CN-121763709-B - Dynamic optimization method and system for stacking robot running path

CN121763709BCN 121763709 BCN121763709 BCN 121763709BCN-121763709-B

Abstract

The application provides a dynamic optimization method and a system for a stacking robot running path, which relate to the technical field of path optimization, and the method comprises the steps of acquiring a dynamic semantic map carrying goods space priority and path traffic class and real-time running data containing real-time loads and multi-task sequences, and analyzing to obtain the association relation of running energy consumption, path steering and real-time loads; the method comprises the steps of carrying out multi-objective optimization by adopting an improved genetic algorithm and combining goods space priority and path traffic level to generate a path optimal solution set, then constructing a moving track model by using a time elastic band algorithm by taking the path optimal solution set as input, carrying out online re-planning by combining dynamic updating to obtain a reference track and a speed profile, and finally tracking and generating track control parameters by a PID controller to drive a stacking robot to execute tasks. The application comprehensively optimizes the operation efficiency, stability and execution precision of the stacking robot in a dynamic environment.

Inventors

ZHENG XIAOSHENG
KANG SHE
SHAN CHAOLAN
MA FANG
ZHU YING
LIU YING
YIN GUOPING

Assignees

北京东方国凯工业装备有限公司

Dates

Publication Date: 20260508
Application Date: 20260305

Claims (10)

1. A method for dynamically optimizing a travel path of a palletizing robot, comprising: acquiring a dynamic semantic map and real-time operation data of the stacking robot, wherein the dynamic semantic map is marked with goods position priority and path traffic level, and the real-time operation data comprises real-time load data and a multitasking sequence; Performing coupling analysis processing on the dynamic semantic map and the real-time operation data to obtain an association relationship of operation energy consumption, path steering and real-time load of the stacking robot; Based on the association relation, the goods space priority, the path passing grade and the multi-task sequence, adopting an improved genetic algorithm to carry out multi-objective optimization to generate a path optimal solution set, wherein the improved genetic algorithm comprises a screening stage based on load adaptation degree, an elite retaining stage based on energy consumption estimation and distance estimation and a diversity supplementing stage based on task waiting time, the load adaptation degree is a score between 0 and 100, and the higher the score is the safer the path curve and gradient of the current load passing through; Taking the path optimal solution set as input, and adopting a time elastic band algorithm to construct a space-time track under the condition of meeting multiple constraint conditions to obtain a moving track model; Combining the increment updating of the dynamic semantic map and the change of the real-time load data, and carrying out on-line re-planning and space-time constraint optimization processing on the moving track model to obtain a reference track and a corresponding speed profile of the stacking robot; And tracking control is carried out through a PID controller according to the reference track and the speed profile so as to generate track control parameters, and the stacking robot is controlled to execute the multitasking sequence according to the track control parameters.
2. The method of claim 1, wherein the performing multi-objective optimization with a modified genetic algorithm based on the association, the cargo space priority, the path traffic class, and the multi-task sequence to generate a path optimal solution set, wherein the modified genetic algorithm includes three stages including a screening stage based on load adaptation, an elite retention stage based on energy consumption estimation and distance estimation, and a diversity replenishment stage based on task waiting time, comprising: Taking each goods space coordinate in the multitasking sequence as a node to form an initial path population; based on the association relation, calculating the energy consumption estimated value, the distance estimated value and the load adaptation degree corresponding to each individual in the initial path population, and simultaneously giving priority weights to tasks according to the goods space priority and giving passing weights to paths according to the path passing grades; Eliminating individuals with the load adaptation degree lower than a preset load threshold value or the sum of the passing weights of the paths passing through being greater than a preset weight upper limit in the initial path population through a screening strategy of an improved genetic algorithm to obtain a first intermediate population; Sorting the individuals in the first intermediate population by a non-dominant sorting method by taking the sum of the corresponding energy consumption estimated value, the distance estimated value and the passing weight of the passed path as an optimization target through an elite retention strategy of an improved genetic algorithm so as to retain the individuals in a first non-dominant level, obtain a second intermediate population, and take the second intermediate population as elite offspring; According to a guiding strategy of an improved genetic algorithm, when the elite offspring does not meet a preset convergence condition, calculating task waiting time of the first intermediate population for removing individuals of the second intermediate population according to the priority weight, and selecting a preset number of individuals with minimum task waiting time to obtain diversified offspring; combining the elite offspring with the diversity offspring through an optimization strategy of an improved genetic algorithm to obtain an effective population, and performing self-adaptive crossover operation and local mutation operation on the effective population to obtain a middle path population; and repeatedly executing the operation of calculating and estimating to the local variation on the intermediate path population until elite offspring meet a preset convergence condition, and taking the elite offspring of the last generation as a path optimal solution set.
3. The method of claim 2, wherein the step of calculating the task waiting time of the first intermediate population to remove individuals of the second intermediate population according to the priority weight when the elite offspring do not meet a preset convergence condition by a guidance strategy of a modified genetic algorithm comprises: acquiring an operation sequence of removing individuals of the second intermediate population in the first intermediate population; Performing time sequence conflict analysis by combining the operation sequence, the path traffic level and the real-time load data to obtain conflict pairs with overlapped execution time; aiming at the conflict pair, according to the priority weights of the conflict parties, a penalty function method is applied to carry out conflict delay calculation to obtain single delay time, wherein the single delay is generated from the task with the small priority weight in the conflict parties to the task with the large priority weight; based on the topological relation of the operation sequence, the single delay time is transmitted backwards, delay propagation is executed, and accumulated delay time is obtained; and carrying out weighted summation on all the accumulated delay time to obtain the task waiting time of the first intermediate population for removing the individuals of the second intermediate population.
4. The method of claim 1, wherein the performing space-time trajectory construction using a time elastic band algorithm with the path optimal solution set as an input under the condition of meeting multiple constraints to obtain a movement trajectory model includes: Performing time stamp distribution on the path optimal solution set to generate an initial path sequence; performing spatial interpolation on the initial path sequence by adopting a cubic spline interpolation algorithm to obtain a dense sequence; constructing a track optimization problem by taking the dense sequence as an initial solution, wherein constraint conditions of the track optimization problem comprise path curvature constraint, acceleration constraint, path feasibility constraint and load dynamics constraint; Solving the track optimization problem through a time elastic band algorithm to obtain an optimization solution, and optimizing the space positions and time stamps of all path points in the dense sequence based on the optimization solution to generate an optimized path sequence; Performing dynamic feasibility verification on the optimized path sequence, and performing relaxation adjustment on path points which do not meet the path feasibility constraint or the load dynamics constraint in verification to obtain a target path sequence; And packaging the target path sequence and each constraint item to obtain a movement track model.
5. The method of claim 4, wherein the dynamically verifying the optimized path sequence and performing slack adjustment on path points in the verification that do not satisfy the path feasibility constraint or the load dynamics constraint, to obtain a target path sequence, comprises: constructing a cost evaluation model through a graph neural network based on the path traffic grade and the real-time load data; Calculating the traffic state and the load state of each path point in the optimized path sequence through the cost evaluation model; Comparing the passing state of each path point with a preset passing state threshold value, and comparing the load state of each path point with a preset load state threshold value to identify the path points of which the passing state and the load state exceed the corresponding threshold values respectively, so as to obtain an adjustment point set; Aiming at each path point in the set of the adjustment points, performing neighborhood exploration to obtain a space-time neighborhood corresponding to each path point, and performing sampling processing meeting preset connection conditions in the space-time neighborhood to generate a plurality of alternative candidate points; executing decision evaluation processing on each of the alternative candidate points to obtain a comprehensive evaluation value; Selecting an optimal candidate point from the replacement candidate points according to the comprehensive evaluation value, and replacing the corresponding path point in the set of the adjustment points with the optimal candidate point to obtain a final point set; and performing iterative optimization operation of curvature continuity and speed continuity on the final point set by using a gradient descent method to obtain a target path sequence.
6. The method of claim 1, wherein the combining the incremental update of the dynamic semantic map with the change of the real-time load data performs online re-planning and space-time constraint optimization on the movement trajectory model to obtain a reference trajectory and a corresponding speed profile of the palletizing robot, comprising: Performing space conflict detection on the incremental updating information of the dynamic semantic map and the path point sequence in the moving track model to obtain a space conflict area, and performing mutation analysis on the time sequence change of the real-time load data to obtain an abnormal period; positioning a local track segment to be adjusted in a path point sequence of the moving track model based on the space conflict area and the abnormal time period; Constructing a space-time feasible region for the local track segment according to the traffic level in the incremental updating information and the real-time load data; Carrying out random sampling processing meeting preset starting and stopping states in the space-time feasible domain to obtain a plurality of alternative sequences; Carrying out solving and evaluating treatment on the alternative sequence by using a track optimization algorithm and taking constraint parameters in the moving track model as conditions to obtain candidate track segments; Splicing the candidate track segments with a target path sequence in the moving track model through a track splicing algorithm, and smoothing connection points to obtain a global optimized path; and executing kinematic interpolation processing on the global optimization path to generate a reference track of the stacking robot, and executing first-order differential processing on the reference track to generate a speed profile.
7. The method of claim 1, wherein tracking control by a PID controller based on the reference trajectory and the velocity profile to generate trajectory control parameters comprises: Performing error distribution processing on the reference track, the speed profile and the running state data to obtain a multidimensional error vector; the multidimensional error vector is respectively input into a first control channel, a second control channel and a feedforward compensation channel in the PID controller; performing pose tracking processing through the first control channel to generate basic control quantity of proportion, integration and differentiation, and performing speed tracking processing through the second control channel to generate speed control quantity; performing disturbance observation processing according to the real-time load data and the path traffic class through the feedforward compensation channel to generate feedforward compensation quantity; Inputting the basic control quantity, the speed control quantity and the feedforward compensation quantity to an instruction fusion unit of a PID controller to perform weighted fusion so as to generate a composite control instruction; Performing boundary adaptation and instruction smoothing on the curvature information of the composite control instruction and the reference track by using an instruction limiter in the PID controller so as to generate a modulated control instruction; and converting the modulated control instruction into a pulse width modulation signal to obtain a track control parameter.
8. A system for dynamic optimization of a palletizing robot travel path, comprising: The system comprises an acquisition module, a stacking module and a storage module, wherein the acquisition module is used for acquiring a dynamic semantic map and real-time operation data of the stacking robot, wherein the dynamic semantic map is marked with goods position priority and path passing grade, and the real-time operation data comprises real-time load data and a multi-task sequence; the analysis module is used for carrying out coupling analysis processing on the dynamic semantic map and the real-time operation data to obtain the association relation of the operation energy consumption, the path steering and the real-time load of the stacking robot; The first optimization module is used for performing multi-objective optimization by adopting an improved genetic algorithm based on the association relation, the goods space priority, the path passing grade and the multi-task sequence to generate a path optimal solution set, wherein the improved genetic algorithm comprises three stages, namely a screening stage based on load adaptation degree, an elite retaining stage based on energy consumption estimation and distance estimation and a diversity supplementing stage based on task waiting time, the load adaptation degree is a score between 0 and 100, and the higher the score is the safer the path curve and gradient of the current load passing; the construction module is used for taking the path optimal solution set as input, and adopting a time elastic band algorithm to carry out space-time track construction processing under the condition of meeting multiple constraint conditions so as to obtain a movement track model; The second optimization module is used for carrying out online re-planning and space-time constraint optimization processing on the moving track model by combining the increment updating of the dynamic semantic map and the change of the real-time load data so as to obtain a reference track and a corresponding speed profile of the stacking robot; And the control module is used for tracking control through a PID controller according to the reference track and the speed profile so as to generate track control parameters and controlling the stacking robot to execute the multi-task sequence according to the track control parameters.
9. An electronic device, comprising: A memory for storing a computer program; a processor for implementing the steps of the method for dynamic optimization of a palletizing robot path according to any of claims 1 to 7 when executing said computer program.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when executed by a processor, enables a method of dynamic optimization of the stacking robot path of any one of claims 1 to 7.

Description

Dynamic optimization method and system for stacking robot running path Technical Field The application relates to the technical field of path optimization, in particular to a dynamic optimization method and a system for a stacking robot running path. Background The optimization of the stacking robot running path is a core link of efficient operation of the automatic stereoscopic warehouse, and aims to improve the overall running efficiency and economic benefit of a warehouse system, and the effective path planning can reduce equipment idle load, reduce energy consumption, flexibly respond to dynamically-changed warehouse tasks and have wide application value. At present, path optimization of stacking robots is mostly dependent on a scheduling strategy based on preset rules, or a path planning method based on a single index, such as a shortest distance, and in addition, a method of adjusting an operation route through manual experience intervention or simple environment judgment exists. However, it is generally difficult to coordinate and run multiple key factors such as energy consumption, cargo processing sequence, real-time state of path, and load change, which often lacks full utilization of dynamic environment information, and fails to effectively integrate and synergistically optimize the above factors, so that adaptability and economy of the planning result in a practical complex operation scene are poor. Disclosure of Invention The application aims to provide a dynamic optimization method and a system for a stacking robot running path, which are used for solving the problem that the running efficiency, the energy consumption economy and the environmental adaptability are difficult to realize collaborative optimization in a dynamic working environment in the prior art. In order to solve the above technical problems, in a first aspect, the present application provides a method for dynamically optimizing a stacking robot running path, including: acquiring a dynamic semantic map and real-time operation data of the stacking robot, wherein the dynamic semantic map is marked with goods position priority and path traffic level, and the real-time operation data comprises real-time load data and a multitasking sequence; Performing coupling analysis processing on the dynamic semantic map and the real-time operation data to obtain an association relationship of operation energy consumption, path steering and real-time load of the stacking robot; Based on the association relation, the goods space priority, the path passing grade and the multi-task sequence, adopting an improved genetic algorithm to carry out multi-objective optimization to generate a path optimal solution set, wherein the improved genetic algorithm comprises a screening stage based on load adaptation degree, an elite retaining stage based on energy consumption estimation and distance estimation and a diversity supplementing stage based on task waiting time; Taking the path optimal solution set as input, and adopting a time elastic band algorithm to construct a space-time track under the condition of meeting multiple constraint conditions to obtain a moving track model; Combining the increment updating of the dynamic semantic map and the change of the real-time load data, and carrying out on-line re-planning and space-time constraint optimization processing on the moving track model to obtain a reference track and a corresponding speed profile of the stacking robot; And tracking control is carried out through a PID controller according to the reference track and the speed profile so as to generate track control parameters, and the stacking robot is controlled to execute the multitasking sequence according to the track control parameters. Optionally, the multi-objective optimization is performed by adopting an improved genetic algorithm based on the association relation, the goods space priority, the path passing level and the multi-task sequence to generate a path optimal solution set, wherein the improved genetic algorithm comprises a screening stage based on load adaptation degree, an elite retaining stage based on energy consumption estimation and distance estimation and a diversity supplementing stage based on task waiting time, and the method comprises the following three stages: Taking each goods space coordinate in the multitasking sequence as a node to form an initial path population; based on the association relation, calculating the energy consumption estimated value, the distance estimated value and the load adaptation degree corresponding to each individual in the initial path population, and simultaneously giving priority weights to tasks according to the goods space priority and giving passing weights to paths according to the path passing grades; Eliminating individuals with the load adaptation degree lower than a preset load threshold value or the sum of the passing weights of the paths passing through being greater than a preset w