver configuration; if the driver configuration of the candidate solutions remains unchanged, then output the minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions. S414: The minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions are fed back to the outer loop; S415: Repeat steps S413 and S414 to complete the evaluation of all candidate solutions in the feasibility candidate solution set. The minimum cycle time, optimal demand envelope, and optimal driving selection of all candidate solutions form the feasibility candidate solution solution set. S416: Conduct high-fidelity finite element certification on all candidate schemes in the re-evaluation and certification set to obtain high-fidelity values ​​of the candidate schemes. Based on the high-fidelity values ​​of the candidate schemes, screen the candidate schemes in the re-evaluation and certification set to obtain candidate schemes that pass certification. Perform an inner loop evaluation on the candidate schemes that pass certification to obtain a solution set of certified candidate schemes. S417: Merge the solution sets of feasible candidate solutions and the solution sets of certified candidate solutions to form a multi-objective optimization solution set.

[0012] Furthermore, in step S4, the working steps of the inner loop include: S421: Iteratively solve the minimum cycle time and optimal requirement envelope of candidate solutions using a trajectory-driven performance quantization interface function; S422: Perform driver matching in a compact set of candidate drivers based on the optimal requirement envelope to obtain the optimal driver selection.

[0013] Furthermore, the performance quantization interface function for trajectory-driven operation is as follows: in, For trajectory-driven performance quantization interface, For periodic time, for The torque during the cycle of time, for The rotational speed during the time cycle, For the demand envelope, For scale structure decision variables, For trajectory planning related parameters, This is the output.

[0014] Furthermore, the optimization variables include scale variables, structural variables, and drive selection variables. The scale variables are used to characterize the geometric dimensions of the connecting rod length; the structural variables are used to characterize the key cross-sectional dimensions and plate thickness; and the drive selection variables are used to characterize the selection scheme of the motor-reducer. The constraints include structural constraints, driving capability constraints, and engineering constraints. The multi-objective optimization function aims at minimizing cycle time, overall machine mass, and end-effector deformation.

[0015] Furthermore, the parametric performance modeling module in step S3 includes rod inertia parameter mapping, kinematic model, and dynamic model.

[0016] Furthermore, the demand envelope includes the maximum operating torque, the root mean square value of the torque, and the maximum speed within one operating cycle.

[0017] The above-described one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: This invention breaks away from the traditional serial design paradigm and achieves integrated collaborative optimization of scale, structure, and drive. By constructing a "two-stage + inner and outer double loop" framework, discrete drive selection variables and continuous scale-structure variables are incorporated into a unified system-level optimization model. The outer loop is responsible for the global search of the scale-structure space, while the inner loop realizes the joint optimal matching of trajectory parameters and drive configuration. This fundamentally solves the problem of the separation between ontology design and drive selection in traditional methods.

[0018] This invention achieves pre-quantization and active optimization of cycle time, reducing the minimum cycle time. As an optimizable system-level objective, the closed-loop iteration of "trajectory update → Eval evaluation → drive matching" can proactively compress cycle time during the scale-structure design stage, overcoming the limitations of traditional methods where trajectory planning passively adapts to existing design parameters.

[0019] This invention significantly reduces the cost of high-fidelity evaluation and the scale of discrete combinatorial search. It constructs a parameterized performance modeling module and a Gaussian process stiffness surrogate model, achieving low-cost pre-screening of structural feasibility through a binary gating function. Simultaneously, it compresses the original driver library into a compact candidate set through pattern filtering, reducing the scale of discrete combinatorial search from the source. Delayed authentication and batch high-fidelity re-evaluation mechanisms further concentrate expensive simulation calls on the most promising candidate solutions, achieving efficient allocation of computational resources.

[0020] This invention supports multi-objective engineering trade-offs and global optimal design. It constructs a multi-dimensional objective vector with cycle time, overall machine mass, and end deformation, and uses a multi-objective optimization algorithm to drive the outer loop Pareto search. The final solution set fully presents the set of optimal design schemes under different trade-off preferences, providing a quantitative basis for engineering design decisions with a global perspective.

[0021] This invention possesses excellent scalability and engineering adaptability. The trajectory-driven performance quantization interface, parameterized performance modeling module, and binary gating function of this invention are all reusable modular designs, capable of adapting to the design requirements of high-speed, heavy-duty robots with different configurations and task scenarios, and have broad industrial application prospects.

[0022] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating a scale-structure-drive collaborative optimization design method for a high-speed, heavy-duty robot provided by the present invention.

[0025] Figure 2 This is a schematic diagram of a three-dimensional model of a high-speed heavy-duty robot according to an embodiment of the present invention.

[0026] Figure 3 These are the optimized rod design variables in the embodiments of the present invention.

[0027] Figure 4 This is the multi-objective optimization solution set in the embodiments of the present invention.

[0028] Figure 5 This is a scatter plot of the Pareto optimal solution set obtained by solving the multi-objective optimization problem of the robot in this embodiment of the invention, wherein... Figure 5 Figure (a) in the figure is a scatter plot of the mass-stiffness correlation. Figure 5 Figure (b) in the figure is a scatter plot of the mass-cycle time correlation. Figure 5 Figure (c) is a scatter plot of stiffness-period-time correlation.

[0029] Figure 6 This is a torque diagram of the robot's cyclic operation manufactured using the present invention, wherein... Figure 6 Figure (a) shows the real-time torque variation curve of the active locating joint D1 during one complete motion cycle. Figure 6 Figure (b) shows the real-time torque variation curve of the second active rotary joint R2 during one complete motion cycle. Figure 6 Figure (c) shows the real-time torque variation curve of the third active rotary joint R3 during one complete motion cycle.

[0030] Figure 7 This is a schematic diagram of the outer loop solution of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but cannot be used to limit the scope of this invention.

[0032] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0033] The following is combined Figures 1 to 7 This invention describes a scale-structure-drive collaborative optimization design method for a high-speed, heavy-duty robot.

[0034] like Figure 1 As shown, a scale-structure-drive collaborative optimization design method for high-speed heavy-duty robots includes: S1: Construct a coupled optimization model, which includes optimization variables, constraints, and a multi-objective optimization function. The coupled optimization model aims to optimize the multi-objective optimization function by adjusting the optimization variables while satisfying the constraints. Optimization variables include scaling variables Structural variables Driving selection variables Scale variables are used to characterize the geometric dimensions of the connecting rod length; structural variables are used to characterize key cross-sectional dimensions, plate thickness, etc.; drive selection variables are used to characterize the selection scheme of motor-reducer. The robot body collaborative design parameters are divided into: , as well as ,in , Design space for scale parameters, , For the design space of structural parameters, , To drive the product library.

[0035] Define scale structure decision variables , System design decision variables are defined as follows: .

[0036] The constraints include structural constraints, driving capability constraints, and engineering constraints, and the calculation expression is as follows: in, For constraint function, For trajectory planning related parameters, These are engineering constraints related to geometric boundaries, assembly space, and manufacturability. This is the upper limit of the engineering constraints. In the first Various working conditions The end deformation index when a fixed reference load is applied. The maximum allowable deformation threshold, This is a set of task-level requirement envelope metrics. To drive selection variables Capability boundaries For each element to be unequal.

[0037] Multi-objective optimization function with minimum periodic time Overall machine quality and end deformation For the objective, the calculation expression is: in, For multi-objective optimization functions, , For the feasible region of trajectory parameters, Scale structure decision variables Trajectory planning related parameters The cycle time, This is the optimal driver corresponding to the scale-structure parameters.

[0038] Coupled optimization model: Under the premise of satisfying constraints, the optimization model adjusts the optimization variables to make the multi-objective optimization function optimal. It is used to uniformly describe the multivariate optimization relationship between scale variables, structural variables and driving configuration.

[0039] S2: Construct a trajectory-driven performance quantization interface function based on optimization variables, constraints, and multi-objective optimization functions; The performance quantization interface function for trajectory-driven systems is as follows: in, For trajectory-driven performance quantization interface, For periodic time, for The torque during the cycle of time, for The rotational speed during the time cycle, For the demand envelope, For scale structure decision variables, For trajectory planning related parameters, This is the output.

[0040] The trajectory-driven performance quantization interface encapsulates the trajectory planning and dynamics evaluation modules. It generates the requirement envelope for each joint through the trajectory-driven performance quantization interface functions, including: S21: Initialize the Particle Swarm Optimization (PSO) algorithm based on the optimization variables, constraints, and multi-objective optimization function, and set the number of iterations according to the time interval; Initialize PSO at time intervals For time-based decision variables, ,in For the first time interval, For the first The number of time intervals is used as the number of iterations.

[0041] S22: B-splines are used to parameterize the joint trajectory, and the trajectory parameters are optimized using a particle swarm optimization algorithm; right ,structure , ,in The shape parameters, determined by the critical path points and boundary conditions, are parameterized using B-splines. The calculation expression is as follows: in, node vector B-spline curve, To control the vertices, For node vectors No. The first control vertex B-order spline basis functions, To control the number of vertices; B-spline curves The first derivative is: in, for of First derivative, for of The first derivative corresponding to the second derivative One control vertex, node vector No. The first control vertex B-order spline basis functions; Given by the recurrence relation: in, B-spline curve The first derivative One control vertex, B-spline curve The first derivative One control vertex, B-spline curve The first derivative One control vertex, The first node vector One control vertex, The first node vector One control vertex, for The 0th derivative corresponds to the 1st derivative. One control vertex.

[0042] S23: Set smoothness constraints and determine whether the trajectory parameters violate the unified kinematics and smoothness constraints. If they do not comply with the unified kinematics and smoothness constraints, repeat step S22. If they comply with the unified kinematics and smoothness constraints, solve the optimal demand envelope through the trajectory-driven performance quantization interface function, and also obtain the minimum cycle time of the current trajectory parameters. S24: Repeat steps S22 and S23 based on the number of iterations to obtain the optimal trajectory parameters, minimum cycle time, and required envelope under the current drive.

[0043] S3: Construct a parameterized performance modeling module based on the optimization variables. Construct a stiffness surrogate model based on the parameterized performance modeling module through simulation and Gaussian process regression. Construct a binary gating function based on the stiffness surrogate model. Generate the demand envelope of each joint through the performance quantization interface function of trajectory drive. Filter the drive spectrum through the demand envelope of each joint to obtain a compact candidate drive set. Parametric performance modeling module This includes the mapping of inertial parameters of the rods, kinematic models, and dynamic models.

[0044] Member inertia parameter mapping : Regarding the first Member parameter subset Response surface methodology is used for fitting, for arbitrary mass Center of mass Component or inertia component The expression for calculating the response surface is: in, For the predicted output of the response surface model, As the reference offset, The coefficient of the linear term, for The One design variable, for The One design variable, for The One design variable, The coefficient of the quadratic term, The coefficient of the cubic term; Therefore, quality Center of mass Components and inertia components The response surface models can be expressed as follows: in, For the first The quality of the rod, For the first The predicted output of the mass response surface model of the rod. For the first The center of mass of the rod, For the first The predicted output of the center-of-mass response surface model of the bar. For the first The moment of inertia of the rod, For the first Predicted output of the inertia response surface model of the rod.

[0045] Kinematic model Let the joint variable be... , The end pose is The expressions for calculating forward / inverse kinematics are: in, For positive kinematic functions, It is the inverse kinematic function; Dynamics model Constructing the standard form of joint space dynamics based on the Lagrange method: in, Torque / force The inertia matrix, For the Coriolis force and centrifugal force terms, This is the term related to gravity. for The second derivative, for The first derivative; The parameterized performance modeling module is as follows: in, It is a mapping operator.

[0046] Constructing a stiffness proxy model includes: S311: Construct a CAD-FEA parametric joint simulation link based on the parametric performance modeling module. The CAD-FEA parametric joint simulation link automatically solves the scale-structure parameters and end deformation response. S312: Latin hypercube sampling is performed on the scale-structure parameters in the design space to generate a scale-structure parameter sample set. Simulation calculations are performed on each sample through the CAD-FEA parametric joint simulation link to obtain the end deformation dataset. S313: Perform Gaussian process regression on the end-deformation dataset to obtain the stiffness surrogate model. The calculation expression is as follows: in, For end deformation response, For end-deformation datasets, To predict the mean, This represents the predicted standard deviation.

[0047] Constructing a binary gate function : in, For the maximum end deformation response, This is a conservative coefficient used to incorporate uncertainty into the decision-making process and mitigate the risk of erroneous releases. This gate function is used for pre-loop screening: only when At that time, candidates Proceed to the subsequent internal circulation evaluation.

[0048] By filtering the demand envelopes of each joint, a compact candidate driver set was obtained, including: S321: Select reference parameters The performance quantization interface function driven by trajectory generates the requirement envelope of each joint under the reference parameters. ; Select reference parameters composed of extreme value combinations or the most unfavorable operating conditions in the design space. , The calculation expression is: in, For joints The peak torque corresponding to the reference parameters. joint The corresponding root mean square torque under the reference parameters. For joints The corresponding peak speed under the reference parameters.

[0049] S322: Apply engineering margins to the requirement envelope of each joint to obtain conservative requirements. ; S323: Based on the conservative requirement envelope, candidate drivers are screened in the original driver library of the joint through pattern screening to obtain a compact candidate driver set; Conservative demand Below, joints are constructed according to a unified transmission conversion. feasible driver set For motor-reducer combination joints, reducer candidates are selected based on the torque-speed requirements on the joint side, and motors are selected based on the reduction ratio requirements converted to the motor side, retaining combinations where "the reducer is feasible and a matching motor exists".

[0050] The pattern selection involves matching the demand envelope of each joint with the drive product library. The torque, speed, power, reduction ratio, and safety margin of the drive are compared, and drives that do not meet the requirements are removed to obtain a compact set of candidate drives.

[0051] By summing up the feasible sets of each joint, we obtain a system-level compact candidate driver set: in, for A compact set of candidate drivers This represents the set of feasible drives for the first joint. For the first The set of feasible actuations for the joint. For a compact candidate driver set.

[0052] S4: The outer loop generates scale-structure candidate solutions through a multi-objective optimization algorithm, and performs rapid feasibility screening on the scale-structure candidate solutions through a binary gating function to obtain a set of feasible candidate solutions and a set of re-evaluation and certification solutions. The inner loop evaluates the set of feasible candidate solutions on a compact candidate driving set to obtain a set of feasible candidate solution solutions. The re-evaluation and certification set is evaluated with high fidelity and the inner loop is executed to obtain a set of certified candidate solution solutions. The set of feasible candidate solution solutions and the set of certified candidate solution solutions are merged to form a multi-objective optimization solution set.

[0053] like Figure 7 As shown, the working steps of the outer loop include: S411: Initialize the outer loop multi-objective optimization algorithm (NSGA-II) to generate scale-structure candidate solutions; Define the design variable boundaries, population size, number of iterations, and optimization objectives; S412: Use a binary gate function to screen the structural feasibility of scale-structure candidate schemes to obtain a set of feasible candidate schemes and a set of re-evaluation and certification schemes; like Then the aforementioned structure-scale candidate schemes are directly discarded; like Then, a confidence level assessment is performed on the aforementioned structure-scale candidate schemes; like Less than or equal to When the time is right, the structure-scale candidate scheme becomes a feasible candidate scheme and enters the inner loop; like Greater than If so, the structure-scale candidate schemes will be included in the review queue for delayed certification; in, Candidates for structure-scale analysis. The standard deviation of the structure-scale prediction. This is the standard deviation threshold.

[0054] S413: Perform an inner loop evaluation on a compact candidate driver set for the candidate solutions in the feasible candidate solution set to obtain the minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions; if the driver configuration of the candidate solutions is updated, then re-execute the inner loop evaluation under the updated driver configuration; if the driver configuration of the candidate solutions remains unchanged, then output the minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions. S414: The minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions are fed back to the outer loop; S415: Repeat steps S413 and S414 to complete the evaluation of all candidate solutions in the feasibility candidate solution set. The minimum cycle time, optimal demand envelope, and optimal driving selection of all candidate solutions form the feasibility candidate solution solution set. The current population is updated by an outer loop multi-objective optimization algorithm, and the next generation of scale-structure candidate solutions is generated. Steps S413 and S414 are repeated until the maximum number of iterations is reached to obtain a set of feasible candidate solutions. ; S416: Conduct high-fidelity finite element analysis on all candidate solutions in the re-evaluation and certification set to obtain high-fidelity values ​​for the candidate solutions. Based on these high-fidelity values, filter the candidate solutions in the re-evaluation and certification set to obtain certified candidate solutions. Perform an inner loop evaluation on the certified candidate solutions to obtain a solution set of certified candidate solutions. ; Set high fidelity threshold If the candidate solution has a high fidelity Less than or equal to the high-fidelity threshold If the candidate solution is positive, then the candidate solution is considered to be a certified candidate solution.

[0055] S417: Merge the solution sets of feasible candidate solutions and the solution sets of verified candidate solutions to form a multi-objective optimization solution set; in, To merge the solution sets.

[0056] Optimization schemes are selected from the final candidate solution set according to preset engineering criteria, and the optimal scale-structure parameters and driving configuration are output. The multi-objective optimization solution set is called the Pareto solution set.

[0057] The working steps of the inner loop include: S421: Iteratively solve the minimum cycle time and requirement envelope of feasible candidate solutions using a trajectory-driven performance quantization interface function; S422: Perform driver matching in a compact set of candidate drivers based on the optimal requirement envelope to obtain the optimal driver selection.

[0058] The expression for calculating the optimal demand envelope is: in, For maximum operating torque, This is the root mean square value of the torque. This is the maximum speed.

[0059] Driving selection variables The capability boundary vector is The calculation expression is: in, for Maximum operating torque under drive for The root mean square value of the torque under drive. for Maximum speed under drive.

[0060] Feasibility assessment uses element-by-element comparison At the same time, define the capacity constraint margin: in, This is a capacity constraint margin.

[0061] If and only if It was deemed feasible at that time.

[0062] This invention is based on a coupled optimization model and adopts an inner and outer dual-loop collaborative solution mechanism: the outer loop generates and iteratively updates the scale and structural parameters based on a multi-objective optimization algorithm; the inner loop performs a task-driven performance evaluation and driver matching process for each set of design variables generated by the outer loop, to calculate the objective function values ​​such as the running cycle time and joint requirement envelope, and outputs the matched driver selection.

[0063] Specifically, for any set of scale-structure candidate parameters generated by the outer loop, the inner loop first calculates the system operation process based on the task trajectory to obtain the corresponding operation cycle time and joint dynamic requirement information, and further extracts the joint requirement envelope; then, it compares the joint requirement envelope with the capability boundary of the drive system to complete the drive feasibility determination, and matches or updates the drive configuration.

[0064] The task evaluation results from the inner loop are further fed back into the parameter update process of the outer loop, guiding the evolution of scale and structural parameters in a direction that balances task speed performance, structural stiffness constraints, and drive adaptability. Simultaneously, structural feasible region constraints continuously influence candidate selection, eliminating invalid designs that do not meet geometric boundaries, stiffness conditions, and assembly limitations. This ensures that the optimization search remains confined within the feasible region and is used to obtain a Pareto solution set that satisfies multiple constraints.

[0065] This invention breaks away from the traditional serial design paradigm and achieves integrated collaborative optimization of scale, structure, and drive. By constructing a "two-stage + inner and outer double loop" framework, discrete drive selection variables and continuous scale-structure variables are incorporated into a unified system-level optimization model. The outer loop is responsible for the global search of the scale-structure space, while the inner loop realizes the joint optimal matching of trajectory parameters and drive configuration. This fundamentally solves the problem of the separation between ontology design and drive selection in traditional methods.

[0066] This invention achieves pre-quantization and active optimization of cycle time, reducing the minimum cycle time. As an optimizable system-level objective, the closed-loop iteration of "trajectory update → Eval evaluation → drive matching" can proactively compress cycle time during the scale-structure design stage, overcoming the limitations of traditional methods where trajectory planning passively adapts to existing designs.

[0067] This invention significantly reduces the cost of high-fidelity evaluation and the scale of discrete combinatorial search. It constructs a parameterized performance modeling module and a Gaussian process stiffness surrogate model, achieving low-cost pre-screening of structural feasibility through a binary gating function. Simultaneously, it compresses the original driver library into a compact candidate set through pattern filtering, reducing the scale of discrete combinatorial search from the source. Delayed authentication and batch high-fidelity re-evaluation mechanisms further concentrate expensive simulation calls on the most promising candidate solutions, achieving efficient allocation of computational resources.

[0068] This invention supports multi-objective engineering trade-offs and global optimal design. It constructs a multi-dimensional objective vector with cycle time, overall machine mass, and end deformation, and uses a multi-objective optimization algorithm to drive the outer loop Pareto search. The final solution set fully presents the set of optimal design schemes under different trade-off preferences, providing a quantitative basis for engineering design decisions with a global perspective.

[0069] This invention possesses excellent scalability and engineering adaptability. The trajectory-driven performance quantization interface, parameterized performance modeling module, and binary gating function of this invention are all reusable modular designs, capable of adapting to the design requirements of high-speed, heavy-duty robots with different configurations and task scenarios, and have broad industrial application prospects.

[0070] To verify the effectiveness of this invention, a high-speed feeding scenario on a stamping line was used, with an end load of 100 kg and a working space requirement of 2.7-2.9 m. A hybrid mechanism comprising an active moving joint D1, a second active rotating joint R2, and a third active rotating joint R3 was proposed. Figure 2 As shown, the design variables are as follows Figure 3 As shown. Includes the length L1 of rod 1, the wall thickness T1 of rod 1, the length L2 of rod 2, the wall thickness T2 of rod 2, the length L3 of rod 3, and the wall thickness T3 of rod 3.

[0071] Parametric modeling of the rod was performed, 30 sets of data were sampled, and a response surface model of mass, centroid, and inertia was constructed using a quadratic polynomial.

[0072] Using the maximum arm span posture and a load of 100 kg, a dataset was constructed by performing 300 finite element simulations based on Solidworks-Ansys, and a Gaussian process (GP) surrogate model was trained. A binary gating function was employed. Conduct a rapid feasibility screening.

[0073] Based on the performance quantization interface of trajectory drive and the extreme value calculation of design variables, torque / speed requirements are calculated and margins are applied to initially select feasible drive configurations from the mainstream industrial catalog.

[0074] The path includes 14 keypoints defined, symmetrically distributed. The optimization variables are set to 7 time intervals, and quintic B-spline parameterization is used. Under kinematic constraints, the objective is to minimize the period time. The solution is obtained using a particle swarm optimization algorithm with 150 particles, 50 iterations, and inertia weights. The learning factor is 2.

[0075] The NSGA-II algorithm was used, with a population size of 100, 50 generations, a crossover probability of 0.9, a mutation probability of 0.16, a crossover exponent of 20, and a mutation exponent of 20. Under the constraints of stiffness and end-load, the minimum cycle time and total mass were simultaneously optimized. .

[0076] The candidate solutions in the outer loop are first screened by GP gating. Those that pass the screening enter the inner loop to solve the minimum cycle time and are matched and driven. Low confidence samples are backfilled into the Pareto solution set after high-fidelity verification.

[0077] like Figure 4 The multi-objective optimization solution shown obtains the solution set for period time, mass, and stiffness. Figure 5 A scatter plot of the Pareto optimal solution set obtained from multi-objective optimization of the robot. Figure 5 Figure (a) in the figure is a scatter plot of the mass-stiffness correlation; Figure 5 Figure (b) in the figure is a scatter plot of the mass-cycle time correlation; Figure 5 Figure (c) in the diagram is a scatter plot showing the correlation between stiffness and cycle time. This Pareto optimal solution set provides a complete design space for the co-optimization of robot structure and performance. The optimal design scheme can be selected according to actual engineering requirements (such as load capacity, cycle time requirements, and lightweight indicators) to achieve a comprehensive optimal match of mass, stiffness, and cycle time.

[0078] Taking into account multiple engineering design criteria such as cycle time, structural stiffness, and drive margin, the optimal scheme was rounded to an engineering integer, and the final prototype design parameters were determined as follows: The length L1 of rod 1 is 1780 mm, and the thickness T1 is 6 mm; The length L2 of rod 2 is 560 mm, and the thickness T2 is 48 mm; The length L3 of rod 3 is 2200 mm, and the thickness T3 is 30 mm; Active moving joint D1 motor: weight 119 kg, allowable peak torque 2600 Nm; Second active rotary joint R2 servo motor: weight 37 kg, allowable peak torque 100 Nm; Second active rotary joint R2 reducer: mass 72 kg, allowable peak torque 13480 Nm; Third active rotary joint R3 servo motor: weight 37 kg, allowable peak torque 125 Nm; The third active rotary joint R3 reducer: weight 77 kg, allowable peak torque 3488 Nm.

[0079] The total mass of the moving parts in this scheme kg, under the premise of satisfying the end load of 100 kg and stiffness constraints, the cycle time can be achieved. The real-time torque variation curves of the robot's joint motors within one complete motion cycle are shown below. Figure 6 As shown, Figure 6 Figure (a) shows the real-time torque variation curve of the active locating joint D1 during one complete motion cycle. Figure 6 Figure (b) shows the real-time torque variation curve of the second active rotary joint R2 during one complete motion cycle. Figure 6 Figure (c) shows the real-time torque variation curve of the third active rotary joint R3 during one complete motion cycle. This verifies the dynamic performance and motor load rationality of the multi-objective optimization scheme, where the red dashed line represents the motor torque safety limit. Figure 6 The torque curves of the motors at each joint showed good synchronization, and the torques did not exceed the safety limits under extreme conditions. This proves that the proposed multi-objective optimization scheme can effectively ensure the safe and stable operation of the motors at each joint while meeting the requirements of robot motion rhythm, structural rigidity and lightweighting, thus providing dynamic support for reliable operation of the robot under high-speed and heavy-load conditions.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A scale-structure-drive collaborative optimization design method for high-speed heavy-duty robots, characterized in that, include: S1: Construct a coupled optimization model, which includes optimization variables, constraints, and a multi-objective optimization function. The coupled optimization model aims to optimize the multi-objective optimization function by adjusting the optimization variables while satisfying the constraints. S2: Construct a trajectory-driven performance quantization interface function based on optimization variables, constraints, and multi-objective optimization functions; S3: Construct a parameterized performance modeling module based on the optimization variables. Construct a stiffness surrogate model based on the parameterized performance modeling module through simulation and Gaussian process regression. Construct a binary gating function based on the stiffness surrogate model. Generate the demand envelope of each joint through the performance quantization interface function of trajectory drive. Filter the drive spectrum through the demand envelope of each joint to obtain a compact candidate drive set. S4: The outer loop generates scale-structure candidate solutions through a multi-objective optimization algorithm, and performs rapid feasibility screening on the scale-structure candidate solutions through a binary gating function to obtain a set of feasible candidate solutions and a set of re-evaluation and certification solutions. The inner loop evaluates the set of feasible candidate solutions on a compact candidate driving set to obtain a set of feasible candidate solution solutions. The re-evaluation and certification set is evaluated with high fidelity and the inner loop is executed to obtain a set of certified candidate solution solutions. The set of feasible candidate solution solutions and the set of certified candidate solution solutions are merged to form a multi-objective optimization solution set.

2. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, The requirement envelope for each joint is generated through the trajectory-driven performance quantization interface function, including: S21: Initialize the particle swarm optimization algorithm based on the optimization variables, constraints, and multi-objective optimization function, and set the number of iterations according to the time interval; S22: B-splines are used to parameterize the joint trajectory, and the trajectory parameters are optimized using a particle swarm optimization algorithm; S23: Set smoothness constraints and determine whether the trajectory parameters violate the unified kinematics and smoothness constraints. If they do not comply with the unified kinematics and smoothness constraints, repeat step S22. If they comply with the unified kinematics and smoothness constraints, solve the requirement envelope through the trajectory-driven performance quantization interface function, and also obtain the minimum cycle time of the current trajectory parameters. S24: Repeat steps S22 and S23 based on the number of iterations to obtain the optimal trajectory parameters, minimum cycle time, and required envelope under the current drive.

3. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, Constructing a stiffness proxy model includes: S311: Construct a CAD-FEA parametric joint simulation link based on the parametric performance modeling module. The CAD-FEA parametric joint simulation link is used to automatically solve the scale-structure parameters and end deformation response. S312: Latin hypercube sampling is performed on the scale-structure parameters in the design space to generate a scale-structure parameter sample set. Simulation calculations are performed on each sample through the CAD-FEA parametric joint simulation link to obtain the end deformation dataset. S313: Perform Gaussian process regression on the end deformation dataset to obtain a stiffness surrogate model.

4. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, The drive spectrum is filtered by the demand envelope of each joint to obtain a compact candidate drive set, including: S321: Select reference parameters and generate the requirement envelope of each joint under the reference parameters through the performance quantization interface function of trajectory driving; S322: Apply an engineering margin to the requirement envelope of each joint to obtain a conservative requirement envelope; S323: Based on the conservative requirement envelope, candidate drivers are screened in the original driver library of the joint through pattern screening to obtain a compact candidate driver set; The model selection process compares the requirement envelope of each joint with the torque, speed, power, reduction ratio, and safety margin of the drive in the original drive library, deletes drives that do not meet the requirements, and obtains a compact set of candidate drives.

5. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, In step S4, the outer loop's working steps include: S411: Initialize the outer loop multi-objective optimization algorithm and generate scale-structure candidate schemes through the outer loop multi-objective optimization algorithm; S412: Use a binary gate function to screen the structural feasibility of scale-structure candidate schemes to obtain a set of feasible candidate schemes and a set of re-evaluation and certification schemes; S413: Perform an inner loop evaluation on a compact candidate driver set for the candidate solutions in the feasible candidate solution set to obtain the minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions; if the driver configuration of the candidate solutions is updated, then re-execute the inner loop evaluation under the updated driver configuration; if the driver configuration of the candidate solutions remains unchanged, then output the minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions. S414: The minimum cycle time, optimal demand envelope, and optimal driver selection of the candidate solutions are fed back to the outer loop; S415: Repeat steps S413 and S414 to complete the evaluation of all candidate solutions in the feasibility candidate solution set. The minimum cycle time, optimal demand envelope, and optimal driving selection of all candidate solutions form the feasibility candidate solution solution set. S416: Conduct high-fidelity finite element certification on all candidate schemes in the re-evaluation and certification set to obtain high-fidelity values ​​of the candidate schemes. Based on the high-fidelity values ​​of the candidate schemes, screen the candidate schemes in the re-evaluation and certification set to obtain candidate schemes that pass certification. Perform an inner loop evaluation on the candidate schemes that pass certification to obtain a solution set of certified candidate schemes. S417: Merge the solution sets of feasible candidate solutions and the solution sets of certified candidate solutions to form a multi-objective optimization solution set.

6. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, In step S4, the inner loop's working steps include: S421: Iteratively solve the minimum cycle time and optimal requirement envelope of candidate solutions using a trajectory-driven performance quantization interface function; S422: Perform driver matching in a compact set of candidate drivers based on the optimal requirement envelope to obtain the optimal driver selection.

7. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, The performance quantization interface function for trajectory-driven operation is: in, For trajectory-driven performance quantization interface, For periodic time, for The torque during the cycle of time, for The rotational speed during the time cycle, For the demand envelope, For scale structure decision variables, For trajectory planning related parameters, This is the output.

8. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, The optimization variables include scale variables, structural variables, and drive selection variables. The scale variables are used to characterize the geometric dimensions of the link length. The structural variables are used to characterize key cross-sectional dimensions and plate thickness; The drive selection variables are used to characterize the selection scheme of motor-reducer; The constraints include structural constraints, driving capability constraints, and engineering constraints. The multi-objective optimization function aims at minimizing cycle time, overall machine mass, and end-effector deformation.

9. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, The parametric performance modeling module in step S3 includes rod inertia parameter mapping, kinematic model, and dynamic model.

10. The scale-structure-drive collaborative optimization design method for a high-speed heavy-duty robot according to claim 1, characterized in that, The demand envelope includes the maximum operating torque, the root mean square value of the torque, and the maximum speed within one operating cycle.

Images